1 /* 2 * linux/fs/ext4/super.c 3 * 4 * Copyright (C) 1992, 1993, 1994, 1995 5 * Remy Card (card@masi.ibp.fr) 6 * Laboratoire MASI - Institut Blaise Pascal 7 * Universite Pierre et Marie Curie (Paris VI) 8 * 9 * from 10 * 11 * linux/fs/minix/inode.c 12 * 13 * Copyright (C) 1991, 1992 Linus Torvalds 14 * 15 * Big-endian to little-endian byte-swapping/bitmaps by 16 * David S. Miller (davem@caip.rutgers.edu), 1995 17 */ 18 19 #include <linux/module.h> 20 #include <linux/string.h> 21 #include <linux/fs.h> 22 #include <linux/time.h> 23 #include <linux/vmalloc.h> 24 #include <linux/slab.h> 25 #include <linux/init.h> 26 #include <linux/blkdev.h> 27 #include <linux/backing-dev.h> 28 #include <linux/parser.h> 29 #include <linux/buffer_head.h> 30 #include <linux/exportfs.h> 31 #include <linux/vfs.h> 32 #include <linux/random.h> 33 #include <linux/mount.h> 34 #include <linux/namei.h> 35 #include <linux/quotaops.h> 36 #include <linux/seq_file.h> 37 #include <linux/ctype.h> 38 #include <linux/log2.h> 39 #include <linux/crc16.h> 40 #include <linux/cleancache.h> 41 #include <asm/uaccess.h> 42 43 #include <linux/kthread.h> 44 #include <linux/freezer.h> 45 46 #include "ext4.h" 47 #include "ext4_extents.h" /* Needed for trace points definition */ 48 #include "ext4_jbd2.h" 49 #include "xattr.h" 50 #include "acl.h" 51 #include "mballoc.h" 52 53 #define CREATE_TRACE_POINTS 54 #include <trace/events/ext4.h> 55 56 static struct ext4_lazy_init *ext4_li_info; 57 static struct mutex ext4_li_mtx; 58 static struct ratelimit_state ext4_mount_msg_ratelimit; 59 60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 61 unsigned long journal_devnum); 62 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 63 static int ext4_commit_super(struct super_block *sb, int sync); 64 static void ext4_mark_recovery_complete(struct super_block *sb, 65 struct ext4_super_block *es); 66 static void ext4_clear_journal_err(struct super_block *sb, 67 struct ext4_super_block *es); 68 static int ext4_sync_fs(struct super_block *sb, int wait); 69 static int ext4_remount(struct super_block *sb, int *flags, char *data); 70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 71 static int ext4_unfreeze(struct super_block *sb); 72 static int ext4_freeze(struct super_block *sb); 73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 74 const char *dev_name, void *data); 75 static inline int ext2_feature_set_ok(struct super_block *sb); 76 static inline int ext3_feature_set_ok(struct super_block *sb); 77 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 78 static void ext4_destroy_lazyinit_thread(void); 79 static void ext4_unregister_li_request(struct super_block *sb); 80 static void ext4_clear_request_list(void); 81 static struct inode *ext4_get_journal_inode(struct super_block *sb, 82 unsigned int journal_inum); 83 84 /* 85 * Lock ordering 86 * 87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 88 * i_mmap_rwsem (inode->i_mmap_rwsem)! 89 * 90 * page fault path: 91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 92 * page lock -> i_data_sem (rw) 93 * 94 * buffered write path: 95 * sb_start_write -> i_mutex -> mmap_sem 96 * sb_start_write -> i_mutex -> transaction start -> page lock -> 97 * i_data_sem (rw) 98 * 99 * truncate: 100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) -> 101 * i_mmap_rwsem (w) -> page lock 102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) -> 103 * transaction start -> i_data_sem (rw) 104 * 105 * direct IO: 106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem 107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> 108 * transaction start -> i_data_sem (rw) 109 * 110 * writepages: 111 * transaction start -> page lock(s) -> i_data_sem (rw) 112 */ 113 114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 115 static struct file_system_type ext2_fs_type = { 116 .owner = THIS_MODULE, 117 .name = "ext2", 118 .mount = ext4_mount, 119 .kill_sb = kill_block_super, 120 .fs_flags = FS_REQUIRES_DEV, 121 }; 122 MODULE_ALIAS_FS("ext2"); 123 MODULE_ALIAS("ext2"); 124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 125 #else 126 #define IS_EXT2_SB(sb) (0) 127 #endif 128 129 130 static struct file_system_type ext3_fs_type = { 131 .owner = THIS_MODULE, 132 .name = "ext3", 133 .mount = ext4_mount, 134 .kill_sb = kill_block_super, 135 .fs_flags = FS_REQUIRES_DEV, 136 }; 137 MODULE_ALIAS_FS("ext3"); 138 MODULE_ALIAS("ext3"); 139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 140 141 static int ext4_verify_csum_type(struct super_block *sb, 142 struct ext4_super_block *es) 143 { 144 if (!ext4_has_feature_metadata_csum(sb)) 145 return 1; 146 147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 148 } 149 150 static __le32 ext4_superblock_csum(struct super_block *sb, 151 struct ext4_super_block *es) 152 { 153 struct ext4_sb_info *sbi = EXT4_SB(sb); 154 int offset = offsetof(struct ext4_super_block, s_checksum); 155 __u32 csum; 156 157 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 158 159 return cpu_to_le32(csum); 160 } 161 162 static int ext4_superblock_csum_verify(struct super_block *sb, 163 struct ext4_super_block *es) 164 { 165 if (!ext4_has_metadata_csum(sb)) 166 return 1; 167 168 return es->s_checksum == ext4_superblock_csum(sb, es); 169 } 170 171 void ext4_superblock_csum_set(struct super_block *sb) 172 { 173 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 174 175 if (!ext4_has_metadata_csum(sb)) 176 return; 177 178 es->s_checksum = ext4_superblock_csum(sb, es); 179 } 180 181 void *ext4_kvmalloc(size_t size, gfp_t flags) 182 { 183 void *ret; 184 185 ret = kmalloc(size, flags | __GFP_NOWARN); 186 if (!ret) 187 ret = __vmalloc(size, flags, PAGE_KERNEL); 188 return ret; 189 } 190 191 void *ext4_kvzalloc(size_t size, gfp_t flags) 192 { 193 void *ret; 194 195 ret = kzalloc(size, flags | __GFP_NOWARN); 196 if (!ret) 197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 198 return ret; 199 } 200 201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 202 struct ext4_group_desc *bg) 203 { 204 return le32_to_cpu(bg->bg_block_bitmap_lo) | 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 207 } 208 209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 210 struct ext4_group_desc *bg) 211 { 212 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 215 } 216 217 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 218 struct ext4_group_desc *bg) 219 { 220 return le32_to_cpu(bg->bg_inode_table_lo) | 221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 223 } 224 225 __u32 ext4_free_group_clusters(struct super_block *sb, 226 struct ext4_group_desc *bg) 227 { 228 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 231 } 232 233 __u32 ext4_free_inodes_count(struct super_block *sb, 234 struct ext4_group_desc *bg) 235 { 236 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 239 } 240 241 __u32 ext4_used_dirs_count(struct super_block *sb, 242 struct ext4_group_desc *bg) 243 { 244 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 247 } 248 249 __u32 ext4_itable_unused_count(struct super_block *sb, 250 struct ext4_group_desc *bg) 251 { 252 return le16_to_cpu(bg->bg_itable_unused_lo) | 253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 255 } 256 257 void ext4_block_bitmap_set(struct super_block *sb, 258 struct ext4_group_desc *bg, ext4_fsblk_t blk) 259 { 260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 263 } 264 265 void ext4_inode_bitmap_set(struct super_block *sb, 266 struct ext4_group_desc *bg, ext4_fsblk_t blk) 267 { 268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 271 } 272 273 void ext4_inode_table_set(struct super_block *sb, 274 struct ext4_group_desc *bg, ext4_fsblk_t blk) 275 { 276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 279 } 280 281 void ext4_free_group_clusters_set(struct super_block *sb, 282 struct ext4_group_desc *bg, __u32 count) 283 { 284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 287 } 288 289 void ext4_free_inodes_set(struct super_block *sb, 290 struct ext4_group_desc *bg, __u32 count) 291 { 292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 295 } 296 297 void ext4_used_dirs_set(struct super_block *sb, 298 struct ext4_group_desc *bg, __u32 count) 299 { 300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 303 } 304 305 void ext4_itable_unused_set(struct super_block *sb, 306 struct ext4_group_desc *bg, __u32 count) 307 { 308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 311 } 312 313 314 static void __save_error_info(struct super_block *sb, const char *func, 315 unsigned int line) 316 { 317 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 318 319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 320 if (bdev_read_only(sb->s_bdev)) 321 return; 322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 323 es->s_last_error_time = cpu_to_le32(get_seconds()); 324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 325 es->s_last_error_line = cpu_to_le32(line); 326 if (!es->s_first_error_time) { 327 es->s_first_error_time = es->s_last_error_time; 328 strncpy(es->s_first_error_func, func, 329 sizeof(es->s_first_error_func)); 330 es->s_first_error_line = cpu_to_le32(line); 331 es->s_first_error_ino = es->s_last_error_ino; 332 es->s_first_error_block = es->s_last_error_block; 333 } 334 /* 335 * Start the daily error reporting function if it hasn't been 336 * started already 337 */ 338 if (!es->s_error_count) 339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 340 le32_add_cpu(&es->s_error_count, 1); 341 } 342 343 static void save_error_info(struct super_block *sb, const char *func, 344 unsigned int line) 345 { 346 __save_error_info(sb, func, line); 347 ext4_commit_super(sb, 1); 348 } 349 350 /* 351 * The del_gendisk() function uninitializes the disk-specific data 352 * structures, including the bdi structure, without telling anyone 353 * else. Once this happens, any attempt to call mark_buffer_dirty() 354 * (for example, by ext4_commit_super), will cause a kernel OOPS. 355 * This is a kludge to prevent these oops until we can put in a proper 356 * hook in del_gendisk() to inform the VFS and file system layers. 357 */ 358 static int block_device_ejected(struct super_block *sb) 359 { 360 struct inode *bd_inode = sb->s_bdev->bd_inode; 361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 362 363 return bdi->dev == NULL; 364 } 365 366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 367 { 368 struct super_block *sb = journal->j_private; 369 struct ext4_sb_info *sbi = EXT4_SB(sb); 370 int error = is_journal_aborted(journal); 371 struct ext4_journal_cb_entry *jce; 372 373 BUG_ON(txn->t_state == T_FINISHED); 374 spin_lock(&sbi->s_md_lock); 375 while (!list_empty(&txn->t_private_list)) { 376 jce = list_entry(txn->t_private_list.next, 377 struct ext4_journal_cb_entry, jce_list); 378 list_del_init(&jce->jce_list); 379 spin_unlock(&sbi->s_md_lock); 380 jce->jce_func(sb, jce, error); 381 spin_lock(&sbi->s_md_lock); 382 } 383 spin_unlock(&sbi->s_md_lock); 384 } 385 386 /* Deal with the reporting of failure conditions on a filesystem such as 387 * inconsistencies detected or read IO failures. 388 * 389 * On ext2, we can store the error state of the filesystem in the 390 * superblock. That is not possible on ext4, because we may have other 391 * write ordering constraints on the superblock which prevent us from 392 * writing it out straight away; and given that the journal is about to 393 * be aborted, we can't rely on the current, or future, transactions to 394 * write out the superblock safely. 395 * 396 * We'll just use the jbd2_journal_abort() error code to record an error in 397 * the journal instead. On recovery, the journal will complain about 398 * that error until we've noted it down and cleared it. 399 */ 400 401 static void ext4_handle_error(struct super_block *sb) 402 { 403 if (sb->s_flags & MS_RDONLY) 404 return; 405 406 if (!test_opt(sb, ERRORS_CONT)) { 407 journal_t *journal = EXT4_SB(sb)->s_journal; 408 409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 410 if (journal) 411 jbd2_journal_abort(journal, -EIO); 412 } 413 if (test_opt(sb, ERRORS_RO)) { 414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 415 /* 416 * Make sure updated value of ->s_mount_flags will be visible 417 * before ->s_flags update 418 */ 419 smp_wmb(); 420 sb->s_flags |= MS_RDONLY; 421 } 422 if (test_opt(sb, ERRORS_PANIC)) { 423 if (EXT4_SB(sb)->s_journal && 424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 425 return; 426 panic("EXT4-fs (device %s): panic forced after error\n", 427 sb->s_id); 428 } 429 } 430 431 #define ext4_error_ratelimit(sb) \ 432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 433 "EXT4-fs error") 434 435 void __ext4_error(struct super_block *sb, const char *function, 436 unsigned int line, const char *fmt, ...) 437 { 438 struct va_format vaf; 439 va_list args; 440 441 if (ext4_error_ratelimit(sb)) { 442 va_start(args, fmt); 443 vaf.fmt = fmt; 444 vaf.va = &args; 445 printk(KERN_CRIT 446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 447 sb->s_id, function, line, current->comm, &vaf); 448 va_end(args); 449 } 450 save_error_info(sb, function, line); 451 ext4_handle_error(sb); 452 } 453 454 void __ext4_error_inode(struct inode *inode, const char *function, 455 unsigned int line, ext4_fsblk_t block, 456 const char *fmt, ...) 457 { 458 va_list args; 459 struct va_format vaf; 460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 461 462 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 463 es->s_last_error_block = cpu_to_le64(block); 464 if (ext4_error_ratelimit(inode->i_sb)) { 465 va_start(args, fmt); 466 vaf.fmt = fmt; 467 vaf.va = &args; 468 if (block) 469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 470 "inode #%lu: block %llu: comm %s: %pV\n", 471 inode->i_sb->s_id, function, line, inode->i_ino, 472 block, current->comm, &vaf); 473 else 474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 475 "inode #%lu: comm %s: %pV\n", 476 inode->i_sb->s_id, function, line, inode->i_ino, 477 current->comm, &vaf); 478 va_end(args); 479 } 480 save_error_info(inode->i_sb, function, line); 481 ext4_handle_error(inode->i_sb); 482 } 483 484 void __ext4_error_file(struct file *file, const char *function, 485 unsigned int line, ext4_fsblk_t block, 486 const char *fmt, ...) 487 { 488 va_list args; 489 struct va_format vaf; 490 struct ext4_super_block *es; 491 struct inode *inode = file_inode(file); 492 char pathname[80], *path; 493 494 es = EXT4_SB(inode->i_sb)->s_es; 495 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 496 if (ext4_error_ratelimit(inode->i_sb)) { 497 path = file_path(file, pathname, sizeof(pathname)); 498 if (IS_ERR(path)) 499 path = "(unknown)"; 500 va_start(args, fmt); 501 vaf.fmt = fmt; 502 vaf.va = &args; 503 if (block) 504 printk(KERN_CRIT 505 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 506 "block %llu: comm %s: path %s: %pV\n", 507 inode->i_sb->s_id, function, line, inode->i_ino, 508 block, current->comm, path, &vaf); 509 else 510 printk(KERN_CRIT 511 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 512 "comm %s: path %s: %pV\n", 513 inode->i_sb->s_id, function, line, inode->i_ino, 514 current->comm, path, &vaf); 515 va_end(args); 516 } 517 save_error_info(inode->i_sb, function, line); 518 ext4_handle_error(inode->i_sb); 519 } 520 521 const char *ext4_decode_error(struct super_block *sb, int errno, 522 char nbuf[16]) 523 { 524 char *errstr = NULL; 525 526 switch (errno) { 527 case -EFSCORRUPTED: 528 errstr = "Corrupt filesystem"; 529 break; 530 case -EFSBADCRC: 531 errstr = "Filesystem failed CRC"; 532 break; 533 case -EIO: 534 errstr = "IO failure"; 535 break; 536 case -ENOMEM: 537 errstr = "Out of memory"; 538 break; 539 case -EROFS: 540 if (!sb || (EXT4_SB(sb)->s_journal && 541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 542 errstr = "Journal has aborted"; 543 else 544 errstr = "Readonly filesystem"; 545 break; 546 default: 547 /* If the caller passed in an extra buffer for unknown 548 * errors, textualise them now. Else we just return 549 * NULL. */ 550 if (nbuf) { 551 /* Check for truncated error codes... */ 552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 553 errstr = nbuf; 554 } 555 break; 556 } 557 558 return errstr; 559 } 560 561 /* __ext4_std_error decodes expected errors from journaling functions 562 * automatically and invokes the appropriate error response. */ 563 564 void __ext4_std_error(struct super_block *sb, const char *function, 565 unsigned int line, int errno) 566 { 567 char nbuf[16]; 568 const char *errstr; 569 570 /* Special case: if the error is EROFS, and we're not already 571 * inside a transaction, then there's really no point in logging 572 * an error. */ 573 if (errno == -EROFS && journal_current_handle() == NULL && 574 (sb->s_flags & MS_RDONLY)) 575 return; 576 577 if (ext4_error_ratelimit(sb)) { 578 errstr = ext4_decode_error(sb, errno, nbuf); 579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 580 sb->s_id, function, line, errstr); 581 } 582 583 save_error_info(sb, function, line); 584 ext4_handle_error(sb); 585 } 586 587 /* 588 * ext4_abort is a much stronger failure handler than ext4_error. The 589 * abort function may be used to deal with unrecoverable failures such 590 * as journal IO errors or ENOMEM at a critical moment in log management. 591 * 592 * We unconditionally force the filesystem into an ABORT|READONLY state, 593 * unless the error response on the fs has been set to panic in which 594 * case we take the easy way out and panic immediately. 595 */ 596 597 void __ext4_abort(struct super_block *sb, const char *function, 598 unsigned int line, const char *fmt, ...) 599 { 600 struct va_format vaf; 601 va_list args; 602 603 save_error_info(sb, function, line); 604 va_start(args, fmt); 605 vaf.fmt = fmt; 606 vaf.va = &args; 607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 608 sb->s_id, function, line, &vaf); 609 va_end(args); 610 611 if ((sb->s_flags & MS_RDONLY) == 0) { 612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 614 /* 615 * Make sure updated value of ->s_mount_flags will be visible 616 * before ->s_flags update 617 */ 618 smp_wmb(); 619 sb->s_flags |= MS_RDONLY; 620 if (EXT4_SB(sb)->s_journal) 621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 622 save_error_info(sb, function, line); 623 } 624 if (test_opt(sb, ERRORS_PANIC)) { 625 if (EXT4_SB(sb)->s_journal && 626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 627 return; 628 panic("EXT4-fs panic from previous error\n"); 629 } 630 } 631 632 void __ext4_msg(struct super_block *sb, 633 const char *prefix, const char *fmt, ...) 634 { 635 struct va_format vaf; 636 va_list args; 637 638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 639 return; 640 641 va_start(args, fmt); 642 vaf.fmt = fmt; 643 vaf.va = &args; 644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 645 va_end(args); 646 } 647 648 #define ext4_warning_ratelimit(sb) \ 649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ 650 "EXT4-fs warning") 651 652 void __ext4_warning(struct super_block *sb, const char *function, 653 unsigned int line, const char *fmt, ...) 654 { 655 struct va_format vaf; 656 va_list args; 657 658 if (!ext4_warning_ratelimit(sb)) 659 return; 660 661 va_start(args, fmt); 662 vaf.fmt = fmt; 663 vaf.va = &args; 664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 665 sb->s_id, function, line, &vaf); 666 va_end(args); 667 } 668 669 void __ext4_warning_inode(const struct inode *inode, const char *function, 670 unsigned int line, const char *fmt, ...) 671 { 672 struct va_format vaf; 673 va_list args; 674 675 if (!ext4_warning_ratelimit(inode->i_sb)) 676 return; 677 678 va_start(args, fmt); 679 vaf.fmt = fmt; 680 vaf.va = &args; 681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 683 function, line, inode->i_ino, current->comm, &vaf); 684 va_end(args); 685 } 686 687 void __ext4_grp_locked_error(const char *function, unsigned int line, 688 struct super_block *sb, ext4_group_t grp, 689 unsigned long ino, ext4_fsblk_t block, 690 const char *fmt, ...) 691 __releases(bitlock) 692 __acquires(bitlock) 693 { 694 struct va_format vaf; 695 va_list args; 696 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 697 698 es->s_last_error_ino = cpu_to_le32(ino); 699 es->s_last_error_block = cpu_to_le64(block); 700 __save_error_info(sb, function, line); 701 702 if (ext4_error_ratelimit(sb)) { 703 va_start(args, fmt); 704 vaf.fmt = fmt; 705 vaf.va = &args; 706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 707 sb->s_id, function, line, grp); 708 if (ino) 709 printk(KERN_CONT "inode %lu: ", ino); 710 if (block) 711 printk(KERN_CONT "block %llu:", 712 (unsigned long long) block); 713 printk(KERN_CONT "%pV\n", &vaf); 714 va_end(args); 715 } 716 717 if (test_opt(sb, ERRORS_CONT)) { 718 ext4_commit_super(sb, 0); 719 return; 720 } 721 722 ext4_unlock_group(sb, grp); 723 ext4_handle_error(sb); 724 /* 725 * We only get here in the ERRORS_RO case; relocking the group 726 * may be dangerous, but nothing bad will happen since the 727 * filesystem will have already been marked read/only and the 728 * journal has been aborted. We return 1 as a hint to callers 729 * who might what to use the return value from 730 * ext4_grp_locked_error() to distinguish between the 731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 732 * aggressively from the ext4 function in question, with a 733 * more appropriate error code. 734 */ 735 ext4_lock_group(sb, grp); 736 return; 737 } 738 739 void ext4_update_dynamic_rev(struct super_block *sb) 740 { 741 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 742 743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 744 return; 745 746 ext4_warning(sb, 747 "updating to rev %d because of new feature flag, " 748 "running e2fsck is recommended", 749 EXT4_DYNAMIC_REV); 750 751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 754 /* leave es->s_feature_*compat flags alone */ 755 /* es->s_uuid will be set by e2fsck if empty */ 756 757 /* 758 * The rest of the superblock fields should be zero, and if not it 759 * means they are likely already in use, so leave them alone. We 760 * can leave it up to e2fsck to clean up any inconsistencies there. 761 */ 762 } 763 764 /* 765 * Open the external journal device 766 */ 767 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 768 { 769 struct block_device *bdev; 770 char b[BDEVNAME_SIZE]; 771 772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 773 if (IS_ERR(bdev)) 774 goto fail; 775 return bdev; 776 777 fail: 778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 779 __bdevname(dev, b), PTR_ERR(bdev)); 780 return NULL; 781 } 782 783 /* 784 * Release the journal device 785 */ 786 static void ext4_blkdev_put(struct block_device *bdev) 787 { 788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 789 } 790 791 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 792 { 793 struct block_device *bdev; 794 bdev = sbi->journal_bdev; 795 if (bdev) { 796 ext4_blkdev_put(bdev); 797 sbi->journal_bdev = NULL; 798 } 799 } 800 801 static inline struct inode *orphan_list_entry(struct list_head *l) 802 { 803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 804 } 805 806 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 807 { 808 struct list_head *l; 809 810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 811 le32_to_cpu(sbi->s_es->s_last_orphan)); 812 813 printk(KERN_ERR "sb_info orphan list:\n"); 814 list_for_each(l, &sbi->s_orphan) { 815 struct inode *inode = orphan_list_entry(l); 816 printk(KERN_ERR " " 817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 818 inode->i_sb->s_id, inode->i_ino, inode, 819 inode->i_mode, inode->i_nlink, 820 NEXT_ORPHAN(inode)); 821 } 822 } 823 824 static void ext4_put_super(struct super_block *sb) 825 { 826 struct ext4_sb_info *sbi = EXT4_SB(sb); 827 struct ext4_super_block *es = sbi->s_es; 828 int i, err; 829 830 ext4_unregister_li_request(sb); 831 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 832 833 flush_workqueue(sbi->rsv_conversion_wq); 834 destroy_workqueue(sbi->rsv_conversion_wq); 835 836 if (sbi->s_journal) { 837 err = jbd2_journal_destroy(sbi->s_journal); 838 sbi->s_journal = NULL; 839 if (err < 0) 840 ext4_abort(sb, "Couldn't clean up the journal"); 841 } 842 843 ext4_unregister_sysfs(sb); 844 ext4_es_unregister_shrinker(sbi); 845 del_timer_sync(&sbi->s_err_report); 846 ext4_release_system_zone(sb); 847 ext4_mb_release(sb); 848 ext4_ext_release(sb); 849 850 if (!(sb->s_flags & MS_RDONLY)) { 851 ext4_clear_feature_journal_needs_recovery(sb); 852 es->s_state = cpu_to_le16(sbi->s_mount_state); 853 } 854 if (!(sb->s_flags & MS_RDONLY)) 855 ext4_commit_super(sb, 1); 856 857 for (i = 0; i < sbi->s_gdb_count; i++) 858 brelse(sbi->s_group_desc[i]); 859 kvfree(sbi->s_group_desc); 860 kvfree(sbi->s_flex_groups); 861 percpu_counter_destroy(&sbi->s_freeclusters_counter); 862 percpu_counter_destroy(&sbi->s_freeinodes_counter); 863 percpu_counter_destroy(&sbi->s_dirs_counter); 864 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 865 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 866 brelse(sbi->s_sbh); 867 #ifdef CONFIG_QUOTA 868 for (i = 0; i < EXT4_MAXQUOTAS; i++) 869 kfree(sbi->s_qf_names[i]); 870 #endif 871 872 /* Debugging code just in case the in-memory inode orphan list 873 * isn't empty. The on-disk one can be non-empty if we've 874 * detected an error and taken the fs readonly, but the 875 * in-memory list had better be clean by this point. */ 876 if (!list_empty(&sbi->s_orphan)) 877 dump_orphan_list(sb, sbi); 878 J_ASSERT(list_empty(&sbi->s_orphan)); 879 880 sync_blockdev(sb->s_bdev); 881 invalidate_bdev(sb->s_bdev); 882 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 883 /* 884 * Invalidate the journal device's buffers. We don't want them 885 * floating about in memory - the physical journal device may 886 * hotswapped, and it breaks the `ro-after' testing code. 887 */ 888 sync_blockdev(sbi->journal_bdev); 889 invalidate_bdev(sbi->journal_bdev); 890 ext4_blkdev_remove(sbi); 891 } 892 if (sbi->s_mb_cache) { 893 ext4_xattr_destroy_cache(sbi->s_mb_cache); 894 sbi->s_mb_cache = NULL; 895 } 896 if (sbi->s_mmp_tsk) 897 kthread_stop(sbi->s_mmp_tsk); 898 sb->s_fs_info = NULL; 899 /* 900 * Now that we are completely done shutting down the 901 * superblock, we need to actually destroy the kobject. 902 */ 903 kobject_put(&sbi->s_kobj); 904 wait_for_completion(&sbi->s_kobj_unregister); 905 if (sbi->s_chksum_driver) 906 crypto_free_shash(sbi->s_chksum_driver); 907 kfree(sbi->s_blockgroup_lock); 908 kfree(sbi); 909 } 910 911 static struct kmem_cache *ext4_inode_cachep; 912 913 /* 914 * Called inside transaction, so use GFP_NOFS 915 */ 916 static struct inode *ext4_alloc_inode(struct super_block *sb) 917 { 918 struct ext4_inode_info *ei; 919 920 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 921 if (!ei) 922 return NULL; 923 924 ei->vfs_inode.i_version = 1; 925 spin_lock_init(&ei->i_raw_lock); 926 INIT_LIST_HEAD(&ei->i_prealloc_list); 927 spin_lock_init(&ei->i_prealloc_lock); 928 ext4_es_init_tree(&ei->i_es_tree); 929 rwlock_init(&ei->i_es_lock); 930 INIT_LIST_HEAD(&ei->i_es_list); 931 ei->i_es_all_nr = 0; 932 ei->i_es_shk_nr = 0; 933 ei->i_es_shrink_lblk = 0; 934 ei->i_reserved_data_blocks = 0; 935 ei->i_reserved_meta_blocks = 0; 936 ei->i_allocated_meta_blocks = 0; 937 ei->i_da_metadata_calc_len = 0; 938 ei->i_da_metadata_calc_last_lblock = 0; 939 spin_lock_init(&(ei->i_block_reservation_lock)); 940 #ifdef CONFIG_QUOTA 941 ei->i_reserved_quota = 0; 942 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 943 #endif 944 ei->jinode = NULL; 945 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 946 spin_lock_init(&ei->i_completed_io_lock); 947 ei->i_sync_tid = 0; 948 ei->i_datasync_tid = 0; 949 atomic_set(&ei->i_unwritten, 0); 950 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 951 return &ei->vfs_inode; 952 } 953 954 static int ext4_drop_inode(struct inode *inode) 955 { 956 int drop = generic_drop_inode(inode); 957 958 trace_ext4_drop_inode(inode, drop); 959 return drop; 960 } 961 962 static void ext4_i_callback(struct rcu_head *head) 963 { 964 struct inode *inode = container_of(head, struct inode, i_rcu); 965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 966 } 967 968 static void ext4_destroy_inode(struct inode *inode) 969 { 970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 971 ext4_msg(inode->i_sb, KERN_ERR, 972 "Inode %lu (%p): orphan list check failed!", 973 inode->i_ino, EXT4_I(inode)); 974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 975 EXT4_I(inode), sizeof(struct ext4_inode_info), 976 true); 977 dump_stack(); 978 } 979 call_rcu(&inode->i_rcu, ext4_i_callback); 980 } 981 982 static void init_once(void *foo) 983 { 984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 985 986 INIT_LIST_HEAD(&ei->i_orphan); 987 init_rwsem(&ei->xattr_sem); 988 init_rwsem(&ei->i_data_sem); 989 init_rwsem(&ei->i_mmap_sem); 990 inode_init_once(&ei->vfs_inode); 991 } 992 993 static int __init init_inodecache(void) 994 { 995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 996 sizeof(struct ext4_inode_info), 997 0, (SLAB_RECLAIM_ACCOUNT| 998 SLAB_MEM_SPREAD|SLAB_ACCOUNT), 999 init_once); 1000 if (ext4_inode_cachep == NULL) 1001 return -ENOMEM; 1002 return 0; 1003 } 1004 1005 static void destroy_inodecache(void) 1006 { 1007 /* 1008 * Make sure all delayed rcu free inodes are flushed before we 1009 * destroy cache. 1010 */ 1011 rcu_barrier(); 1012 kmem_cache_destroy(ext4_inode_cachep); 1013 } 1014 1015 void ext4_clear_inode(struct inode *inode) 1016 { 1017 invalidate_inode_buffers(inode); 1018 clear_inode(inode); 1019 dquot_drop(inode); 1020 ext4_discard_preallocations(inode); 1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1022 if (EXT4_I(inode)->jinode) { 1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1024 EXT4_I(inode)->jinode); 1025 jbd2_free_inode(EXT4_I(inode)->jinode); 1026 EXT4_I(inode)->jinode = NULL; 1027 } 1028 #ifdef CONFIG_EXT4_FS_ENCRYPTION 1029 fscrypt_put_encryption_info(inode, NULL); 1030 #endif 1031 } 1032 1033 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1034 u64 ino, u32 generation) 1035 { 1036 struct inode *inode; 1037 1038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 1039 return ERR_PTR(-ESTALE); 1040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 1041 return ERR_PTR(-ESTALE); 1042 1043 /* iget isn't really right if the inode is currently unallocated!! 1044 * 1045 * ext4_read_inode will return a bad_inode if the inode had been 1046 * deleted, so we should be safe. 1047 * 1048 * Currently we don't know the generation for parent directory, so 1049 * a generation of 0 means "accept any" 1050 */ 1051 inode = ext4_iget_normal(sb, ino); 1052 if (IS_ERR(inode)) 1053 return ERR_CAST(inode); 1054 if (generation && inode->i_generation != generation) { 1055 iput(inode); 1056 return ERR_PTR(-ESTALE); 1057 } 1058 1059 return inode; 1060 } 1061 1062 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1063 int fh_len, int fh_type) 1064 { 1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1066 ext4_nfs_get_inode); 1067 } 1068 1069 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1070 int fh_len, int fh_type) 1071 { 1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1073 ext4_nfs_get_inode); 1074 } 1075 1076 /* 1077 * Try to release metadata pages (indirect blocks, directories) which are 1078 * mapped via the block device. Since these pages could have journal heads 1079 * which would prevent try_to_free_buffers() from freeing them, we must use 1080 * jbd2 layer's try_to_free_buffers() function to release them. 1081 */ 1082 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1083 gfp_t wait) 1084 { 1085 journal_t *journal = EXT4_SB(sb)->s_journal; 1086 1087 WARN_ON(PageChecked(page)); 1088 if (!page_has_buffers(page)) 1089 return 0; 1090 if (journal) 1091 return jbd2_journal_try_to_free_buffers(journal, page, 1092 wait & ~__GFP_DIRECT_RECLAIM); 1093 return try_to_free_buffers(page); 1094 } 1095 1096 #ifdef CONFIG_EXT4_FS_ENCRYPTION 1097 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1098 { 1099 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1101 } 1102 1103 static int ext4_key_prefix(struct inode *inode, u8 **key) 1104 { 1105 *key = EXT4_SB(inode->i_sb)->key_prefix; 1106 return EXT4_SB(inode->i_sb)->key_prefix_size; 1107 } 1108 1109 static int ext4_prepare_context(struct inode *inode) 1110 { 1111 return ext4_convert_inline_data(inode); 1112 } 1113 1114 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1115 void *fs_data) 1116 { 1117 handle_t *handle; 1118 int res, res2; 1119 1120 /* fs_data is null when internally used. */ 1121 if (fs_data) { 1122 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1123 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, 1124 len, 0); 1125 if (!res) { 1126 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1127 ext4_clear_inode_state(inode, 1128 EXT4_STATE_MAY_INLINE_DATA); 1129 } 1130 return res; 1131 } 1132 1133 handle = ext4_journal_start(inode, EXT4_HT_MISC, 1134 ext4_jbd2_credits_xattr(inode)); 1135 if (IS_ERR(handle)) 1136 return PTR_ERR(handle); 1137 1138 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1139 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, 1140 len, 0); 1141 if (!res) { 1142 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1143 res = ext4_mark_inode_dirty(handle, inode); 1144 if (res) 1145 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1146 } 1147 res2 = ext4_journal_stop(handle); 1148 if (!res) 1149 res = res2; 1150 return res; 1151 } 1152 1153 static int ext4_dummy_context(struct inode *inode) 1154 { 1155 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); 1156 } 1157 1158 static unsigned ext4_max_namelen(struct inode *inode) 1159 { 1160 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize : 1161 EXT4_NAME_LEN; 1162 } 1163 1164 static struct fscrypt_operations ext4_cryptops = { 1165 .get_context = ext4_get_context, 1166 .key_prefix = ext4_key_prefix, 1167 .prepare_context = ext4_prepare_context, 1168 .set_context = ext4_set_context, 1169 .dummy_context = ext4_dummy_context, 1170 .is_encrypted = ext4_encrypted_inode, 1171 .empty_dir = ext4_empty_dir, 1172 .max_namelen = ext4_max_namelen, 1173 }; 1174 #else 1175 static struct fscrypt_operations ext4_cryptops = { 1176 .is_encrypted = ext4_encrypted_inode, 1177 }; 1178 #endif 1179 1180 #ifdef CONFIG_QUOTA 1181 static char *quotatypes[] = INITQFNAMES; 1182 #define QTYPE2NAME(t) (quotatypes[t]) 1183 1184 static int ext4_write_dquot(struct dquot *dquot); 1185 static int ext4_acquire_dquot(struct dquot *dquot); 1186 static int ext4_release_dquot(struct dquot *dquot); 1187 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1188 static int ext4_write_info(struct super_block *sb, int type); 1189 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1190 struct path *path); 1191 static int ext4_quota_off(struct super_block *sb, int type); 1192 static int ext4_quota_on_mount(struct super_block *sb, int type); 1193 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1194 size_t len, loff_t off); 1195 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1196 const char *data, size_t len, loff_t off); 1197 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1198 unsigned int flags); 1199 static int ext4_enable_quotas(struct super_block *sb); 1200 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); 1201 1202 static struct dquot **ext4_get_dquots(struct inode *inode) 1203 { 1204 return EXT4_I(inode)->i_dquot; 1205 } 1206 1207 static const struct dquot_operations ext4_quota_operations = { 1208 .get_reserved_space = ext4_get_reserved_space, 1209 .write_dquot = ext4_write_dquot, 1210 .acquire_dquot = ext4_acquire_dquot, 1211 .release_dquot = ext4_release_dquot, 1212 .mark_dirty = ext4_mark_dquot_dirty, 1213 .write_info = ext4_write_info, 1214 .alloc_dquot = dquot_alloc, 1215 .destroy_dquot = dquot_destroy, 1216 .get_projid = ext4_get_projid, 1217 .get_next_id = ext4_get_next_id, 1218 }; 1219 1220 static const struct quotactl_ops ext4_qctl_operations = { 1221 .quota_on = ext4_quota_on, 1222 .quota_off = ext4_quota_off, 1223 .quota_sync = dquot_quota_sync, 1224 .get_state = dquot_get_state, 1225 .set_info = dquot_set_dqinfo, 1226 .get_dqblk = dquot_get_dqblk, 1227 .set_dqblk = dquot_set_dqblk, 1228 .get_nextdqblk = dquot_get_next_dqblk, 1229 }; 1230 #endif 1231 1232 static const struct super_operations ext4_sops = { 1233 .alloc_inode = ext4_alloc_inode, 1234 .destroy_inode = ext4_destroy_inode, 1235 .write_inode = ext4_write_inode, 1236 .dirty_inode = ext4_dirty_inode, 1237 .drop_inode = ext4_drop_inode, 1238 .evict_inode = ext4_evict_inode, 1239 .put_super = ext4_put_super, 1240 .sync_fs = ext4_sync_fs, 1241 .freeze_fs = ext4_freeze, 1242 .unfreeze_fs = ext4_unfreeze, 1243 .statfs = ext4_statfs, 1244 .remount_fs = ext4_remount, 1245 .show_options = ext4_show_options, 1246 #ifdef CONFIG_QUOTA 1247 .quota_read = ext4_quota_read, 1248 .quota_write = ext4_quota_write, 1249 .get_dquots = ext4_get_dquots, 1250 #endif 1251 .bdev_try_to_free_page = bdev_try_to_free_page, 1252 }; 1253 1254 static const struct export_operations ext4_export_ops = { 1255 .fh_to_dentry = ext4_fh_to_dentry, 1256 .fh_to_parent = ext4_fh_to_parent, 1257 .get_parent = ext4_get_parent, 1258 }; 1259 1260 enum { 1261 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1262 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1263 Opt_nouid32, Opt_debug, Opt_removed, 1264 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1265 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1266 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1267 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1268 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1269 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1270 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1271 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1272 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1273 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, 1274 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 1275 Opt_lazytime, Opt_nolazytime, 1276 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1277 Opt_inode_readahead_blks, Opt_journal_ioprio, 1278 Opt_dioread_nolock, Opt_dioread_lock, 1279 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1280 Opt_max_dir_size_kb, Opt_nojournal_checksum, 1281 }; 1282 1283 static const match_table_t tokens = { 1284 {Opt_bsd_df, "bsddf"}, 1285 {Opt_minix_df, "minixdf"}, 1286 {Opt_grpid, "grpid"}, 1287 {Opt_grpid, "bsdgroups"}, 1288 {Opt_nogrpid, "nogrpid"}, 1289 {Opt_nogrpid, "sysvgroups"}, 1290 {Opt_resgid, "resgid=%u"}, 1291 {Opt_resuid, "resuid=%u"}, 1292 {Opt_sb, "sb=%u"}, 1293 {Opt_err_cont, "errors=continue"}, 1294 {Opt_err_panic, "errors=panic"}, 1295 {Opt_err_ro, "errors=remount-ro"}, 1296 {Opt_nouid32, "nouid32"}, 1297 {Opt_debug, "debug"}, 1298 {Opt_removed, "oldalloc"}, 1299 {Opt_removed, "orlov"}, 1300 {Opt_user_xattr, "user_xattr"}, 1301 {Opt_nouser_xattr, "nouser_xattr"}, 1302 {Opt_acl, "acl"}, 1303 {Opt_noacl, "noacl"}, 1304 {Opt_noload, "norecovery"}, 1305 {Opt_noload, "noload"}, 1306 {Opt_removed, "nobh"}, 1307 {Opt_removed, "bh"}, 1308 {Opt_commit, "commit=%u"}, 1309 {Opt_min_batch_time, "min_batch_time=%u"}, 1310 {Opt_max_batch_time, "max_batch_time=%u"}, 1311 {Opt_journal_dev, "journal_dev=%u"}, 1312 {Opt_journal_path, "journal_path=%s"}, 1313 {Opt_journal_checksum, "journal_checksum"}, 1314 {Opt_nojournal_checksum, "nojournal_checksum"}, 1315 {Opt_journal_async_commit, "journal_async_commit"}, 1316 {Opt_abort, "abort"}, 1317 {Opt_data_journal, "data=journal"}, 1318 {Opt_data_ordered, "data=ordered"}, 1319 {Opt_data_writeback, "data=writeback"}, 1320 {Opt_data_err_abort, "data_err=abort"}, 1321 {Opt_data_err_ignore, "data_err=ignore"}, 1322 {Opt_offusrjquota, "usrjquota="}, 1323 {Opt_usrjquota, "usrjquota=%s"}, 1324 {Opt_offgrpjquota, "grpjquota="}, 1325 {Opt_grpjquota, "grpjquota=%s"}, 1326 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1327 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1328 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1329 {Opt_grpquota, "grpquota"}, 1330 {Opt_noquota, "noquota"}, 1331 {Opt_quota, "quota"}, 1332 {Opt_usrquota, "usrquota"}, 1333 {Opt_prjquota, "prjquota"}, 1334 {Opt_barrier, "barrier=%u"}, 1335 {Opt_barrier, "barrier"}, 1336 {Opt_nobarrier, "nobarrier"}, 1337 {Opt_i_version, "i_version"}, 1338 {Opt_dax, "dax"}, 1339 {Opt_stripe, "stripe=%u"}, 1340 {Opt_delalloc, "delalloc"}, 1341 {Opt_lazytime, "lazytime"}, 1342 {Opt_nolazytime, "nolazytime"}, 1343 {Opt_nodelalloc, "nodelalloc"}, 1344 {Opt_removed, "mblk_io_submit"}, 1345 {Opt_removed, "nomblk_io_submit"}, 1346 {Opt_block_validity, "block_validity"}, 1347 {Opt_noblock_validity, "noblock_validity"}, 1348 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1349 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1350 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1351 {Opt_auto_da_alloc, "auto_da_alloc"}, 1352 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1353 {Opt_dioread_nolock, "dioread_nolock"}, 1354 {Opt_dioread_lock, "dioread_lock"}, 1355 {Opt_discard, "discard"}, 1356 {Opt_nodiscard, "nodiscard"}, 1357 {Opt_init_itable, "init_itable=%u"}, 1358 {Opt_init_itable, "init_itable"}, 1359 {Opt_noinit_itable, "noinit_itable"}, 1360 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1361 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1362 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1363 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1364 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1365 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1366 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1367 {Opt_err, NULL}, 1368 }; 1369 1370 static ext4_fsblk_t get_sb_block(void **data) 1371 { 1372 ext4_fsblk_t sb_block; 1373 char *options = (char *) *data; 1374 1375 if (!options || strncmp(options, "sb=", 3) != 0) 1376 return 1; /* Default location */ 1377 1378 options += 3; 1379 /* TODO: use simple_strtoll with >32bit ext4 */ 1380 sb_block = simple_strtoul(options, &options, 0); 1381 if (*options && *options != ',') { 1382 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1383 (char *) *data); 1384 return 1; 1385 } 1386 if (*options == ',') 1387 options++; 1388 *data = (void *) options; 1389 1390 return sb_block; 1391 } 1392 1393 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1394 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1395 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1396 1397 #ifdef CONFIG_QUOTA 1398 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1399 { 1400 struct ext4_sb_info *sbi = EXT4_SB(sb); 1401 char *qname; 1402 int ret = -1; 1403 1404 if (sb_any_quota_loaded(sb) && 1405 !sbi->s_qf_names[qtype]) { 1406 ext4_msg(sb, KERN_ERR, 1407 "Cannot change journaled " 1408 "quota options when quota turned on"); 1409 return -1; 1410 } 1411 if (ext4_has_feature_quota(sb)) { 1412 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1413 "ignored when QUOTA feature is enabled"); 1414 return 1; 1415 } 1416 qname = match_strdup(args); 1417 if (!qname) { 1418 ext4_msg(sb, KERN_ERR, 1419 "Not enough memory for storing quotafile name"); 1420 return -1; 1421 } 1422 if (sbi->s_qf_names[qtype]) { 1423 if (strcmp(sbi->s_qf_names[qtype], qname) == 0) 1424 ret = 1; 1425 else 1426 ext4_msg(sb, KERN_ERR, 1427 "%s quota file already specified", 1428 QTYPE2NAME(qtype)); 1429 goto errout; 1430 } 1431 if (strchr(qname, '/')) { 1432 ext4_msg(sb, KERN_ERR, 1433 "quotafile must be on filesystem root"); 1434 goto errout; 1435 } 1436 sbi->s_qf_names[qtype] = qname; 1437 set_opt(sb, QUOTA); 1438 return 1; 1439 errout: 1440 kfree(qname); 1441 return ret; 1442 } 1443 1444 static int clear_qf_name(struct super_block *sb, int qtype) 1445 { 1446 1447 struct ext4_sb_info *sbi = EXT4_SB(sb); 1448 1449 if (sb_any_quota_loaded(sb) && 1450 sbi->s_qf_names[qtype]) { 1451 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1452 " when quota turned on"); 1453 return -1; 1454 } 1455 kfree(sbi->s_qf_names[qtype]); 1456 sbi->s_qf_names[qtype] = NULL; 1457 return 1; 1458 } 1459 #endif 1460 1461 #define MOPT_SET 0x0001 1462 #define MOPT_CLEAR 0x0002 1463 #define MOPT_NOSUPPORT 0x0004 1464 #define MOPT_EXPLICIT 0x0008 1465 #define MOPT_CLEAR_ERR 0x0010 1466 #define MOPT_GTE0 0x0020 1467 #ifdef CONFIG_QUOTA 1468 #define MOPT_Q 0 1469 #define MOPT_QFMT 0x0040 1470 #else 1471 #define MOPT_Q MOPT_NOSUPPORT 1472 #define MOPT_QFMT MOPT_NOSUPPORT 1473 #endif 1474 #define MOPT_DATAJ 0x0080 1475 #define MOPT_NO_EXT2 0x0100 1476 #define MOPT_NO_EXT3 0x0200 1477 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1478 #define MOPT_STRING 0x0400 1479 1480 static const struct mount_opts { 1481 int token; 1482 int mount_opt; 1483 int flags; 1484 } ext4_mount_opts[] = { 1485 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1486 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1487 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1488 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1489 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1490 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1491 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1492 MOPT_EXT4_ONLY | MOPT_SET}, 1493 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1494 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1495 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1496 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1497 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1498 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1499 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1500 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1501 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1502 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1503 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1504 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1505 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1506 EXT4_MOUNT_JOURNAL_CHECKSUM), 1507 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1508 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1509 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1510 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1511 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1512 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1513 MOPT_NO_EXT2}, 1514 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1515 MOPT_NO_EXT2}, 1516 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1517 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1518 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1519 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1520 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1521 {Opt_commit, 0, MOPT_GTE0}, 1522 {Opt_max_batch_time, 0, MOPT_GTE0}, 1523 {Opt_min_batch_time, 0, MOPT_GTE0}, 1524 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1525 {Opt_init_itable, 0, MOPT_GTE0}, 1526 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 1527 {Opt_stripe, 0, MOPT_GTE0}, 1528 {Opt_resuid, 0, MOPT_GTE0}, 1529 {Opt_resgid, 0, MOPT_GTE0}, 1530 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1531 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1532 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1533 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1534 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1535 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1536 MOPT_NO_EXT2 | MOPT_DATAJ}, 1537 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1538 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1539 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1540 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1541 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1542 #else 1543 {Opt_acl, 0, MOPT_NOSUPPORT}, 1544 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1545 #endif 1546 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1547 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1548 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1549 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1550 MOPT_SET | MOPT_Q}, 1551 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1552 MOPT_SET | MOPT_Q}, 1553 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1554 MOPT_SET | MOPT_Q}, 1555 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1556 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1557 MOPT_CLEAR | MOPT_Q}, 1558 {Opt_usrjquota, 0, MOPT_Q}, 1559 {Opt_grpjquota, 0, MOPT_Q}, 1560 {Opt_offusrjquota, 0, MOPT_Q}, 1561 {Opt_offgrpjquota, 0, MOPT_Q}, 1562 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1563 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1564 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1565 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1566 {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1567 {Opt_err, 0, 0} 1568 }; 1569 1570 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1571 substring_t *args, unsigned long *journal_devnum, 1572 unsigned int *journal_ioprio, int is_remount) 1573 { 1574 struct ext4_sb_info *sbi = EXT4_SB(sb); 1575 const struct mount_opts *m; 1576 kuid_t uid; 1577 kgid_t gid; 1578 int arg = 0; 1579 1580 #ifdef CONFIG_QUOTA 1581 if (token == Opt_usrjquota) 1582 return set_qf_name(sb, USRQUOTA, &args[0]); 1583 else if (token == Opt_grpjquota) 1584 return set_qf_name(sb, GRPQUOTA, &args[0]); 1585 else if (token == Opt_offusrjquota) 1586 return clear_qf_name(sb, USRQUOTA); 1587 else if (token == Opt_offgrpjquota) 1588 return clear_qf_name(sb, GRPQUOTA); 1589 #endif 1590 switch (token) { 1591 case Opt_noacl: 1592 case Opt_nouser_xattr: 1593 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1594 break; 1595 case Opt_sb: 1596 return 1; /* handled by get_sb_block() */ 1597 case Opt_removed: 1598 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1599 return 1; 1600 case Opt_abort: 1601 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1602 return 1; 1603 case Opt_i_version: 1604 sb->s_flags |= MS_I_VERSION; 1605 return 1; 1606 case Opt_lazytime: 1607 sb->s_flags |= MS_LAZYTIME; 1608 return 1; 1609 case Opt_nolazytime: 1610 sb->s_flags &= ~MS_LAZYTIME; 1611 return 1; 1612 } 1613 1614 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1615 if (token == m->token) 1616 break; 1617 1618 if (m->token == Opt_err) { 1619 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1620 "or missing value", opt); 1621 return -1; 1622 } 1623 1624 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1625 ext4_msg(sb, KERN_ERR, 1626 "Mount option \"%s\" incompatible with ext2", opt); 1627 return -1; 1628 } 1629 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1630 ext4_msg(sb, KERN_ERR, 1631 "Mount option \"%s\" incompatible with ext3", opt); 1632 return -1; 1633 } 1634 1635 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1636 return -1; 1637 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1638 return -1; 1639 if (m->flags & MOPT_EXPLICIT) { 1640 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 1641 set_opt2(sb, EXPLICIT_DELALLOC); 1642 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 1643 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1644 } else 1645 return -1; 1646 } 1647 if (m->flags & MOPT_CLEAR_ERR) 1648 clear_opt(sb, ERRORS_MASK); 1649 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1650 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1651 "options when quota turned on"); 1652 return -1; 1653 } 1654 1655 if (m->flags & MOPT_NOSUPPORT) { 1656 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1657 } else if (token == Opt_commit) { 1658 if (arg == 0) 1659 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1660 sbi->s_commit_interval = HZ * arg; 1661 } else if (token == Opt_max_batch_time) { 1662 sbi->s_max_batch_time = arg; 1663 } else if (token == Opt_min_batch_time) { 1664 sbi->s_min_batch_time = arg; 1665 } else if (token == Opt_inode_readahead_blks) { 1666 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1667 ext4_msg(sb, KERN_ERR, 1668 "EXT4-fs: inode_readahead_blks must be " 1669 "0 or a power of 2 smaller than 2^31"); 1670 return -1; 1671 } 1672 sbi->s_inode_readahead_blks = arg; 1673 } else if (token == Opt_init_itable) { 1674 set_opt(sb, INIT_INODE_TABLE); 1675 if (!args->from) 1676 arg = EXT4_DEF_LI_WAIT_MULT; 1677 sbi->s_li_wait_mult = arg; 1678 } else if (token == Opt_max_dir_size_kb) { 1679 sbi->s_max_dir_size_kb = arg; 1680 } else if (token == Opt_stripe) { 1681 sbi->s_stripe = arg; 1682 } else if (token == Opt_resuid) { 1683 uid = make_kuid(current_user_ns(), arg); 1684 if (!uid_valid(uid)) { 1685 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1686 return -1; 1687 } 1688 sbi->s_resuid = uid; 1689 } else if (token == Opt_resgid) { 1690 gid = make_kgid(current_user_ns(), arg); 1691 if (!gid_valid(gid)) { 1692 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1693 return -1; 1694 } 1695 sbi->s_resgid = gid; 1696 } else if (token == Opt_journal_dev) { 1697 if (is_remount) { 1698 ext4_msg(sb, KERN_ERR, 1699 "Cannot specify journal on remount"); 1700 return -1; 1701 } 1702 *journal_devnum = arg; 1703 } else if (token == Opt_journal_path) { 1704 char *journal_path; 1705 struct inode *journal_inode; 1706 struct path path; 1707 int error; 1708 1709 if (is_remount) { 1710 ext4_msg(sb, KERN_ERR, 1711 "Cannot specify journal on remount"); 1712 return -1; 1713 } 1714 journal_path = match_strdup(&args[0]); 1715 if (!journal_path) { 1716 ext4_msg(sb, KERN_ERR, "error: could not dup " 1717 "journal device string"); 1718 return -1; 1719 } 1720 1721 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1722 if (error) { 1723 ext4_msg(sb, KERN_ERR, "error: could not find " 1724 "journal device path: error %d", error); 1725 kfree(journal_path); 1726 return -1; 1727 } 1728 1729 journal_inode = d_inode(path.dentry); 1730 if (!S_ISBLK(journal_inode->i_mode)) { 1731 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1732 "is not a block device", journal_path); 1733 path_put(&path); 1734 kfree(journal_path); 1735 return -1; 1736 } 1737 1738 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1739 path_put(&path); 1740 kfree(journal_path); 1741 } else if (token == Opt_journal_ioprio) { 1742 if (arg > 7) { 1743 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1744 " (must be 0-7)"); 1745 return -1; 1746 } 1747 *journal_ioprio = 1748 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1749 } else if (token == Opt_test_dummy_encryption) { 1750 #ifdef CONFIG_EXT4_FS_ENCRYPTION 1751 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 1752 ext4_msg(sb, KERN_WARNING, 1753 "Test dummy encryption mode enabled"); 1754 #else 1755 ext4_msg(sb, KERN_WARNING, 1756 "Test dummy encryption mount option ignored"); 1757 #endif 1758 } else if (m->flags & MOPT_DATAJ) { 1759 if (is_remount) { 1760 if (!sbi->s_journal) 1761 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1762 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 1763 ext4_msg(sb, KERN_ERR, 1764 "Cannot change data mode on remount"); 1765 return -1; 1766 } 1767 } else { 1768 clear_opt(sb, DATA_FLAGS); 1769 sbi->s_mount_opt |= m->mount_opt; 1770 } 1771 #ifdef CONFIG_QUOTA 1772 } else if (m->flags & MOPT_QFMT) { 1773 if (sb_any_quota_loaded(sb) && 1774 sbi->s_jquota_fmt != m->mount_opt) { 1775 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 1776 "quota options when quota turned on"); 1777 return -1; 1778 } 1779 if (ext4_has_feature_quota(sb)) { 1780 ext4_msg(sb, KERN_INFO, 1781 "Quota format mount options ignored " 1782 "when QUOTA feature is enabled"); 1783 return 1; 1784 } 1785 sbi->s_jquota_fmt = m->mount_opt; 1786 #endif 1787 } else if (token == Opt_dax) { 1788 #ifdef CONFIG_FS_DAX 1789 ext4_msg(sb, KERN_WARNING, 1790 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 1791 sbi->s_mount_opt |= m->mount_opt; 1792 #else 1793 ext4_msg(sb, KERN_INFO, "dax option not supported"); 1794 return -1; 1795 #endif 1796 } else if (token == Opt_data_err_abort) { 1797 sbi->s_mount_opt |= m->mount_opt; 1798 } else if (token == Opt_data_err_ignore) { 1799 sbi->s_mount_opt &= ~m->mount_opt; 1800 } else { 1801 if (!args->from) 1802 arg = 1; 1803 if (m->flags & MOPT_CLEAR) 1804 arg = !arg; 1805 else if (unlikely(!(m->flags & MOPT_SET))) { 1806 ext4_msg(sb, KERN_WARNING, 1807 "buggy handling of option %s", opt); 1808 WARN_ON(1); 1809 return -1; 1810 } 1811 if (arg != 0) 1812 sbi->s_mount_opt |= m->mount_opt; 1813 else 1814 sbi->s_mount_opt &= ~m->mount_opt; 1815 } 1816 return 1; 1817 } 1818 1819 static int parse_options(char *options, struct super_block *sb, 1820 unsigned long *journal_devnum, 1821 unsigned int *journal_ioprio, 1822 int is_remount) 1823 { 1824 struct ext4_sb_info *sbi = EXT4_SB(sb); 1825 char *p; 1826 substring_t args[MAX_OPT_ARGS]; 1827 int token; 1828 1829 if (!options) 1830 return 1; 1831 1832 while ((p = strsep(&options, ",")) != NULL) { 1833 if (!*p) 1834 continue; 1835 /* 1836 * Initialize args struct so we know whether arg was 1837 * found; some options take optional arguments. 1838 */ 1839 args[0].to = args[0].from = NULL; 1840 token = match_token(p, tokens, args); 1841 if (handle_mount_opt(sb, p, token, args, journal_devnum, 1842 journal_ioprio, is_remount) < 0) 1843 return 0; 1844 } 1845 #ifdef CONFIG_QUOTA 1846 /* 1847 * We do the test below only for project quotas. 'usrquota' and 1848 * 'grpquota' mount options are allowed even without quota feature 1849 * to support legacy quotas in quota files. 1850 */ 1851 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 1852 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 1853 "Cannot enable project quota enforcement."); 1854 return 0; 1855 } 1856 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1857 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1858 clear_opt(sb, USRQUOTA); 1859 1860 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1861 clear_opt(sb, GRPQUOTA); 1862 1863 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1864 ext4_msg(sb, KERN_ERR, "old and new quota " 1865 "format mixing"); 1866 return 0; 1867 } 1868 1869 if (!sbi->s_jquota_fmt) { 1870 ext4_msg(sb, KERN_ERR, "journaled quota format " 1871 "not specified"); 1872 return 0; 1873 } 1874 } 1875 #endif 1876 if (test_opt(sb, DIOREAD_NOLOCK)) { 1877 int blocksize = 1878 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 1879 1880 if (blocksize < PAGE_SIZE) { 1881 ext4_msg(sb, KERN_ERR, "can't mount with " 1882 "dioread_nolock if block size != PAGE_SIZE"); 1883 return 0; 1884 } 1885 } 1886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 1887 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 1888 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit " 1889 "in data=ordered mode"); 1890 return 0; 1891 } 1892 return 1; 1893 } 1894 1895 static inline void ext4_show_quota_options(struct seq_file *seq, 1896 struct super_block *sb) 1897 { 1898 #if defined(CONFIG_QUOTA) 1899 struct ext4_sb_info *sbi = EXT4_SB(sb); 1900 1901 if (sbi->s_jquota_fmt) { 1902 char *fmtname = ""; 1903 1904 switch (sbi->s_jquota_fmt) { 1905 case QFMT_VFS_OLD: 1906 fmtname = "vfsold"; 1907 break; 1908 case QFMT_VFS_V0: 1909 fmtname = "vfsv0"; 1910 break; 1911 case QFMT_VFS_V1: 1912 fmtname = "vfsv1"; 1913 break; 1914 } 1915 seq_printf(seq, ",jqfmt=%s", fmtname); 1916 } 1917 1918 if (sbi->s_qf_names[USRQUOTA]) 1919 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); 1920 1921 if (sbi->s_qf_names[GRPQUOTA]) 1922 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); 1923 #endif 1924 } 1925 1926 static const char *token2str(int token) 1927 { 1928 const struct match_token *t; 1929 1930 for (t = tokens; t->token != Opt_err; t++) 1931 if (t->token == token && !strchr(t->pattern, '=')) 1932 break; 1933 return t->pattern; 1934 } 1935 1936 /* 1937 * Show an option if 1938 * - it's set to a non-default value OR 1939 * - if the per-sb default is different from the global default 1940 */ 1941 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 1942 int nodefs) 1943 { 1944 struct ext4_sb_info *sbi = EXT4_SB(sb); 1945 struct ext4_super_block *es = sbi->s_es; 1946 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; 1947 const struct mount_opts *m; 1948 char sep = nodefs ? '\n' : ','; 1949 1950 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 1951 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 1952 1953 if (sbi->s_sb_block != 1) 1954 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 1955 1956 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 1957 int want_set = m->flags & MOPT_SET; 1958 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 1959 (m->flags & MOPT_CLEAR_ERR)) 1960 continue; 1961 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 1962 continue; /* skip if same as the default */ 1963 if ((want_set && 1964 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 1965 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 1966 continue; /* select Opt_noFoo vs Opt_Foo */ 1967 SEQ_OPTS_PRINT("%s", token2str(m->token)); 1968 } 1969 1970 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 1971 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 1972 SEQ_OPTS_PRINT("resuid=%u", 1973 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 1974 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 1975 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 1976 SEQ_OPTS_PRINT("resgid=%u", 1977 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 1978 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 1979 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 1980 SEQ_OPTS_PUTS("errors=remount-ro"); 1981 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 1982 SEQ_OPTS_PUTS("errors=continue"); 1983 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 1984 SEQ_OPTS_PUTS("errors=panic"); 1985 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 1986 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 1987 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 1988 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 1989 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 1990 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 1991 if (sb->s_flags & MS_I_VERSION) 1992 SEQ_OPTS_PUTS("i_version"); 1993 if (nodefs || sbi->s_stripe) 1994 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 1995 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { 1996 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 1997 SEQ_OPTS_PUTS("data=journal"); 1998 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 1999 SEQ_OPTS_PUTS("data=ordered"); 2000 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2001 SEQ_OPTS_PUTS("data=writeback"); 2002 } 2003 if (nodefs || 2004 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2005 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2006 sbi->s_inode_readahead_blks); 2007 2008 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && 2009 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2010 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2011 if (nodefs || sbi->s_max_dir_size_kb) 2012 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2013 if (test_opt(sb, DATA_ERR_ABORT)) 2014 SEQ_OPTS_PUTS("data_err=abort"); 2015 2016 ext4_show_quota_options(seq, sb); 2017 return 0; 2018 } 2019 2020 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2021 { 2022 return _ext4_show_options(seq, root->d_sb, 0); 2023 } 2024 2025 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2026 { 2027 struct super_block *sb = seq->private; 2028 int rc; 2029 2030 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw"); 2031 rc = _ext4_show_options(seq, sb, 1); 2032 seq_puts(seq, "\n"); 2033 return rc; 2034 } 2035 2036 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2037 int read_only) 2038 { 2039 struct ext4_sb_info *sbi = EXT4_SB(sb); 2040 int res = 0; 2041 2042 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2043 ext4_msg(sb, KERN_ERR, "revision level too high, " 2044 "forcing read-only mode"); 2045 res = MS_RDONLY; 2046 } 2047 if (read_only) 2048 goto done; 2049 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2050 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2051 "running e2fsck is recommended"); 2052 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2053 ext4_msg(sb, KERN_WARNING, 2054 "warning: mounting fs with errors, " 2055 "running e2fsck is recommended"); 2056 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2057 le16_to_cpu(es->s_mnt_count) >= 2058 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2059 ext4_msg(sb, KERN_WARNING, 2060 "warning: maximal mount count reached, " 2061 "running e2fsck is recommended"); 2062 else if (le32_to_cpu(es->s_checkinterval) && 2063 (le32_to_cpu(es->s_lastcheck) + 2064 le32_to_cpu(es->s_checkinterval) <= get_seconds())) 2065 ext4_msg(sb, KERN_WARNING, 2066 "warning: checktime reached, " 2067 "running e2fsck is recommended"); 2068 if (!sbi->s_journal) 2069 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2070 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2071 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2072 le16_add_cpu(&es->s_mnt_count, 1); 2073 es->s_mtime = cpu_to_le32(get_seconds()); 2074 ext4_update_dynamic_rev(sb); 2075 if (sbi->s_journal) 2076 ext4_set_feature_journal_needs_recovery(sb); 2077 2078 ext4_commit_super(sb, 1); 2079 done: 2080 if (test_opt(sb, DEBUG)) 2081 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2082 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2083 sb->s_blocksize, 2084 sbi->s_groups_count, 2085 EXT4_BLOCKS_PER_GROUP(sb), 2086 EXT4_INODES_PER_GROUP(sb), 2087 sbi->s_mount_opt, sbi->s_mount_opt2); 2088 2089 cleancache_init_fs(sb); 2090 return res; 2091 } 2092 2093 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2094 { 2095 struct ext4_sb_info *sbi = EXT4_SB(sb); 2096 struct flex_groups *new_groups; 2097 int size; 2098 2099 if (!sbi->s_log_groups_per_flex) 2100 return 0; 2101 2102 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2103 if (size <= sbi->s_flex_groups_allocated) 2104 return 0; 2105 2106 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2107 new_groups = ext4_kvzalloc(size, GFP_KERNEL); 2108 if (!new_groups) { 2109 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2110 size / (int) sizeof(struct flex_groups)); 2111 return -ENOMEM; 2112 } 2113 2114 if (sbi->s_flex_groups) { 2115 memcpy(new_groups, sbi->s_flex_groups, 2116 (sbi->s_flex_groups_allocated * 2117 sizeof(struct flex_groups))); 2118 kvfree(sbi->s_flex_groups); 2119 } 2120 sbi->s_flex_groups = new_groups; 2121 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2122 return 0; 2123 } 2124 2125 static int ext4_fill_flex_info(struct super_block *sb) 2126 { 2127 struct ext4_sb_info *sbi = EXT4_SB(sb); 2128 struct ext4_group_desc *gdp = NULL; 2129 ext4_group_t flex_group; 2130 int i, err; 2131 2132 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2133 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2134 sbi->s_log_groups_per_flex = 0; 2135 return 1; 2136 } 2137 2138 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2139 if (err) 2140 goto failed; 2141 2142 for (i = 0; i < sbi->s_groups_count; i++) { 2143 gdp = ext4_get_group_desc(sb, i, NULL); 2144 2145 flex_group = ext4_flex_group(sbi, i); 2146 atomic_add(ext4_free_inodes_count(sb, gdp), 2147 &sbi->s_flex_groups[flex_group].free_inodes); 2148 atomic64_add(ext4_free_group_clusters(sb, gdp), 2149 &sbi->s_flex_groups[flex_group].free_clusters); 2150 atomic_add(ext4_used_dirs_count(sb, gdp), 2151 &sbi->s_flex_groups[flex_group].used_dirs); 2152 } 2153 2154 return 1; 2155 failed: 2156 return 0; 2157 } 2158 2159 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2160 struct ext4_group_desc *gdp) 2161 { 2162 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2163 __u16 crc = 0; 2164 __le32 le_group = cpu_to_le32(block_group); 2165 struct ext4_sb_info *sbi = EXT4_SB(sb); 2166 2167 if (ext4_has_metadata_csum(sbi->s_sb)) { 2168 /* Use new metadata_csum algorithm */ 2169 __u32 csum32; 2170 __u16 dummy_csum = 0; 2171 2172 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2173 sizeof(le_group)); 2174 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2175 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2176 sizeof(dummy_csum)); 2177 offset += sizeof(dummy_csum); 2178 if (offset < sbi->s_desc_size) 2179 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2180 sbi->s_desc_size - offset); 2181 2182 crc = csum32 & 0xFFFF; 2183 goto out; 2184 } 2185 2186 /* old crc16 code */ 2187 if (!ext4_has_feature_gdt_csum(sb)) 2188 return 0; 2189 2190 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2191 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2192 crc = crc16(crc, (__u8 *)gdp, offset); 2193 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2194 /* for checksum of struct ext4_group_desc do the rest...*/ 2195 if (ext4_has_feature_64bit(sb) && 2196 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2197 crc = crc16(crc, (__u8 *)gdp + offset, 2198 le16_to_cpu(sbi->s_es->s_desc_size) - 2199 offset); 2200 2201 out: 2202 return cpu_to_le16(crc); 2203 } 2204 2205 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2206 struct ext4_group_desc *gdp) 2207 { 2208 if (ext4_has_group_desc_csum(sb) && 2209 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2210 return 0; 2211 2212 return 1; 2213 } 2214 2215 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2216 struct ext4_group_desc *gdp) 2217 { 2218 if (!ext4_has_group_desc_csum(sb)) 2219 return; 2220 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2221 } 2222 2223 /* Called at mount-time, super-block is locked */ 2224 static int ext4_check_descriptors(struct super_block *sb, 2225 ext4_fsblk_t sb_block, 2226 ext4_group_t *first_not_zeroed) 2227 { 2228 struct ext4_sb_info *sbi = EXT4_SB(sb); 2229 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2230 ext4_fsblk_t last_block; 2231 ext4_fsblk_t block_bitmap; 2232 ext4_fsblk_t inode_bitmap; 2233 ext4_fsblk_t inode_table; 2234 int flexbg_flag = 0; 2235 ext4_group_t i, grp = sbi->s_groups_count; 2236 2237 if (ext4_has_feature_flex_bg(sb)) 2238 flexbg_flag = 1; 2239 2240 ext4_debug("Checking group descriptors"); 2241 2242 for (i = 0; i < sbi->s_groups_count; i++) { 2243 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2244 2245 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2246 last_block = ext4_blocks_count(sbi->s_es) - 1; 2247 else 2248 last_block = first_block + 2249 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2250 2251 if ((grp == sbi->s_groups_count) && 2252 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2253 grp = i; 2254 2255 block_bitmap = ext4_block_bitmap(sb, gdp); 2256 if (block_bitmap == sb_block) { 2257 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2258 "Block bitmap for group %u overlaps " 2259 "superblock", i); 2260 } 2261 if (block_bitmap < first_block || block_bitmap > last_block) { 2262 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2263 "Block bitmap for group %u not in group " 2264 "(block %llu)!", i, block_bitmap); 2265 return 0; 2266 } 2267 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2268 if (inode_bitmap == sb_block) { 2269 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2270 "Inode bitmap for group %u overlaps " 2271 "superblock", i); 2272 } 2273 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2274 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2275 "Inode bitmap for group %u not in group " 2276 "(block %llu)!", i, inode_bitmap); 2277 return 0; 2278 } 2279 inode_table = ext4_inode_table(sb, gdp); 2280 if (inode_table == sb_block) { 2281 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2282 "Inode table for group %u overlaps " 2283 "superblock", i); 2284 } 2285 if (inode_table < first_block || 2286 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2288 "Inode table for group %u not in group " 2289 "(block %llu)!", i, inode_table); 2290 return 0; 2291 } 2292 ext4_lock_group(sb, i); 2293 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2295 "Checksum for group %u failed (%u!=%u)", 2296 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2297 gdp)), le16_to_cpu(gdp->bg_checksum)); 2298 if (!(sb->s_flags & MS_RDONLY)) { 2299 ext4_unlock_group(sb, i); 2300 return 0; 2301 } 2302 } 2303 ext4_unlock_group(sb, i); 2304 if (!flexbg_flag) 2305 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2306 } 2307 if (NULL != first_not_zeroed) 2308 *first_not_zeroed = grp; 2309 return 1; 2310 } 2311 2312 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2313 * the superblock) which were deleted from all directories, but held open by 2314 * a process at the time of a crash. We walk the list and try to delete these 2315 * inodes at recovery time (only with a read-write filesystem). 2316 * 2317 * In order to keep the orphan inode chain consistent during traversal (in 2318 * case of crash during recovery), we link each inode into the superblock 2319 * orphan list_head and handle it the same way as an inode deletion during 2320 * normal operation (which journals the operations for us). 2321 * 2322 * We only do an iget() and an iput() on each inode, which is very safe if we 2323 * accidentally point at an in-use or already deleted inode. The worst that 2324 * can happen in this case is that we get a "bit already cleared" message from 2325 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2326 * e2fsck was run on this filesystem, and it must have already done the orphan 2327 * inode cleanup for us, so we can safely abort without any further action. 2328 */ 2329 static void ext4_orphan_cleanup(struct super_block *sb, 2330 struct ext4_super_block *es) 2331 { 2332 unsigned int s_flags = sb->s_flags; 2333 int nr_orphans = 0, nr_truncates = 0; 2334 #ifdef CONFIG_QUOTA 2335 int i; 2336 #endif 2337 if (!es->s_last_orphan) { 2338 jbd_debug(4, "no orphan inodes to clean up\n"); 2339 return; 2340 } 2341 2342 if (bdev_read_only(sb->s_bdev)) { 2343 ext4_msg(sb, KERN_ERR, "write access " 2344 "unavailable, skipping orphan cleanup"); 2345 return; 2346 } 2347 2348 /* Check if feature set would not allow a r/w mount */ 2349 if (!ext4_feature_set_ok(sb, 0)) { 2350 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2351 "unknown ROCOMPAT features"); 2352 return; 2353 } 2354 2355 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2356 /* don't clear list on RO mount w/ errors */ 2357 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { 2358 ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2359 "clearing orphan list.\n"); 2360 es->s_last_orphan = 0; 2361 } 2362 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2363 return; 2364 } 2365 2366 if (s_flags & MS_RDONLY) { 2367 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2368 sb->s_flags &= ~MS_RDONLY; 2369 } 2370 #ifdef CONFIG_QUOTA 2371 /* Needed for iput() to work correctly and not trash data */ 2372 sb->s_flags |= MS_ACTIVE; 2373 /* Turn on quotas so that they are updated correctly */ 2374 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2375 if (EXT4_SB(sb)->s_qf_names[i]) { 2376 int ret = ext4_quota_on_mount(sb, i); 2377 if (ret < 0) 2378 ext4_msg(sb, KERN_ERR, 2379 "Cannot turn on journaled " 2380 "quota: error %d", ret); 2381 } 2382 } 2383 #endif 2384 2385 while (es->s_last_orphan) { 2386 struct inode *inode; 2387 2388 /* 2389 * We may have encountered an error during cleanup; if 2390 * so, skip the rest. 2391 */ 2392 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2393 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2394 es->s_last_orphan = 0; 2395 break; 2396 } 2397 2398 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2399 if (IS_ERR(inode)) { 2400 es->s_last_orphan = 0; 2401 break; 2402 } 2403 2404 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2405 dquot_initialize(inode); 2406 if (inode->i_nlink) { 2407 if (test_opt(sb, DEBUG)) 2408 ext4_msg(sb, KERN_DEBUG, 2409 "%s: truncating inode %lu to %lld bytes", 2410 __func__, inode->i_ino, inode->i_size); 2411 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2412 inode->i_ino, inode->i_size); 2413 inode_lock(inode); 2414 truncate_inode_pages(inode->i_mapping, inode->i_size); 2415 ext4_truncate(inode); 2416 inode_unlock(inode); 2417 nr_truncates++; 2418 } else { 2419 if (test_opt(sb, DEBUG)) 2420 ext4_msg(sb, KERN_DEBUG, 2421 "%s: deleting unreferenced inode %lu", 2422 __func__, inode->i_ino); 2423 jbd_debug(2, "deleting unreferenced inode %lu\n", 2424 inode->i_ino); 2425 nr_orphans++; 2426 } 2427 iput(inode); /* The delete magic happens here! */ 2428 } 2429 2430 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2431 2432 if (nr_orphans) 2433 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2434 PLURAL(nr_orphans)); 2435 if (nr_truncates) 2436 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2437 PLURAL(nr_truncates)); 2438 #ifdef CONFIG_QUOTA 2439 /* Turn quotas off */ 2440 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2441 if (sb_dqopt(sb)->files[i]) 2442 dquot_quota_off(sb, i); 2443 } 2444 #endif 2445 sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2446 } 2447 2448 /* 2449 * Maximal extent format file size. 2450 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2451 * extent format containers, within a sector_t, and within i_blocks 2452 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2453 * so that won't be a limiting factor. 2454 * 2455 * However there is other limiting factor. We do store extents in the form 2456 * of starting block and length, hence the resulting length of the extent 2457 * covering maximum file size must fit into on-disk format containers as 2458 * well. Given that length is always by 1 unit bigger than max unit (because 2459 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2460 * 2461 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2462 */ 2463 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2464 { 2465 loff_t res; 2466 loff_t upper_limit = MAX_LFS_FILESIZE; 2467 2468 /* small i_blocks in vfs inode? */ 2469 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2470 /* 2471 * CONFIG_LBDAF is not enabled implies the inode 2472 * i_block represent total blocks in 512 bytes 2473 * 32 == size of vfs inode i_blocks * 8 2474 */ 2475 upper_limit = (1LL << 32) - 1; 2476 2477 /* total blocks in file system block size */ 2478 upper_limit >>= (blkbits - 9); 2479 upper_limit <<= blkbits; 2480 } 2481 2482 /* 2483 * 32-bit extent-start container, ee_block. We lower the maxbytes 2484 * by one fs block, so ee_len can cover the extent of maximum file 2485 * size 2486 */ 2487 res = (1LL << 32) - 1; 2488 res <<= blkbits; 2489 2490 /* Sanity check against vm- & vfs- imposed limits */ 2491 if (res > upper_limit) 2492 res = upper_limit; 2493 2494 return res; 2495 } 2496 2497 /* 2498 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2499 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2500 * We need to be 1 filesystem block less than the 2^48 sector limit. 2501 */ 2502 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2503 { 2504 loff_t res = EXT4_NDIR_BLOCKS; 2505 int meta_blocks; 2506 loff_t upper_limit; 2507 /* This is calculated to be the largest file size for a dense, block 2508 * mapped file such that the file's total number of 512-byte sectors, 2509 * including data and all indirect blocks, does not exceed (2^48 - 1). 2510 * 2511 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2512 * number of 512-byte sectors of the file. 2513 */ 2514 2515 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2516 /* 2517 * !has_huge_files or CONFIG_LBDAF not enabled implies that 2518 * the inode i_block field represents total file blocks in 2519 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 2520 */ 2521 upper_limit = (1LL << 32) - 1; 2522 2523 /* total blocks in file system block size */ 2524 upper_limit >>= (bits - 9); 2525 2526 } else { 2527 /* 2528 * We use 48 bit ext4_inode i_blocks 2529 * With EXT4_HUGE_FILE_FL set the i_blocks 2530 * represent total number of blocks in 2531 * file system block size 2532 */ 2533 upper_limit = (1LL << 48) - 1; 2534 2535 } 2536 2537 /* indirect blocks */ 2538 meta_blocks = 1; 2539 /* double indirect blocks */ 2540 meta_blocks += 1 + (1LL << (bits-2)); 2541 /* tripple indirect blocks */ 2542 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2543 2544 upper_limit -= meta_blocks; 2545 upper_limit <<= bits; 2546 2547 res += 1LL << (bits-2); 2548 res += 1LL << (2*(bits-2)); 2549 res += 1LL << (3*(bits-2)); 2550 res <<= bits; 2551 if (res > upper_limit) 2552 res = upper_limit; 2553 2554 if (res > MAX_LFS_FILESIZE) 2555 res = MAX_LFS_FILESIZE; 2556 2557 return res; 2558 } 2559 2560 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2561 ext4_fsblk_t logical_sb_block, int nr) 2562 { 2563 struct ext4_sb_info *sbi = EXT4_SB(sb); 2564 ext4_group_t bg, first_meta_bg; 2565 int has_super = 0; 2566 2567 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2568 2569 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 2570 return logical_sb_block + nr + 1; 2571 bg = sbi->s_desc_per_block * nr; 2572 if (ext4_bg_has_super(sb, bg)) 2573 has_super = 1; 2574 2575 /* 2576 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2577 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2578 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2579 * compensate. 2580 */ 2581 if (sb->s_blocksize == 1024 && nr == 0 && 2582 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0) 2583 has_super++; 2584 2585 return (has_super + ext4_group_first_block_no(sb, bg)); 2586 } 2587 2588 /** 2589 * ext4_get_stripe_size: Get the stripe size. 2590 * @sbi: In memory super block info 2591 * 2592 * If we have specified it via mount option, then 2593 * use the mount option value. If the value specified at mount time is 2594 * greater than the blocks per group use the super block value. 2595 * If the super block value is greater than blocks per group return 0. 2596 * Allocator needs it be less than blocks per group. 2597 * 2598 */ 2599 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2600 { 2601 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2602 unsigned long stripe_width = 2603 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2604 int ret; 2605 2606 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2607 ret = sbi->s_stripe; 2608 else if (stripe_width <= sbi->s_blocks_per_group) 2609 ret = stripe_width; 2610 else if (stride <= sbi->s_blocks_per_group) 2611 ret = stride; 2612 else 2613 ret = 0; 2614 2615 /* 2616 * If the stripe width is 1, this makes no sense and 2617 * we set it to 0 to turn off stripe handling code. 2618 */ 2619 if (ret <= 1) 2620 ret = 0; 2621 2622 return ret; 2623 } 2624 2625 /* 2626 * Check whether this filesystem can be mounted based on 2627 * the features present and the RDONLY/RDWR mount requested. 2628 * Returns 1 if this filesystem can be mounted as requested, 2629 * 0 if it cannot be. 2630 */ 2631 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2632 { 2633 if (ext4_has_unknown_ext4_incompat_features(sb)) { 2634 ext4_msg(sb, KERN_ERR, 2635 "Couldn't mount because of " 2636 "unsupported optional features (%x)", 2637 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2638 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2639 return 0; 2640 } 2641 2642 if (readonly) 2643 return 1; 2644 2645 if (ext4_has_feature_readonly(sb)) { 2646 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 2647 sb->s_flags |= MS_RDONLY; 2648 return 1; 2649 } 2650 2651 /* Check that feature set is OK for a read-write mount */ 2652 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2653 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2654 "unsupported optional features (%x)", 2655 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2656 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2657 return 0; 2658 } 2659 /* 2660 * Large file size enabled file system can only be mounted 2661 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2662 */ 2663 if (ext4_has_feature_huge_file(sb)) { 2664 if (sizeof(blkcnt_t) < sizeof(u64)) { 2665 ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2666 "cannot be mounted RDWR without " 2667 "CONFIG_LBDAF"); 2668 return 0; 2669 } 2670 } 2671 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2672 ext4_msg(sb, KERN_ERR, 2673 "Can't support bigalloc feature without " 2674 "extents feature\n"); 2675 return 0; 2676 } 2677 2678 #ifndef CONFIG_QUOTA 2679 if (ext4_has_feature_quota(sb) && !readonly) { 2680 ext4_msg(sb, KERN_ERR, 2681 "Filesystem with quota feature cannot be mounted RDWR " 2682 "without CONFIG_QUOTA"); 2683 return 0; 2684 } 2685 if (ext4_has_feature_project(sb) && !readonly) { 2686 ext4_msg(sb, KERN_ERR, 2687 "Filesystem with project quota feature cannot be mounted RDWR " 2688 "without CONFIG_QUOTA"); 2689 return 0; 2690 } 2691 #endif /* CONFIG_QUOTA */ 2692 return 1; 2693 } 2694 2695 /* 2696 * This function is called once a day if we have errors logged 2697 * on the file system 2698 */ 2699 static void print_daily_error_info(unsigned long arg) 2700 { 2701 struct super_block *sb = (struct super_block *) arg; 2702 struct ext4_sb_info *sbi; 2703 struct ext4_super_block *es; 2704 2705 sbi = EXT4_SB(sb); 2706 es = sbi->s_es; 2707 2708 if (es->s_error_count) 2709 /* fsck newer than v1.41.13 is needed to clean this condition. */ 2710 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 2711 le32_to_cpu(es->s_error_count)); 2712 if (es->s_first_error_time) { 2713 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d", 2714 sb->s_id, le32_to_cpu(es->s_first_error_time), 2715 (int) sizeof(es->s_first_error_func), 2716 es->s_first_error_func, 2717 le32_to_cpu(es->s_first_error_line)); 2718 if (es->s_first_error_ino) 2719 printk(KERN_CONT ": inode %u", 2720 le32_to_cpu(es->s_first_error_ino)); 2721 if (es->s_first_error_block) 2722 printk(KERN_CONT ": block %llu", (unsigned long long) 2723 le64_to_cpu(es->s_first_error_block)); 2724 printk(KERN_CONT "\n"); 2725 } 2726 if (es->s_last_error_time) { 2727 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d", 2728 sb->s_id, le32_to_cpu(es->s_last_error_time), 2729 (int) sizeof(es->s_last_error_func), 2730 es->s_last_error_func, 2731 le32_to_cpu(es->s_last_error_line)); 2732 if (es->s_last_error_ino) 2733 printk(KERN_CONT ": inode %u", 2734 le32_to_cpu(es->s_last_error_ino)); 2735 if (es->s_last_error_block) 2736 printk(KERN_CONT ": block %llu", (unsigned long long) 2737 le64_to_cpu(es->s_last_error_block)); 2738 printk(KERN_CONT "\n"); 2739 } 2740 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 2741 } 2742 2743 /* Find next suitable group and run ext4_init_inode_table */ 2744 static int ext4_run_li_request(struct ext4_li_request *elr) 2745 { 2746 struct ext4_group_desc *gdp = NULL; 2747 ext4_group_t group, ngroups; 2748 struct super_block *sb; 2749 unsigned long timeout = 0; 2750 int ret = 0; 2751 2752 sb = elr->lr_super; 2753 ngroups = EXT4_SB(sb)->s_groups_count; 2754 2755 for (group = elr->lr_next_group; group < ngroups; group++) { 2756 gdp = ext4_get_group_desc(sb, group, NULL); 2757 if (!gdp) { 2758 ret = 1; 2759 break; 2760 } 2761 2762 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2763 break; 2764 } 2765 2766 if (group >= ngroups) 2767 ret = 1; 2768 2769 if (!ret) { 2770 timeout = jiffies; 2771 ret = ext4_init_inode_table(sb, group, 2772 elr->lr_timeout ? 0 : 1); 2773 if (elr->lr_timeout == 0) { 2774 timeout = (jiffies - timeout) * 2775 elr->lr_sbi->s_li_wait_mult; 2776 elr->lr_timeout = timeout; 2777 } 2778 elr->lr_next_sched = jiffies + elr->lr_timeout; 2779 elr->lr_next_group = group + 1; 2780 } 2781 return ret; 2782 } 2783 2784 /* 2785 * Remove lr_request from the list_request and free the 2786 * request structure. Should be called with li_list_mtx held 2787 */ 2788 static void ext4_remove_li_request(struct ext4_li_request *elr) 2789 { 2790 struct ext4_sb_info *sbi; 2791 2792 if (!elr) 2793 return; 2794 2795 sbi = elr->lr_sbi; 2796 2797 list_del(&elr->lr_request); 2798 sbi->s_li_request = NULL; 2799 kfree(elr); 2800 } 2801 2802 static void ext4_unregister_li_request(struct super_block *sb) 2803 { 2804 mutex_lock(&ext4_li_mtx); 2805 if (!ext4_li_info) { 2806 mutex_unlock(&ext4_li_mtx); 2807 return; 2808 } 2809 2810 mutex_lock(&ext4_li_info->li_list_mtx); 2811 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2812 mutex_unlock(&ext4_li_info->li_list_mtx); 2813 mutex_unlock(&ext4_li_mtx); 2814 } 2815 2816 static struct task_struct *ext4_lazyinit_task; 2817 2818 /* 2819 * This is the function where ext4lazyinit thread lives. It walks 2820 * through the request list searching for next scheduled filesystem. 2821 * When such a fs is found, run the lazy initialization request 2822 * (ext4_rn_li_request) and keep track of the time spend in this 2823 * function. Based on that time we compute next schedule time of 2824 * the request. When walking through the list is complete, compute 2825 * next waking time and put itself into sleep. 2826 */ 2827 static int ext4_lazyinit_thread(void *arg) 2828 { 2829 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2830 struct list_head *pos, *n; 2831 struct ext4_li_request *elr; 2832 unsigned long next_wakeup, cur; 2833 2834 BUG_ON(NULL == eli); 2835 2836 cont_thread: 2837 while (true) { 2838 next_wakeup = MAX_JIFFY_OFFSET; 2839 2840 mutex_lock(&eli->li_list_mtx); 2841 if (list_empty(&eli->li_request_list)) { 2842 mutex_unlock(&eli->li_list_mtx); 2843 goto exit_thread; 2844 } 2845 list_for_each_safe(pos, n, &eli->li_request_list) { 2846 int err = 0; 2847 int progress = 0; 2848 elr = list_entry(pos, struct ext4_li_request, 2849 lr_request); 2850 2851 if (time_before(jiffies, elr->lr_next_sched)) { 2852 if (time_before(elr->lr_next_sched, next_wakeup)) 2853 next_wakeup = elr->lr_next_sched; 2854 continue; 2855 } 2856 if (down_read_trylock(&elr->lr_super->s_umount)) { 2857 if (sb_start_write_trylock(elr->lr_super)) { 2858 progress = 1; 2859 /* 2860 * We hold sb->s_umount, sb can not 2861 * be removed from the list, it is 2862 * now safe to drop li_list_mtx 2863 */ 2864 mutex_unlock(&eli->li_list_mtx); 2865 err = ext4_run_li_request(elr); 2866 sb_end_write(elr->lr_super); 2867 mutex_lock(&eli->li_list_mtx); 2868 n = pos->next; 2869 } 2870 up_read((&elr->lr_super->s_umount)); 2871 } 2872 /* error, remove the lazy_init job */ 2873 if (err) { 2874 ext4_remove_li_request(elr); 2875 continue; 2876 } 2877 if (!progress) { 2878 elr->lr_next_sched = jiffies + 2879 (prandom_u32() 2880 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 2881 } 2882 if (time_before(elr->lr_next_sched, next_wakeup)) 2883 next_wakeup = elr->lr_next_sched; 2884 } 2885 mutex_unlock(&eli->li_list_mtx); 2886 2887 try_to_freeze(); 2888 2889 cur = jiffies; 2890 if ((time_after_eq(cur, next_wakeup)) || 2891 (MAX_JIFFY_OFFSET == next_wakeup)) { 2892 cond_resched(); 2893 continue; 2894 } 2895 2896 schedule_timeout_interruptible(next_wakeup - cur); 2897 2898 if (kthread_should_stop()) { 2899 ext4_clear_request_list(); 2900 goto exit_thread; 2901 } 2902 } 2903 2904 exit_thread: 2905 /* 2906 * It looks like the request list is empty, but we need 2907 * to check it under the li_list_mtx lock, to prevent any 2908 * additions into it, and of course we should lock ext4_li_mtx 2909 * to atomically free the list and ext4_li_info, because at 2910 * this point another ext4 filesystem could be registering 2911 * new one. 2912 */ 2913 mutex_lock(&ext4_li_mtx); 2914 mutex_lock(&eli->li_list_mtx); 2915 if (!list_empty(&eli->li_request_list)) { 2916 mutex_unlock(&eli->li_list_mtx); 2917 mutex_unlock(&ext4_li_mtx); 2918 goto cont_thread; 2919 } 2920 mutex_unlock(&eli->li_list_mtx); 2921 kfree(ext4_li_info); 2922 ext4_li_info = NULL; 2923 mutex_unlock(&ext4_li_mtx); 2924 2925 return 0; 2926 } 2927 2928 static void ext4_clear_request_list(void) 2929 { 2930 struct list_head *pos, *n; 2931 struct ext4_li_request *elr; 2932 2933 mutex_lock(&ext4_li_info->li_list_mtx); 2934 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 2935 elr = list_entry(pos, struct ext4_li_request, 2936 lr_request); 2937 ext4_remove_li_request(elr); 2938 } 2939 mutex_unlock(&ext4_li_info->li_list_mtx); 2940 } 2941 2942 static int ext4_run_lazyinit_thread(void) 2943 { 2944 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 2945 ext4_li_info, "ext4lazyinit"); 2946 if (IS_ERR(ext4_lazyinit_task)) { 2947 int err = PTR_ERR(ext4_lazyinit_task); 2948 ext4_clear_request_list(); 2949 kfree(ext4_li_info); 2950 ext4_li_info = NULL; 2951 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 2952 "initialization thread\n", 2953 err); 2954 return err; 2955 } 2956 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 2957 return 0; 2958 } 2959 2960 /* 2961 * Check whether it make sense to run itable init. thread or not. 2962 * If there is at least one uninitialized inode table, return 2963 * corresponding group number, else the loop goes through all 2964 * groups and return total number of groups. 2965 */ 2966 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 2967 { 2968 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 2969 struct ext4_group_desc *gdp = NULL; 2970 2971 for (group = 0; group < ngroups; group++) { 2972 gdp = ext4_get_group_desc(sb, group, NULL); 2973 if (!gdp) 2974 continue; 2975 2976 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2977 break; 2978 } 2979 2980 return group; 2981 } 2982 2983 static int ext4_li_info_new(void) 2984 { 2985 struct ext4_lazy_init *eli = NULL; 2986 2987 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 2988 if (!eli) 2989 return -ENOMEM; 2990 2991 INIT_LIST_HEAD(&eli->li_request_list); 2992 mutex_init(&eli->li_list_mtx); 2993 2994 eli->li_state |= EXT4_LAZYINIT_QUIT; 2995 2996 ext4_li_info = eli; 2997 2998 return 0; 2999 } 3000 3001 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3002 ext4_group_t start) 3003 { 3004 struct ext4_sb_info *sbi = EXT4_SB(sb); 3005 struct ext4_li_request *elr; 3006 3007 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3008 if (!elr) 3009 return NULL; 3010 3011 elr->lr_super = sb; 3012 elr->lr_sbi = sbi; 3013 elr->lr_next_group = start; 3014 3015 /* 3016 * Randomize first schedule time of the request to 3017 * spread the inode table initialization requests 3018 * better. 3019 */ 3020 elr->lr_next_sched = jiffies + (prandom_u32() % 3021 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3022 return elr; 3023 } 3024 3025 int ext4_register_li_request(struct super_block *sb, 3026 ext4_group_t first_not_zeroed) 3027 { 3028 struct ext4_sb_info *sbi = EXT4_SB(sb); 3029 struct ext4_li_request *elr = NULL; 3030 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3031 int ret = 0; 3032 3033 mutex_lock(&ext4_li_mtx); 3034 if (sbi->s_li_request != NULL) { 3035 /* 3036 * Reset timeout so it can be computed again, because 3037 * s_li_wait_mult might have changed. 3038 */ 3039 sbi->s_li_request->lr_timeout = 0; 3040 goto out; 3041 } 3042 3043 if (first_not_zeroed == ngroups || 3044 (sb->s_flags & MS_RDONLY) || 3045 !test_opt(sb, INIT_INODE_TABLE)) 3046 goto out; 3047 3048 elr = ext4_li_request_new(sb, first_not_zeroed); 3049 if (!elr) { 3050 ret = -ENOMEM; 3051 goto out; 3052 } 3053 3054 if (NULL == ext4_li_info) { 3055 ret = ext4_li_info_new(); 3056 if (ret) 3057 goto out; 3058 } 3059 3060 mutex_lock(&ext4_li_info->li_list_mtx); 3061 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3062 mutex_unlock(&ext4_li_info->li_list_mtx); 3063 3064 sbi->s_li_request = elr; 3065 /* 3066 * set elr to NULL here since it has been inserted to 3067 * the request_list and the removal and free of it is 3068 * handled by ext4_clear_request_list from now on. 3069 */ 3070 elr = NULL; 3071 3072 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3073 ret = ext4_run_lazyinit_thread(); 3074 if (ret) 3075 goto out; 3076 } 3077 out: 3078 mutex_unlock(&ext4_li_mtx); 3079 if (ret) 3080 kfree(elr); 3081 return ret; 3082 } 3083 3084 /* 3085 * We do not need to lock anything since this is called on 3086 * module unload. 3087 */ 3088 static void ext4_destroy_lazyinit_thread(void) 3089 { 3090 /* 3091 * If thread exited earlier 3092 * there's nothing to be done. 3093 */ 3094 if (!ext4_li_info || !ext4_lazyinit_task) 3095 return; 3096 3097 kthread_stop(ext4_lazyinit_task); 3098 } 3099 3100 static int set_journal_csum_feature_set(struct super_block *sb) 3101 { 3102 int ret = 1; 3103 int compat, incompat; 3104 struct ext4_sb_info *sbi = EXT4_SB(sb); 3105 3106 if (ext4_has_metadata_csum(sb)) { 3107 /* journal checksum v3 */ 3108 compat = 0; 3109 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3110 } else { 3111 /* journal checksum v1 */ 3112 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3113 incompat = 0; 3114 } 3115 3116 jbd2_journal_clear_features(sbi->s_journal, 3117 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3118 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3119 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3120 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3121 ret = jbd2_journal_set_features(sbi->s_journal, 3122 compat, 0, 3123 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3124 incompat); 3125 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3126 ret = jbd2_journal_set_features(sbi->s_journal, 3127 compat, 0, 3128 incompat); 3129 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3130 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3131 } else { 3132 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3133 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3134 } 3135 3136 return ret; 3137 } 3138 3139 /* 3140 * Note: calculating the overhead so we can be compatible with 3141 * historical BSD practice is quite difficult in the face of 3142 * clusters/bigalloc. This is because multiple metadata blocks from 3143 * different block group can end up in the same allocation cluster. 3144 * Calculating the exact overhead in the face of clustered allocation 3145 * requires either O(all block bitmaps) in memory or O(number of block 3146 * groups**2) in time. We will still calculate the superblock for 3147 * older file systems --- and if we come across with a bigalloc file 3148 * system with zero in s_overhead_clusters the estimate will be close to 3149 * correct especially for very large cluster sizes --- but for newer 3150 * file systems, it's better to calculate this figure once at mkfs 3151 * time, and store it in the superblock. If the superblock value is 3152 * present (even for non-bigalloc file systems), we will use it. 3153 */ 3154 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3155 char *buf) 3156 { 3157 struct ext4_sb_info *sbi = EXT4_SB(sb); 3158 struct ext4_group_desc *gdp; 3159 ext4_fsblk_t first_block, last_block, b; 3160 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3161 int s, j, count = 0; 3162 3163 if (!ext4_has_feature_bigalloc(sb)) 3164 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3165 sbi->s_itb_per_group + 2); 3166 3167 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3168 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3169 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3170 for (i = 0; i < ngroups; i++) { 3171 gdp = ext4_get_group_desc(sb, i, NULL); 3172 b = ext4_block_bitmap(sb, gdp); 3173 if (b >= first_block && b <= last_block) { 3174 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3175 count++; 3176 } 3177 b = ext4_inode_bitmap(sb, gdp); 3178 if (b >= first_block && b <= last_block) { 3179 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3180 count++; 3181 } 3182 b = ext4_inode_table(sb, gdp); 3183 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3184 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3185 int c = EXT4_B2C(sbi, b - first_block); 3186 ext4_set_bit(c, buf); 3187 count++; 3188 } 3189 if (i != grp) 3190 continue; 3191 s = 0; 3192 if (ext4_bg_has_super(sb, grp)) { 3193 ext4_set_bit(s++, buf); 3194 count++; 3195 } 3196 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { 3197 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3198 count++; 3199 } 3200 } 3201 if (!count) 3202 return 0; 3203 return EXT4_CLUSTERS_PER_GROUP(sb) - 3204 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3205 } 3206 3207 /* 3208 * Compute the overhead and stash it in sbi->s_overhead 3209 */ 3210 int ext4_calculate_overhead(struct super_block *sb) 3211 { 3212 struct ext4_sb_info *sbi = EXT4_SB(sb); 3213 struct ext4_super_block *es = sbi->s_es; 3214 struct inode *j_inode; 3215 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3216 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3217 ext4_fsblk_t overhead = 0; 3218 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3219 3220 if (!buf) 3221 return -ENOMEM; 3222 3223 /* 3224 * Compute the overhead (FS structures). This is constant 3225 * for a given filesystem unless the number of block groups 3226 * changes so we cache the previous value until it does. 3227 */ 3228 3229 /* 3230 * All of the blocks before first_data_block are overhead 3231 */ 3232 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3233 3234 /* 3235 * Add the overhead found in each block group 3236 */ 3237 for (i = 0; i < ngroups; i++) { 3238 int blks; 3239 3240 blks = count_overhead(sb, i, buf); 3241 overhead += blks; 3242 if (blks) 3243 memset(buf, 0, PAGE_SIZE); 3244 cond_resched(); 3245 } 3246 3247 /* 3248 * Add the internal journal blocks whether the journal has been 3249 * loaded or not 3250 */ 3251 if (sbi->s_journal && !sbi->journal_bdev) 3252 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3253 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { 3254 j_inode = ext4_get_journal_inode(sb, j_inum); 3255 if (j_inode) { 3256 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3257 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3258 iput(j_inode); 3259 } else { 3260 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3261 } 3262 } 3263 sbi->s_overhead = overhead; 3264 smp_wmb(); 3265 free_page((unsigned long) buf); 3266 return 0; 3267 } 3268 3269 static void ext4_set_resv_clusters(struct super_block *sb) 3270 { 3271 ext4_fsblk_t resv_clusters; 3272 struct ext4_sb_info *sbi = EXT4_SB(sb); 3273 3274 /* 3275 * There's no need to reserve anything when we aren't using extents. 3276 * The space estimates are exact, there are no unwritten extents, 3277 * hole punching doesn't need new metadata... This is needed especially 3278 * to keep ext2/3 backward compatibility. 3279 */ 3280 if (!ext4_has_feature_extents(sb)) 3281 return; 3282 /* 3283 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3284 * This should cover the situations where we can not afford to run 3285 * out of space like for example punch hole, or converting 3286 * unwritten extents in delalloc path. In most cases such 3287 * allocation would require 1, or 2 blocks, higher numbers are 3288 * very rare. 3289 */ 3290 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3291 sbi->s_cluster_bits); 3292 3293 do_div(resv_clusters, 50); 3294 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3295 3296 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3297 } 3298 3299 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3300 { 3301 char *orig_data = kstrdup(data, GFP_KERNEL); 3302 struct buffer_head *bh; 3303 struct ext4_super_block *es = NULL; 3304 struct ext4_sb_info *sbi; 3305 ext4_fsblk_t block; 3306 ext4_fsblk_t sb_block = get_sb_block(&data); 3307 ext4_fsblk_t logical_sb_block; 3308 unsigned long offset = 0; 3309 unsigned long journal_devnum = 0; 3310 unsigned long def_mount_opts; 3311 struct inode *root; 3312 const char *descr; 3313 int ret = -ENOMEM; 3314 int blocksize, clustersize; 3315 unsigned int db_count; 3316 unsigned int i; 3317 int needs_recovery, has_huge_files, has_bigalloc; 3318 __u64 blocks_count; 3319 int err = 0; 3320 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3321 ext4_group_t first_not_zeroed; 3322 3323 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3324 if (!sbi) 3325 goto out_free_orig; 3326 3327 sbi->s_blockgroup_lock = 3328 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3329 if (!sbi->s_blockgroup_lock) { 3330 kfree(sbi); 3331 goto out_free_orig; 3332 } 3333 sb->s_fs_info = sbi; 3334 sbi->s_sb = sb; 3335 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3336 sbi->s_sb_block = sb_block; 3337 if (sb->s_bdev->bd_part) 3338 sbi->s_sectors_written_start = 3339 part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3340 3341 /* Cleanup superblock name */ 3342 strreplace(sb->s_id, '/', '!'); 3343 3344 /* -EINVAL is default */ 3345 ret = -EINVAL; 3346 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3347 if (!blocksize) { 3348 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3349 goto out_fail; 3350 } 3351 3352 /* 3353 * The ext4 superblock will not be buffer aligned for other than 1kB 3354 * block sizes. We need to calculate the offset from buffer start. 3355 */ 3356 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3357 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3358 offset = do_div(logical_sb_block, blocksize); 3359 } else { 3360 logical_sb_block = sb_block; 3361 } 3362 3363 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3364 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3365 goto out_fail; 3366 } 3367 /* 3368 * Note: s_es must be initialized as soon as possible because 3369 * some ext4 macro-instructions depend on its value 3370 */ 3371 es = (struct ext4_super_block *) (bh->b_data + offset); 3372 sbi->s_es = es; 3373 sb->s_magic = le16_to_cpu(es->s_magic); 3374 if (sb->s_magic != EXT4_SUPER_MAGIC) 3375 goto cantfind_ext4; 3376 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3377 3378 /* Warn if metadata_csum and gdt_csum are both set. */ 3379 if (ext4_has_feature_metadata_csum(sb) && 3380 ext4_has_feature_gdt_csum(sb)) 3381 ext4_warning(sb, "metadata_csum and uninit_bg are " 3382 "redundant flags; please run fsck."); 3383 3384 /* Check for a known checksum algorithm */ 3385 if (!ext4_verify_csum_type(sb, es)) { 3386 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3387 "unknown checksum algorithm."); 3388 silent = 1; 3389 goto cantfind_ext4; 3390 } 3391 3392 /* Load the checksum driver */ 3393 if (ext4_has_feature_metadata_csum(sb)) { 3394 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3395 if (IS_ERR(sbi->s_chksum_driver)) { 3396 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3397 ret = PTR_ERR(sbi->s_chksum_driver); 3398 sbi->s_chksum_driver = NULL; 3399 goto failed_mount; 3400 } 3401 } 3402 3403 /* Check superblock checksum */ 3404 if (!ext4_superblock_csum_verify(sb, es)) { 3405 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3406 "invalid superblock checksum. Run e2fsck?"); 3407 silent = 1; 3408 ret = -EFSBADCRC; 3409 goto cantfind_ext4; 3410 } 3411 3412 /* Precompute checksum seed for all metadata */ 3413 if (ext4_has_feature_csum_seed(sb)) 3414 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3415 else if (ext4_has_metadata_csum(sb)) 3416 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3417 sizeof(es->s_uuid)); 3418 3419 /* Set defaults before we parse the mount options */ 3420 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3421 set_opt(sb, INIT_INODE_TABLE); 3422 if (def_mount_opts & EXT4_DEFM_DEBUG) 3423 set_opt(sb, DEBUG); 3424 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3425 set_opt(sb, GRPID); 3426 if (def_mount_opts & EXT4_DEFM_UID16) 3427 set_opt(sb, NO_UID32); 3428 /* xattr user namespace & acls are now defaulted on */ 3429 set_opt(sb, XATTR_USER); 3430 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3431 set_opt(sb, POSIX_ACL); 3432 #endif 3433 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 3434 if (ext4_has_metadata_csum(sb)) 3435 set_opt(sb, JOURNAL_CHECKSUM); 3436 3437 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3438 set_opt(sb, JOURNAL_DATA); 3439 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3440 set_opt(sb, ORDERED_DATA); 3441 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3442 set_opt(sb, WRITEBACK_DATA); 3443 3444 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3445 set_opt(sb, ERRORS_PANIC); 3446 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3447 set_opt(sb, ERRORS_CONT); 3448 else 3449 set_opt(sb, ERRORS_RO); 3450 /* block_validity enabled by default; disable with noblock_validity */ 3451 set_opt(sb, BLOCK_VALIDITY); 3452 if (def_mount_opts & EXT4_DEFM_DISCARD) 3453 set_opt(sb, DISCARD); 3454 3455 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3456 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3457 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3458 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3459 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3460 3461 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3462 set_opt(sb, BARRIER); 3463 3464 /* 3465 * enable delayed allocation by default 3466 * Use -o nodelalloc to turn it off 3467 */ 3468 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3469 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3470 set_opt(sb, DELALLOC); 3471 3472 /* 3473 * set default s_li_wait_mult for lazyinit, for the case there is 3474 * no mount option specified. 3475 */ 3476 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3477 3478 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3479 &journal_devnum, &journal_ioprio, 0)) { 3480 ext4_msg(sb, KERN_WARNING, 3481 "failed to parse options in superblock: %s", 3482 sbi->s_es->s_mount_opts); 3483 } 3484 sbi->s_def_mount_opt = sbi->s_mount_opt; 3485 if (!parse_options((char *) data, sb, &journal_devnum, 3486 &journal_ioprio, 0)) 3487 goto failed_mount; 3488 3489 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3490 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3491 "with data=journal disables delayed " 3492 "allocation and O_DIRECT support!\n"); 3493 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3494 ext4_msg(sb, KERN_ERR, "can't mount with " 3495 "both data=journal and delalloc"); 3496 goto failed_mount; 3497 } 3498 if (test_opt(sb, DIOREAD_NOLOCK)) { 3499 ext4_msg(sb, KERN_ERR, "can't mount with " 3500 "both data=journal and dioread_nolock"); 3501 goto failed_mount; 3502 } 3503 if (test_opt(sb, DAX)) { 3504 ext4_msg(sb, KERN_ERR, "can't mount with " 3505 "both data=journal and dax"); 3506 goto failed_mount; 3507 } 3508 if (test_opt(sb, DELALLOC)) 3509 clear_opt(sb, DELALLOC); 3510 } else { 3511 sb->s_iflags |= SB_I_CGROUPWB; 3512 } 3513 3514 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3515 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3516 3517 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3518 (ext4_has_compat_features(sb) || 3519 ext4_has_ro_compat_features(sb) || 3520 ext4_has_incompat_features(sb))) 3521 ext4_msg(sb, KERN_WARNING, 3522 "feature flags set on rev 0 fs, " 3523 "running e2fsck is recommended"); 3524 3525 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3526 set_opt2(sb, HURD_COMPAT); 3527 if (ext4_has_feature_64bit(sb)) { 3528 ext4_msg(sb, KERN_ERR, 3529 "The Hurd can't support 64-bit file systems"); 3530 goto failed_mount; 3531 } 3532 } 3533 3534 if (IS_EXT2_SB(sb)) { 3535 if (ext2_feature_set_ok(sb)) 3536 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3537 "using the ext4 subsystem"); 3538 else { 3539 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3540 "to feature incompatibilities"); 3541 goto failed_mount; 3542 } 3543 } 3544 3545 if (IS_EXT3_SB(sb)) { 3546 if (ext3_feature_set_ok(sb)) 3547 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3548 "using the ext4 subsystem"); 3549 else { 3550 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3551 "to feature incompatibilities"); 3552 goto failed_mount; 3553 } 3554 } 3555 3556 /* 3557 * Check feature flags regardless of the revision level, since we 3558 * previously didn't change the revision level when setting the flags, 3559 * so there is a chance incompat flags are set on a rev 0 filesystem. 3560 */ 3561 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3562 goto failed_mount; 3563 3564 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3565 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3566 blocksize > EXT4_MAX_BLOCK_SIZE) { 3567 ext4_msg(sb, KERN_ERR, 3568 "Unsupported filesystem blocksize %d", blocksize); 3569 goto failed_mount; 3570 } 3571 3572 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 3573 ext4_msg(sb, KERN_ERR, 3574 "Number of reserved GDT blocks insanely large: %d", 3575 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 3576 goto failed_mount; 3577 } 3578 3579 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 3580 err = bdev_dax_supported(sb, blocksize); 3581 if (err) 3582 goto failed_mount; 3583 } 3584 3585 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 3586 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 3587 es->s_encryption_level); 3588 goto failed_mount; 3589 } 3590 3591 if (sb->s_blocksize != blocksize) { 3592 /* Validate the filesystem blocksize */ 3593 if (!sb_set_blocksize(sb, blocksize)) { 3594 ext4_msg(sb, KERN_ERR, "bad block size %d", 3595 blocksize); 3596 goto failed_mount; 3597 } 3598 3599 brelse(bh); 3600 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3601 offset = do_div(logical_sb_block, blocksize); 3602 bh = sb_bread_unmovable(sb, logical_sb_block); 3603 if (!bh) { 3604 ext4_msg(sb, KERN_ERR, 3605 "Can't read superblock on 2nd try"); 3606 goto failed_mount; 3607 } 3608 es = (struct ext4_super_block *)(bh->b_data + offset); 3609 sbi->s_es = es; 3610 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3611 ext4_msg(sb, KERN_ERR, 3612 "Magic mismatch, very weird!"); 3613 goto failed_mount; 3614 } 3615 } 3616 3617 has_huge_files = ext4_has_feature_huge_file(sb); 3618 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3619 has_huge_files); 3620 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3621 3622 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3623 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3624 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3625 } else { 3626 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3627 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3628 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3629 (!is_power_of_2(sbi->s_inode_size)) || 3630 (sbi->s_inode_size > blocksize)) { 3631 ext4_msg(sb, KERN_ERR, 3632 "unsupported inode size: %d", 3633 sbi->s_inode_size); 3634 goto failed_mount; 3635 } 3636 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3637 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3638 } 3639 3640 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3641 if (ext4_has_feature_64bit(sb)) { 3642 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 3643 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3644 !is_power_of_2(sbi->s_desc_size)) { 3645 ext4_msg(sb, KERN_ERR, 3646 "unsupported descriptor size %lu", 3647 sbi->s_desc_size); 3648 goto failed_mount; 3649 } 3650 } else 3651 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 3652 3653 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3654 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3655 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3656 goto cantfind_ext4; 3657 3658 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3659 if (sbi->s_inodes_per_block == 0) 3660 goto cantfind_ext4; 3661 sbi->s_itb_per_group = sbi->s_inodes_per_group / 3662 sbi->s_inodes_per_block; 3663 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3664 sbi->s_sbh = bh; 3665 sbi->s_mount_state = le16_to_cpu(es->s_state); 3666 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3667 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 3668 3669 for (i = 0; i < 4; i++) 3670 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3671 sbi->s_def_hash_version = es->s_def_hash_version; 3672 if (ext4_has_feature_dir_index(sb)) { 3673 i = le32_to_cpu(es->s_flags); 3674 if (i & EXT2_FLAGS_UNSIGNED_HASH) 3675 sbi->s_hash_unsigned = 3; 3676 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3677 #ifdef __CHAR_UNSIGNED__ 3678 if (!(sb->s_flags & MS_RDONLY)) 3679 es->s_flags |= 3680 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3681 sbi->s_hash_unsigned = 3; 3682 #else 3683 if (!(sb->s_flags & MS_RDONLY)) 3684 es->s_flags |= 3685 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3686 #endif 3687 } 3688 } 3689 3690 /* Handle clustersize */ 3691 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3692 has_bigalloc = ext4_has_feature_bigalloc(sb); 3693 if (has_bigalloc) { 3694 if (clustersize < blocksize) { 3695 ext4_msg(sb, KERN_ERR, 3696 "cluster size (%d) smaller than " 3697 "block size (%d)", clustersize, blocksize); 3698 goto failed_mount; 3699 } 3700 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3701 le32_to_cpu(es->s_log_block_size); 3702 sbi->s_clusters_per_group = 3703 le32_to_cpu(es->s_clusters_per_group); 3704 if (sbi->s_clusters_per_group > blocksize * 8) { 3705 ext4_msg(sb, KERN_ERR, 3706 "#clusters per group too big: %lu", 3707 sbi->s_clusters_per_group); 3708 goto failed_mount; 3709 } 3710 if (sbi->s_blocks_per_group != 3711 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3712 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3713 "clusters per group (%lu) inconsistent", 3714 sbi->s_blocks_per_group, 3715 sbi->s_clusters_per_group); 3716 goto failed_mount; 3717 } 3718 } else { 3719 if (clustersize != blocksize) { 3720 ext4_warning(sb, "fragment/cluster size (%d) != " 3721 "block size (%d)", clustersize, 3722 blocksize); 3723 clustersize = blocksize; 3724 } 3725 if (sbi->s_blocks_per_group > blocksize * 8) { 3726 ext4_msg(sb, KERN_ERR, 3727 "#blocks per group too big: %lu", 3728 sbi->s_blocks_per_group); 3729 goto failed_mount; 3730 } 3731 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3732 sbi->s_cluster_bits = 0; 3733 } 3734 sbi->s_cluster_ratio = clustersize / blocksize; 3735 3736 if (sbi->s_inodes_per_group > blocksize * 8) { 3737 ext4_msg(sb, KERN_ERR, 3738 "#inodes per group too big: %lu", 3739 sbi->s_inodes_per_group); 3740 goto failed_mount; 3741 } 3742 3743 /* Do we have standard group size of clustersize * 8 blocks ? */ 3744 if (sbi->s_blocks_per_group == clustersize << 3) 3745 set_opt2(sb, STD_GROUP_SIZE); 3746 3747 /* 3748 * Test whether we have more sectors than will fit in sector_t, 3749 * and whether the max offset is addressable by the page cache. 3750 */ 3751 err = generic_check_addressable(sb->s_blocksize_bits, 3752 ext4_blocks_count(es)); 3753 if (err) { 3754 ext4_msg(sb, KERN_ERR, "filesystem" 3755 " too large to mount safely on this system"); 3756 if (sizeof(sector_t) < 8) 3757 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3758 goto failed_mount; 3759 } 3760 3761 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3762 goto cantfind_ext4; 3763 3764 /* check blocks count against device size */ 3765 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 3766 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3767 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3768 "exceeds size of device (%llu blocks)", 3769 ext4_blocks_count(es), blocks_count); 3770 goto failed_mount; 3771 } 3772 3773 /* 3774 * It makes no sense for the first data block to be beyond the end 3775 * of the filesystem. 3776 */ 3777 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3778 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 3779 "block %u is beyond end of filesystem (%llu)", 3780 le32_to_cpu(es->s_first_data_block), 3781 ext4_blocks_count(es)); 3782 goto failed_mount; 3783 } 3784 blocks_count = (ext4_blocks_count(es) - 3785 le32_to_cpu(es->s_first_data_block) + 3786 EXT4_BLOCKS_PER_GROUP(sb) - 1); 3787 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 3788 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3789 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 3790 "(block count %llu, first data block %u, " 3791 "blocks per group %lu)", sbi->s_groups_count, 3792 ext4_blocks_count(es), 3793 le32_to_cpu(es->s_first_data_block), 3794 EXT4_BLOCKS_PER_GROUP(sb)); 3795 goto failed_mount; 3796 } 3797 sbi->s_groups_count = blocks_count; 3798 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3799 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3800 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3801 EXT4_DESC_PER_BLOCK(sb); 3802 sbi->s_group_desc = ext4_kvmalloc(db_count * 3803 sizeof(struct buffer_head *), 3804 GFP_KERNEL); 3805 if (sbi->s_group_desc == NULL) { 3806 ext4_msg(sb, KERN_ERR, "not enough memory"); 3807 ret = -ENOMEM; 3808 goto failed_mount; 3809 } 3810 3811 bgl_lock_init(sbi->s_blockgroup_lock); 3812 3813 for (i = 0; i < db_count; i++) { 3814 block = descriptor_loc(sb, logical_sb_block, i); 3815 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 3816 if (!sbi->s_group_desc[i]) { 3817 ext4_msg(sb, KERN_ERR, 3818 "can't read group descriptor %d", i); 3819 db_count = i; 3820 goto failed_mount2; 3821 } 3822 } 3823 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 3824 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3825 ret = -EFSCORRUPTED; 3826 goto failed_mount2; 3827 } 3828 3829 sbi->s_gdb_count = db_count; 3830 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3831 spin_lock_init(&sbi->s_next_gen_lock); 3832 3833 setup_timer(&sbi->s_err_report, print_daily_error_info, 3834 (unsigned long) sb); 3835 3836 /* Register extent status tree shrinker */ 3837 if (ext4_es_register_shrinker(sbi)) 3838 goto failed_mount3; 3839 3840 sbi->s_stripe = ext4_get_stripe_size(sbi); 3841 sbi->s_extent_max_zeroout_kb = 32; 3842 3843 /* 3844 * set up enough so that it can read an inode 3845 */ 3846 sb->s_op = &ext4_sops; 3847 sb->s_export_op = &ext4_export_ops; 3848 sb->s_xattr = ext4_xattr_handlers; 3849 sb->s_cop = &ext4_cryptops; 3850 #ifdef CONFIG_QUOTA 3851 sb->dq_op = &ext4_quota_operations; 3852 if (ext4_has_feature_quota(sb)) 3853 sb->s_qcop = &dquot_quotactl_sysfile_ops; 3854 else 3855 sb->s_qcop = &ext4_qctl_operations; 3856 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3857 #endif 3858 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3859 3860 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 3861 mutex_init(&sbi->s_orphan_lock); 3862 3863 sb->s_root = NULL; 3864 3865 needs_recovery = (es->s_last_orphan != 0 || 3866 ext4_has_feature_journal_needs_recovery(sb)); 3867 3868 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY)) 3869 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 3870 goto failed_mount3a; 3871 3872 /* 3873 * The first inode we look at is the journal inode. Don't try 3874 * root first: it may be modified in the journal! 3875 */ 3876 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 3877 if (ext4_load_journal(sb, es, journal_devnum)) 3878 goto failed_mount3a; 3879 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 3880 ext4_has_feature_journal_needs_recovery(sb)) { 3881 ext4_msg(sb, KERN_ERR, "required journal recovery " 3882 "suppressed and not mounted read-only"); 3883 goto failed_mount_wq; 3884 } else { 3885 /* Nojournal mode, all journal mount options are illegal */ 3886 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 3887 ext4_msg(sb, KERN_ERR, "can't mount with " 3888 "journal_checksum, fs mounted w/o journal"); 3889 goto failed_mount_wq; 3890 } 3891 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3892 ext4_msg(sb, KERN_ERR, "can't mount with " 3893 "journal_async_commit, fs mounted w/o journal"); 3894 goto failed_mount_wq; 3895 } 3896 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 3897 ext4_msg(sb, KERN_ERR, "can't mount with " 3898 "commit=%lu, fs mounted w/o journal", 3899 sbi->s_commit_interval / HZ); 3900 goto failed_mount_wq; 3901 } 3902 if (EXT4_MOUNT_DATA_FLAGS & 3903 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 3904 ext4_msg(sb, KERN_ERR, "can't mount with " 3905 "data=, fs mounted w/o journal"); 3906 goto failed_mount_wq; 3907 } 3908 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM; 3909 clear_opt(sb, JOURNAL_CHECKSUM); 3910 clear_opt(sb, DATA_FLAGS); 3911 sbi->s_journal = NULL; 3912 needs_recovery = 0; 3913 goto no_journal; 3914 } 3915 3916 if (ext4_has_feature_64bit(sb) && 3917 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3918 JBD2_FEATURE_INCOMPAT_64BIT)) { 3919 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3920 goto failed_mount_wq; 3921 } 3922 3923 if (!set_journal_csum_feature_set(sb)) { 3924 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 3925 "feature set"); 3926 goto failed_mount_wq; 3927 } 3928 3929 /* We have now updated the journal if required, so we can 3930 * validate the data journaling mode. */ 3931 switch (test_opt(sb, DATA_FLAGS)) { 3932 case 0: 3933 /* No mode set, assume a default based on the journal 3934 * capabilities: ORDERED_DATA if the journal can 3935 * cope, else JOURNAL_DATA 3936 */ 3937 if (jbd2_journal_check_available_features 3938 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 3939 set_opt(sb, ORDERED_DATA); 3940 else 3941 set_opt(sb, JOURNAL_DATA); 3942 break; 3943 3944 case EXT4_MOUNT_ORDERED_DATA: 3945 case EXT4_MOUNT_WRITEBACK_DATA: 3946 if (!jbd2_journal_check_available_features 3947 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 3948 ext4_msg(sb, KERN_ERR, "Journal does not support " 3949 "requested data journaling mode"); 3950 goto failed_mount_wq; 3951 } 3952 default: 3953 break; 3954 } 3955 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 3956 3957 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 3958 3959 no_journal: 3960 sbi->s_mb_cache = ext4_xattr_create_cache(); 3961 if (!sbi->s_mb_cache) { 3962 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache"); 3963 goto failed_mount_wq; 3964 } 3965 3966 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && 3967 (blocksize != PAGE_SIZE)) { 3968 ext4_msg(sb, KERN_ERR, 3969 "Unsupported blocksize for fs encryption"); 3970 goto failed_mount_wq; 3971 } 3972 3973 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) && 3974 !ext4_has_feature_encrypt(sb)) { 3975 ext4_set_feature_encrypt(sb); 3976 ext4_commit_super(sb, 1); 3977 } 3978 3979 /* 3980 * Get the # of file system overhead blocks from the 3981 * superblock if present. 3982 */ 3983 if (es->s_overhead_clusters) 3984 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 3985 else { 3986 err = ext4_calculate_overhead(sb); 3987 if (err) 3988 goto failed_mount_wq; 3989 } 3990 3991 /* 3992 * The maximum number of concurrent works can be high and 3993 * concurrency isn't really necessary. Limit it to 1. 3994 */ 3995 EXT4_SB(sb)->rsv_conversion_wq = 3996 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 3997 if (!EXT4_SB(sb)->rsv_conversion_wq) { 3998 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 3999 ret = -ENOMEM; 4000 goto failed_mount4; 4001 } 4002 4003 /* 4004 * The jbd2_journal_load will have done any necessary log recovery, 4005 * so we can safely mount the rest of the filesystem now. 4006 */ 4007 4008 root = ext4_iget(sb, EXT4_ROOT_INO); 4009 if (IS_ERR(root)) { 4010 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4011 ret = PTR_ERR(root); 4012 root = NULL; 4013 goto failed_mount4; 4014 } 4015 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4016 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4017 iput(root); 4018 goto failed_mount4; 4019 } 4020 sb->s_root = d_make_root(root); 4021 if (!sb->s_root) { 4022 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4023 ret = -ENOMEM; 4024 goto failed_mount4; 4025 } 4026 4027 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY)) 4028 sb->s_flags |= MS_RDONLY; 4029 4030 /* determine the minimum size of new large inodes, if present */ 4031 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4032 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4033 EXT4_GOOD_OLD_INODE_SIZE; 4034 if (ext4_has_feature_extra_isize(sb)) { 4035 if (sbi->s_want_extra_isize < 4036 le16_to_cpu(es->s_want_extra_isize)) 4037 sbi->s_want_extra_isize = 4038 le16_to_cpu(es->s_want_extra_isize); 4039 if (sbi->s_want_extra_isize < 4040 le16_to_cpu(es->s_min_extra_isize)) 4041 sbi->s_want_extra_isize = 4042 le16_to_cpu(es->s_min_extra_isize); 4043 } 4044 } 4045 /* Check if enough inode space is available */ 4046 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 4047 sbi->s_inode_size) { 4048 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4049 EXT4_GOOD_OLD_INODE_SIZE; 4050 ext4_msg(sb, KERN_INFO, "required extra inode space not" 4051 "available"); 4052 } 4053 4054 ext4_set_resv_clusters(sb); 4055 4056 err = ext4_setup_system_zone(sb); 4057 if (err) { 4058 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4059 "zone (%d)", err); 4060 goto failed_mount4a; 4061 } 4062 4063 ext4_ext_init(sb); 4064 err = ext4_mb_init(sb); 4065 if (err) { 4066 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4067 err); 4068 goto failed_mount5; 4069 } 4070 4071 block = ext4_count_free_clusters(sb); 4072 ext4_free_blocks_count_set(sbi->s_es, 4073 EXT4_C2B(sbi, block)); 4074 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4075 GFP_KERNEL); 4076 if (!err) { 4077 unsigned long freei = ext4_count_free_inodes(sb); 4078 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4079 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4080 GFP_KERNEL); 4081 } 4082 if (!err) 4083 err = percpu_counter_init(&sbi->s_dirs_counter, 4084 ext4_count_dirs(sb), GFP_KERNEL); 4085 if (!err) 4086 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4087 GFP_KERNEL); 4088 if (!err) 4089 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 4090 4091 if (err) { 4092 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4093 goto failed_mount6; 4094 } 4095 4096 if (ext4_has_feature_flex_bg(sb)) 4097 if (!ext4_fill_flex_info(sb)) { 4098 ext4_msg(sb, KERN_ERR, 4099 "unable to initialize " 4100 "flex_bg meta info!"); 4101 goto failed_mount6; 4102 } 4103 4104 err = ext4_register_li_request(sb, first_not_zeroed); 4105 if (err) 4106 goto failed_mount6; 4107 4108 err = ext4_register_sysfs(sb); 4109 if (err) 4110 goto failed_mount7; 4111 4112 #ifdef CONFIG_QUOTA 4113 /* Enable quota usage during mount. */ 4114 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) { 4115 err = ext4_enable_quotas(sb); 4116 if (err) 4117 goto failed_mount8; 4118 } 4119 #endif /* CONFIG_QUOTA */ 4120 4121 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4122 ext4_orphan_cleanup(sb, es); 4123 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4124 if (needs_recovery) { 4125 ext4_msg(sb, KERN_INFO, "recovery complete"); 4126 ext4_mark_recovery_complete(sb, es); 4127 } 4128 if (EXT4_SB(sb)->s_journal) { 4129 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4130 descr = " journalled data mode"; 4131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4132 descr = " ordered data mode"; 4133 else 4134 descr = " writeback data mode"; 4135 } else 4136 descr = "out journal"; 4137 4138 if (test_opt(sb, DISCARD)) { 4139 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4140 if (!blk_queue_discard(q)) 4141 ext4_msg(sb, KERN_WARNING, 4142 "mounting with \"discard\" option, but " 4143 "the device does not support discard"); 4144 } 4145 4146 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4147 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4148 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 4149 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4150 4151 if (es->s_error_count) 4152 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4153 4154 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4155 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4156 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4157 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4158 4159 kfree(orig_data); 4160 #ifdef CONFIG_EXT4_FS_ENCRYPTION 4161 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX, 4162 EXT4_KEY_DESC_PREFIX_SIZE); 4163 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE; 4164 #endif 4165 return 0; 4166 4167 cantfind_ext4: 4168 if (!silent) 4169 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4170 goto failed_mount; 4171 4172 #ifdef CONFIG_QUOTA 4173 failed_mount8: 4174 ext4_unregister_sysfs(sb); 4175 #endif 4176 failed_mount7: 4177 ext4_unregister_li_request(sb); 4178 failed_mount6: 4179 ext4_mb_release(sb); 4180 if (sbi->s_flex_groups) 4181 kvfree(sbi->s_flex_groups); 4182 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4183 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4184 percpu_counter_destroy(&sbi->s_dirs_counter); 4185 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4186 failed_mount5: 4187 ext4_ext_release(sb); 4188 ext4_release_system_zone(sb); 4189 failed_mount4a: 4190 dput(sb->s_root); 4191 sb->s_root = NULL; 4192 failed_mount4: 4193 ext4_msg(sb, KERN_ERR, "mount failed"); 4194 if (EXT4_SB(sb)->rsv_conversion_wq) 4195 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4196 failed_mount_wq: 4197 if (sbi->s_mb_cache) { 4198 ext4_xattr_destroy_cache(sbi->s_mb_cache); 4199 sbi->s_mb_cache = NULL; 4200 } 4201 if (sbi->s_journal) { 4202 jbd2_journal_destroy(sbi->s_journal); 4203 sbi->s_journal = NULL; 4204 } 4205 failed_mount3a: 4206 ext4_es_unregister_shrinker(sbi); 4207 failed_mount3: 4208 del_timer_sync(&sbi->s_err_report); 4209 if (sbi->s_mmp_tsk) 4210 kthread_stop(sbi->s_mmp_tsk); 4211 failed_mount2: 4212 for (i = 0; i < db_count; i++) 4213 brelse(sbi->s_group_desc[i]); 4214 kvfree(sbi->s_group_desc); 4215 failed_mount: 4216 if (sbi->s_chksum_driver) 4217 crypto_free_shash(sbi->s_chksum_driver); 4218 #ifdef CONFIG_QUOTA 4219 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4220 kfree(sbi->s_qf_names[i]); 4221 #endif 4222 ext4_blkdev_remove(sbi); 4223 brelse(bh); 4224 out_fail: 4225 sb->s_fs_info = NULL; 4226 kfree(sbi->s_blockgroup_lock); 4227 kfree(sbi); 4228 out_free_orig: 4229 kfree(orig_data); 4230 return err ? err : ret; 4231 } 4232 4233 /* 4234 * Setup any per-fs journal parameters now. We'll do this both on 4235 * initial mount, once the journal has been initialised but before we've 4236 * done any recovery; and again on any subsequent remount. 4237 */ 4238 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4239 { 4240 struct ext4_sb_info *sbi = EXT4_SB(sb); 4241 4242 journal->j_commit_interval = sbi->s_commit_interval; 4243 journal->j_min_batch_time = sbi->s_min_batch_time; 4244 journal->j_max_batch_time = sbi->s_max_batch_time; 4245 4246 write_lock(&journal->j_state_lock); 4247 if (test_opt(sb, BARRIER)) 4248 journal->j_flags |= JBD2_BARRIER; 4249 else 4250 journal->j_flags &= ~JBD2_BARRIER; 4251 if (test_opt(sb, DATA_ERR_ABORT)) 4252 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4253 else 4254 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4255 write_unlock(&journal->j_state_lock); 4256 } 4257 4258 static struct inode *ext4_get_journal_inode(struct super_block *sb, 4259 unsigned int journal_inum) 4260 { 4261 struct inode *journal_inode; 4262 4263 /* 4264 * Test for the existence of a valid inode on disk. Bad things 4265 * happen if we iget() an unused inode, as the subsequent iput() 4266 * will try to delete it. 4267 */ 4268 journal_inode = ext4_iget(sb, journal_inum); 4269 if (IS_ERR(journal_inode)) { 4270 ext4_msg(sb, KERN_ERR, "no journal found"); 4271 return NULL; 4272 } 4273 if (!journal_inode->i_nlink) { 4274 make_bad_inode(journal_inode); 4275 iput(journal_inode); 4276 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4277 return NULL; 4278 } 4279 4280 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4281 journal_inode, journal_inode->i_size); 4282 if (!S_ISREG(journal_inode->i_mode)) { 4283 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4284 iput(journal_inode); 4285 return NULL; 4286 } 4287 return journal_inode; 4288 } 4289 4290 static journal_t *ext4_get_journal(struct super_block *sb, 4291 unsigned int journal_inum) 4292 { 4293 struct inode *journal_inode; 4294 journal_t *journal; 4295 4296 BUG_ON(!ext4_has_feature_journal(sb)); 4297 4298 journal_inode = ext4_get_journal_inode(sb, journal_inum); 4299 if (!journal_inode) 4300 return NULL; 4301 4302 journal = jbd2_journal_init_inode(journal_inode); 4303 if (!journal) { 4304 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4305 iput(journal_inode); 4306 return NULL; 4307 } 4308 journal->j_private = sb; 4309 ext4_init_journal_params(sb, journal); 4310 return journal; 4311 } 4312 4313 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4314 dev_t j_dev) 4315 { 4316 struct buffer_head *bh; 4317 journal_t *journal; 4318 ext4_fsblk_t start; 4319 ext4_fsblk_t len; 4320 int hblock, blocksize; 4321 ext4_fsblk_t sb_block; 4322 unsigned long offset; 4323 struct ext4_super_block *es; 4324 struct block_device *bdev; 4325 4326 BUG_ON(!ext4_has_feature_journal(sb)); 4327 4328 bdev = ext4_blkdev_get(j_dev, sb); 4329 if (bdev == NULL) 4330 return NULL; 4331 4332 blocksize = sb->s_blocksize; 4333 hblock = bdev_logical_block_size(bdev); 4334 if (blocksize < hblock) { 4335 ext4_msg(sb, KERN_ERR, 4336 "blocksize too small for journal device"); 4337 goto out_bdev; 4338 } 4339 4340 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4341 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4342 set_blocksize(bdev, blocksize); 4343 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4344 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4345 "external journal"); 4346 goto out_bdev; 4347 } 4348 4349 es = (struct ext4_super_block *) (bh->b_data + offset); 4350 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4351 !(le32_to_cpu(es->s_feature_incompat) & 4352 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4353 ext4_msg(sb, KERN_ERR, "external journal has " 4354 "bad superblock"); 4355 brelse(bh); 4356 goto out_bdev; 4357 } 4358 4359 if ((le32_to_cpu(es->s_feature_ro_compat) & 4360 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4361 es->s_checksum != ext4_superblock_csum(sb, es)) { 4362 ext4_msg(sb, KERN_ERR, "external journal has " 4363 "corrupt superblock"); 4364 brelse(bh); 4365 goto out_bdev; 4366 } 4367 4368 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4369 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4370 brelse(bh); 4371 goto out_bdev; 4372 } 4373 4374 len = ext4_blocks_count(es); 4375 start = sb_block + 1; 4376 brelse(bh); /* we're done with the superblock */ 4377 4378 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4379 start, len, blocksize); 4380 if (!journal) { 4381 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4382 goto out_bdev; 4383 } 4384 journal->j_private = sb; 4385 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4386 wait_on_buffer(journal->j_sb_buffer); 4387 if (!buffer_uptodate(journal->j_sb_buffer)) { 4388 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4389 goto out_journal; 4390 } 4391 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4392 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4393 "user (unsupported) - %d", 4394 be32_to_cpu(journal->j_superblock->s_nr_users)); 4395 goto out_journal; 4396 } 4397 EXT4_SB(sb)->journal_bdev = bdev; 4398 ext4_init_journal_params(sb, journal); 4399 return journal; 4400 4401 out_journal: 4402 jbd2_journal_destroy(journal); 4403 out_bdev: 4404 ext4_blkdev_put(bdev); 4405 return NULL; 4406 } 4407 4408 static int ext4_load_journal(struct super_block *sb, 4409 struct ext4_super_block *es, 4410 unsigned long journal_devnum) 4411 { 4412 journal_t *journal; 4413 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4414 dev_t journal_dev; 4415 int err = 0; 4416 int really_read_only; 4417 4418 BUG_ON(!ext4_has_feature_journal(sb)); 4419 4420 if (journal_devnum && 4421 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4422 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4423 "numbers have changed"); 4424 journal_dev = new_decode_dev(journal_devnum); 4425 } else 4426 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4427 4428 really_read_only = bdev_read_only(sb->s_bdev); 4429 4430 /* 4431 * Are we loading a blank journal or performing recovery after a 4432 * crash? For recovery, we need to check in advance whether we 4433 * can get read-write access to the device. 4434 */ 4435 if (ext4_has_feature_journal_needs_recovery(sb)) { 4436 if (sb->s_flags & MS_RDONLY) { 4437 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4438 "required on readonly filesystem"); 4439 if (really_read_only) { 4440 ext4_msg(sb, KERN_ERR, "write access " 4441 "unavailable, cannot proceed"); 4442 return -EROFS; 4443 } 4444 ext4_msg(sb, KERN_INFO, "write access will " 4445 "be enabled during recovery"); 4446 } 4447 } 4448 4449 if (journal_inum && journal_dev) { 4450 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4451 "and inode journals!"); 4452 return -EINVAL; 4453 } 4454 4455 if (journal_inum) { 4456 if (!(journal = ext4_get_journal(sb, journal_inum))) 4457 return -EINVAL; 4458 } else { 4459 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4460 return -EINVAL; 4461 } 4462 4463 if (!(journal->j_flags & JBD2_BARRIER)) 4464 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4465 4466 if (!ext4_has_feature_journal_needs_recovery(sb)) 4467 err = jbd2_journal_wipe(journal, !really_read_only); 4468 if (!err) { 4469 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4470 if (save) 4471 memcpy(save, ((char *) es) + 4472 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4473 err = jbd2_journal_load(journal); 4474 if (save) 4475 memcpy(((char *) es) + EXT4_S_ERR_START, 4476 save, EXT4_S_ERR_LEN); 4477 kfree(save); 4478 } 4479 4480 if (err) { 4481 ext4_msg(sb, KERN_ERR, "error loading journal"); 4482 jbd2_journal_destroy(journal); 4483 return err; 4484 } 4485 4486 EXT4_SB(sb)->s_journal = journal; 4487 ext4_clear_journal_err(sb, es); 4488 4489 if (!really_read_only && journal_devnum && 4490 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4491 es->s_journal_dev = cpu_to_le32(journal_devnum); 4492 4493 /* Make sure we flush the recovery flag to disk. */ 4494 ext4_commit_super(sb, 1); 4495 } 4496 4497 return 0; 4498 } 4499 4500 static int ext4_commit_super(struct super_block *sb, int sync) 4501 { 4502 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4503 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4504 int error = 0; 4505 4506 if (!sbh || block_device_ejected(sb)) 4507 return error; 4508 /* 4509 * If the file system is mounted read-only, don't update the 4510 * superblock write time. This avoids updating the superblock 4511 * write time when we are mounting the root file system 4512 * read/only but we need to replay the journal; at that point, 4513 * for people who are east of GMT and who make their clock 4514 * tick in localtime for Windows bug-for-bug compatibility, 4515 * the clock is set in the future, and this will cause e2fsck 4516 * to complain and force a full file system check. 4517 */ 4518 if (!(sb->s_flags & MS_RDONLY)) 4519 es->s_wtime = cpu_to_le32(get_seconds()); 4520 if (sb->s_bdev->bd_part) 4521 es->s_kbytes_written = 4522 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4523 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4524 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4525 else 4526 es->s_kbytes_written = 4527 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4528 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 4529 ext4_free_blocks_count_set(es, 4530 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4531 &EXT4_SB(sb)->s_freeclusters_counter))); 4532 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 4533 es->s_free_inodes_count = 4534 cpu_to_le32(percpu_counter_sum_positive( 4535 &EXT4_SB(sb)->s_freeinodes_counter)); 4536 BUFFER_TRACE(sbh, "marking dirty"); 4537 ext4_superblock_csum_set(sb); 4538 lock_buffer(sbh); 4539 if (buffer_write_io_error(sbh)) { 4540 /* 4541 * Oh, dear. A previous attempt to write the 4542 * superblock failed. This could happen because the 4543 * USB device was yanked out. Or it could happen to 4544 * be a transient write error and maybe the block will 4545 * be remapped. Nothing we can do but to retry the 4546 * write and hope for the best. 4547 */ 4548 ext4_msg(sb, KERN_ERR, "previous I/O error to " 4549 "superblock detected"); 4550 clear_buffer_write_io_error(sbh); 4551 set_buffer_uptodate(sbh); 4552 } 4553 mark_buffer_dirty(sbh); 4554 unlock_buffer(sbh); 4555 if (sync) { 4556 error = __sync_dirty_buffer(sbh, 4557 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC); 4558 if (error) 4559 return error; 4560 4561 error = buffer_write_io_error(sbh); 4562 if (error) { 4563 ext4_msg(sb, KERN_ERR, "I/O error while writing " 4564 "superblock"); 4565 clear_buffer_write_io_error(sbh); 4566 set_buffer_uptodate(sbh); 4567 } 4568 } 4569 return error; 4570 } 4571 4572 /* 4573 * Have we just finished recovery? If so, and if we are mounting (or 4574 * remounting) the filesystem readonly, then we will end up with a 4575 * consistent fs on disk. Record that fact. 4576 */ 4577 static void ext4_mark_recovery_complete(struct super_block *sb, 4578 struct ext4_super_block *es) 4579 { 4580 journal_t *journal = EXT4_SB(sb)->s_journal; 4581 4582 if (!ext4_has_feature_journal(sb)) { 4583 BUG_ON(journal != NULL); 4584 return; 4585 } 4586 jbd2_journal_lock_updates(journal); 4587 if (jbd2_journal_flush(journal) < 0) 4588 goto out; 4589 4590 if (ext4_has_feature_journal_needs_recovery(sb) && 4591 sb->s_flags & MS_RDONLY) { 4592 ext4_clear_feature_journal_needs_recovery(sb); 4593 ext4_commit_super(sb, 1); 4594 } 4595 4596 out: 4597 jbd2_journal_unlock_updates(journal); 4598 } 4599 4600 /* 4601 * If we are mounting (or read-write remounting) a filesystem whose journal 4602 * has recorded an error from a previous lifetime, move that error to the 4603 * main filesystem now. 4604 */ 4605 static void ext4_clear_journal_err(struct super_block *sb, 4606 struct ext4_super_block *es) 4607 { 4608 journal_t *journal; 4609 int j_errno; 4610 const char *errstr; 4611 4612 BUG_ON(!ext4_has_feature_journal(sb)); 4613 4614 journal = EXT4_SB(sb)->s_journal; 4615 4616 /* 4617 * Now check for any error status which may have been recorded in the 4618 * journal by a prior ext4_error() or ext4_abort() 4619 */ 4620 4621 j_errno = jbd2_journal_errno(journal); 4622 if (j_errno) { 4623 char nbuf[16]; 4624 4625 errstr = ext4_decode_error(sb, j_errno, nbuf); 4626 ext4_warning(sb, "Filesystem error recorded " 4627 "from previous mount: %s", errstr); 4628 ext4_warning(sb, "Marking fs in need of filesystem check."); 4629 4630 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4631 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4632 ext4_commit_super(sb, 1); 4633 4634 jbd2_journal_clear_err(journal); 4635 jbd2_journal_update_sb_errno(journal); 4636 } 4637 } 4638 4639 /* 4640 * Force the running and committing transactions to commit, 4641 * and wait on the commit. 4642 */ 4643 int ext4_force_commit(struct super_block *sb) 4644 { 4645 journal_t *journal; 4646 4647 if (sb->s_flags & MS_RDONLY) 4648 return 0; 4649 4650 journal = EXT4_SB(sb)->s_journal; 4651 return ext4_journal_force_commit(journal); 4652 } 4653 4654 static int ext4_sync_fs(struct super_block *sb, int wait) 4655 { 4656 int ret = 0; 4657 tid_t target; 4658 bool needs_barrier = false; 4659 struct ext4_sb_info *sbi = EXT4_SB(sb); 4660 4661 trace_ext4_sync_fs(sb, wait); 4662 flush_workqueue(sbi->rsv_conversion_wq); 4663 /* 4664 * Writeback quota in non-journalled quota case - journalled quota has 4665 * no dirty dquots 4666 */ 4667 dquot_writeback_dquots(sb, -1); 4668 /* 4669 * Data writeback is possible w/o journal transaction, so barrier must 4670 * being sent at the end of the function. But we can skip it if 4671 * transaction_commit will do it for us. 4672 */ 4673 if (sbi->s_journal) { 4674 target = jbd2_get_latest_transaction(sbi->s_journal); 4675 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 4676 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 4677 needs_barrier = true; 4678 4679 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4680 if (wait) 4681 ret = jbd2_log_wait_commit(sbi->s_journal, 4682 target); 4683 } 4684 } else if (wait && test_opt(sb, BARRIER)) 4685 needs_barrier = true; 4686 if (needs_barrier) { 4687 int err; 4688 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 4689 if (!ret) 4690 ret = err; 4691 } 4692 4693 return ret; 4694 } 4695 4696 /* 4697 * LVM calls this function before a (read-only) snapshot is created. This 4698 * gives us a chance to flush the journal completely and mark the fs clean. 4699 * 4700 * Note that only this function cannot bring a filesystem to be in a clean 4701 * state independently. It relies on upper layer to stop all data & metadata 4702 * modifications. 4703 */ 4704 static int ext4_freeze(struct super_block *sb) 4705 { 4706 int error = 0; 4707 journal_t *journal; 4708 4709 if (sb->s_flags & MS_RDONLY) 4710 return 0; 4711 4712 journal = EXT4_SB(sb)->s_journal; 4713 4714 if (journal) { 4715 /* Now we set up the journal barrier. */ 4716 jbd2_journal_lock_updates(journal); 4717 4718 /* 4719 * Don't clear the needs_recovery flag if we failed to 4720 * flush the journal. 4721 */ 4722 error = jbd2_journal_flush(journal); 4723 if (error < 0) 4724 goto out; 4725 4726 /* Journal blocked and flushed, clear needs_recovery flag. */ 4727 ext4_clear_feature_journal_needs_recovery(sb); 4728 } 4729 4730 error = ext4_commit_super(sb, 1); 4731 out: 4732 if (journal) 4733 /* we rely on upper layer to stop further updates */ 4734 jbd2_journal_unlock_updates(journal); 4735 return error; 4736 } 4737 4738 /* 4739 * Called by LVM after the snapshot is done. We need to reset the RECOVER 4740 * flag here, even though the filesystem is not technically dirty yet. 4741 */ 4742 static int ext4_unfreeze(struct super_block *sb) 4743 { 4744 if (sb->s_flags & MS_RDONLY) 4745 return 0; 4746 4747 if (EXT4_SB(sb)->s_journal) { 4748 /* Reset the needs_recovery flag before the fs is unlocked. */ 4749 ext4_set_feature_journal_needs_recovery(sb); 4750 } 4751 4752 ext4_commit_super(sb, 1); 4753 return 0; 4754 } 4755 4756 /* 4757 * Structure to save mount options for ext4_remount's benefit 4758 */ 4759 struct ext4_mount_options { 4760 unsigned long s_mount_opt; 4761 unsigned long s_mount_opt2; 4762 kuid_t s_resuid; 4763 kgid_t s_resgid; 4764 unsigned long s_commit_interval; 4765 u32 s_min_batch_time, s_max_batch_time; 4766 #ifdef CONFIG_QUOTA 4767 int s_jquota_fmt; 4768 char *s_qf_names[EXT4_MAXQUOTAS]; 4769 #endif 4770 }; 4771 4772 static int ext4_remount(struct super_block *sb, int *flags, char *data) 4773 { 4774 struct ext4_super_block *es; 4775 struct ext4_sb_info *sbi = EXT4_SB(sb); 4776 unsigned long old_sb_flags; 4777 struct ext4_mount_options old_opts; 4778 int enable_quota = 0; 4779 ext4_group_t g; 4780 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4781 int err = 0; 4782 #ifdef CONFIG_QUOTA 4783 int i, j; 4784 #endif 4785 char *orig_data = kstrdup(data, GFP_KERNEL); 4786 4787 /* Store the original options */ 4788 old_sb_flags = sb->s_flags; 4789 old_opts.s_mount_opt = sbi->s_mount_opt; 4790 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4791 old_opts.s_resuid = sbi->s_resuid; 4792 old_opts.s_resgid = sbi->s_resgid; 4793 old_opts.s_commit_interval = sbi->s_commit_interval; 4794 old_opts.s_min_batch_time = sbi->s_min_batch_time; 4795 old_opts.s_max_batch_time = sbi->s_max_batch_time; 4796 #ifdef CONFIG_QUOTA 4797 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4798 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4799 if (sbi->s_qf_names[i]) { 4800 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], 4801 GFP_KERNEL); 4802 if (!old_opts.s_qf_names[i]) { 4803 for (j = 0; j < i; j++) 4804 kfree(old_opts.s_qf_names[j]); 4805 kfree(orig_data); 4806 return -ENOMEM; 4807 } 4808 } else 4809 old_opts.s_qf_names[i] = NULL; 4810 #endif 4811 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4812 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4813 4814 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 4815 err = -EINVAL; 4816 goto restore_opts; 4817 } 4818 4819 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 4820 test_opt(sb, JOURNAL_CHECKSUM)) { 4821 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 4822 "during remount not supported; ignoring"); 4823 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 4824 } 4825 4826 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4827 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4828 ext4_msg(sb, KERN_ERR, "can't mount with " 4829 "both data=journal and delalloc"); 4830 err = -EINVAL; 4831 goto restore_opts; 4832 } 4833 if (test_opt(sb, DIOREAD_NOLOCK)) { 4834 ext4_msg(sb, KERN_ERR, "can't mount with " 4835 "both data=journal and dioread_nolock"); 4836 err = -EINVAL; 4837 goto restore_opts; 4838 } 4839 if (test_opt(sb, DAX)) { 4840 ext4_msg(sb, KERN_ERR, "can't mount with " 4841 "both data=journal and dax"); 4842 err = -EINVAL; 4843 goto restore_opts; 4844 } 4845 } 4846 4847 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 4848 ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 4849 "dax flag with busy inodes while remounting"); 4850 sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 4851 } 4852 4853 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4854 ext4_abort(sb, "Abort forced by user"); 4855 4856 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4857 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4858 4859 es = sbi->s_es; 4860 4861 if (sbi->s_journal) { 4862 ext4_init_journal_params(sb, sbi->s_journal); 4863 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4864 } 4865 4866 if (*flags & MS_LAZYTIME) 4867 sb->s_flags |= MS_LAZYTIME; 4868 4869 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 4870 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4871 err = -EROFS; 4872 goto restore_opts; 4873 } 4874 4875 if (*flags & MS_RDONLY) { 4876 err = sync_filesystem(sb); 4877 if (err < 0) 4878 goto restore_opts; 4879 err = dquot_suspend(sb, -1); 4880 if (err < 0) 4881 goto restore_opts; 4882 4883 /* 4884 * First of all, the unconditional stuff we have to do 4885 * to disable replay of the journal when we next remount 4886 */ 4887 sb->s_flags |= MS_RDONLY; 4888 4889 /* 4890 * OK, test if we are remounting a valid rw partition 4891 * readonly, and if so set the rdonly flag and then 4892 * mark the partition as valid again. 4893 */ 4894 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4895 (sbi->s_mount_state & EXT4_VALID_FS)) 4896 es->s_state = cpu_to_le16(sbi->s_mount_state); 4897 4898 if (sbi->s_journal) 4899 ext4_mark_recovery_complete(sb, es); 4900 } else { 4901 /* Make sure we can mount this feature set readwrite */ 4902 if (ext4_has_feature_readonly(sb) || 4903 !ext4_feature_set_ok(sb, 0)) { 4904 err = -EROFS; 4905 goto restore_opts; 4906 } 4907 /* 4908 * Make sure the group descriptor checksums 4909 * are sane. If they aren't, refuse to remount r/w. 4910 */ 4911 for (g = 0; g < sbi->s_groups_count; g++) { 4912 struct ext4_group_desc *gdp = 4913 ext4_get_group_desc(sb, g, NULL); 4914 4915 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 4916 ext4_msg(sb, KERN_ERR, 4917 "ext4_remount: Checksum for group %u failed (%u!=%u)", 4918 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 4919 le16_to_cpu(gdp->bg_checksum)); 4920 err = -EFSBADCRC; 4921 goto restore_opts; 4922 } 4923 } 4924 4925 /* 4926 * If we have an unprocessed orphan list hanging 4927 * around from a previously readonly bdev mount, 4928 * require a full umount/remount for now. 4929 */ 4930 if (es->s_last_orphan) { 4931 ext4_msg(sb, KERN_WARNING, "Couldn't " 4932 "remount RDWR because of unprocessed " 4933 "orphan inode list. Please " 4934 "umount/remount instead"); 4935 err = -EINVAL; 4936 goto restore_opts; 4937 } 4938 4939 /* 4940 * Mounting a RDONLY partition read-write, so reread 4941 * and store the current valid flag. (It may have 4942 * been changed by e2fsck since we originally mounted 4943 * the partition.) 4944 */ 4945 if (sbi->s_journal) 4946 ext4_clear_journal_err(sb, es); 4947 sbi->s_mount_state = le16_to_cpu(es->s_state); 4948 if (!ext4_setup_super(sb, es, 0)) 4949 sb->s_flags &= ~MS_RDONLY; 4950 if (ext4_has_feature_mmp(sb)) 4951 if (ext4_multi_mount_protect(sb, 4952 le64_to_cpu(es->s_mmp_block))) { 4953 err = -EROFS; 4954 goto restore_opts; 4955 } 4956 enable_quota = 1; 4957 } 4958 } 4959 4960 /* 4961 * Reinitialize lazy itable initialization thread based on 4962 * current settings 4963 */ 4964 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4965 ext4_unregister_li_request(sb); 4966 else { 4967 ext4_group_t first_not_zeroed; 4968 first_not_zeroed = ext4_has_uninit_itable(sb); 4969 ext4_register_li_request(sb, first_not_zeroed); 4970 } 4971 4972 ext4_setup_system_zone(sb); 4973 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY)) 4974 ext4_commit_super(sb, 1); 4975 4976 #ifdef CONFIG_QUOTA 4977 /* Release old quota file names */ 4978 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4979 kfree(old_opts.s_qf_names[i]); 4980 if (enable_quota) { 4981 if (sb_any_quota_suspended(sb)) 4982 dquot_resume(sb, -1); 4983 else if (ext4_has_feature_quota(sb)) { 4984 err = ext4_enable_quotas(sb); 4985 if (err) 4986 goto restore_opts; 4987 } 4988 } 4989 #endif 4990 4991 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME); 4992 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4993 kfree(orig_data); 4994 return 0; 4995 4996 restore_opts: 4997 sb->s_flags = old_sb_flags; 4998 sbi->s_mount_opt = old_opts.s_mount_opt; 4999 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5000 sbi->s_resuid = old_opts.s_resuid; 5001 sbi->s_resgid = old_opts.s_resgid; 5002 sbi->s_commit_interval = old_opts.s_commit_interval; 5003 sbi->s_min_batch_time = old_opts.s_min_batch_time; 5004 sbi->s_max_batch_time = old_opts.s_max_batch_time; 5005 #ifdef CONFIG_QUOTA 5006 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5007 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 5008 kfree(sbi->s_qf_names[i]); 5009 sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 5010 } 5011 #endif 5012 kfree(orig_data); 5013 return err; 5014 } 5015 5016 #ifdef CONFIG_QUOTA 5017 static int ext4_statfs_project(struct super_block *sb, 5018 kprojid_t projid, struct kstatfs *buf) 5019 { 5020 struct kqid qid; 5021 struct dquot *dquot; 5022 u64 limit; 5023 u64 curblock; 5024 5025 qid = make_kqid_projid(projid); 5026 dquot = dqget(sb, qid); 5027 if (IS_ERR(dquot)) 5028 return PTR_ERR(dquot); 5029 spin_lock(&dq_data_lock); 5030 5031 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 5032 dquot->dq_dqb.dqb_bsoftlimit : 5033 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 5034 if (limit && buf->f_blocks > limit) { 5035 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; 5036 buf->f_blocks = limit; 5037 buf->f_bfree = buf->f_bavail = 5038 (buf->f_blocks > curblock) ? 5039 (buf->f_blocks - curblock) : 0; 5040 } 5041 5042 limit = dquot->dq_dqb.dqb_isoftlimit ? 5043 dquot->dq_dqb.dqb_isoftlimit : 5044 dquot->dq_dqb.dqb_ihardlimit; 5045 if (limit && buf->f_files > limit) { 5046 buf->f_files = limit; 5047 buf->f_ffree = 5048 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5049 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5050 } 5051 5052 spin_unlock(&dq_data_lock); 5053 dqput(dquot); 5054 return 0; 5055 } 5056 #endif 5057 5058 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5059 { 5060 struct super_block *sb = dentry->d_sb; 5061 struct ext4_sb_info *sbi = EXT4_SB(sb); 5062 struct ext4_super_block *es = sbi->s_es; 5063 ext4_fsblk_t overhead = 0, resv_blocks; 5064 u64 fsid; 5065 s64 bfree; 5066 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5067 5068 if (!test_opt(sb, MINIX_DF)) 5069 overhead = sbi->s_overhead; 5070 5071 buf->f_type = EXT4_SUPER_MAGIC; 5072 buf->f_bsize = sb->s_blocksize; 5073 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5074 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5075 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5076 /* prevent underflow in case that few free space is available */ 5077 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5078 buf->f_bavail = buf->f_bfree - 5079 (ext4_r_blocks_count(es) + resv_blocks); 5080 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5081 buf->f_bavail = 0; 5082 buf->f_files = le32_to_cpu(es->s_inodes_count); 5083 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5084 buf->f_namelen = EXT4_NAME_LEN; 5085 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5086 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5087 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5088 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5089 5090 #ifdef CONFIG_QUOTA 5091 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5092 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5093 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5094 #endif 5095 return 0; 5096 } 5097 5098 /* Helper function for writing quotas on sync - we need to start transaction 5099 * before quota file is locked for write. Otherwise the are possible deadlocks: 5100 * Process 1 Process 2 5101 * ext4_create() quota_sync() 5102 * jbd2_journal_start() write_dquot() 5103 * dquot_initialize() down(dqio_mutex) 5104 * down(dqio_mutex) jbd2_journal_start() 5105 * 5106 */ 5107 5108 #ifdef CONFIG_QUOTA 5109 5110 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5111 { 5112 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5113 } 5114 5115 static int ext4_write_dquot(struct dquot *dquot) 5116 { 5117 int ret, err; 5118 handle_t *handle; 5119 struct inode *inode; 5120 5121 inode = dquot_to_inode(dquot); 5122 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5123 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5124 if (IS_ERR(handle)) 5125 return PTR_ERR(handle); 5126 ret = dquot_commit(dquot); 5127 err = ext4_journal_stop(handle); 5128 if (!ret) 5129 ret = err; 5130 return ret; 5131 } 5132 5133 static int ext4_acquire_dquot(struct dquot *dquot) 5134 { 5135 int ret, err; 5136 handle_t *handle; 5137 5138 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5139 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5140 if (IS_ERR(handle)) 5141 return PTR_ERR(handle); 5142 ret = dquot_acquire(dquot); 5143 err = ext4_journal_stop(handle); 5144 if (!ret) 5145 ret = err; 5146 return ret; 5147 } 5148 5149 static int ext4_release_dquot(struct dquot *dquot) 5150 { 5151 int ret, err; 5152 handle_t *handle; 5153 5154 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5155 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5156 if (IS_ERR(handle)) { 5157 /* Release dquot anyway to avoid endless cycle in dqput() */ 5158 dquot_release(dquot); 5159 return PTR_ERR(handle); 5160 } 5161 ret = dquot_release(dquot); 5162 err = ext4_journal_stop(handle); 5163 if (!ret) 5164 ret = err; 5165 return ret; 5166 } 5167 5168 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5169 { 5170 struct super_block *sb = dquot->dq_sb; 5171 struct ext4_sb_info *sbi = EXT4_SB(sb); 5172 5173 /* Are we journaling quotas? */ 5174 if (ext4_has_feature_quota(sb) || 5175 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5176 dquot_mark_dquot_dirty(dquot); 5177 return ext4_write_dquot(dquot); 5178 } else { 5179 return dquot_mark_dquot_dirty(dquot); 5180 } 5181 } 5182 5183 static int ext4_write_info(struct super_block *sb, int type) 5184 { 5185 int ret, err; 5186 handle_t *handle; 5187 5188 /* Data block + inode block */ 5189 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5190 if (IS_ERR(handle)) 5191 return PTR_ERR(handle); 5192 ret = dquot_commit_info(sb, type); 5193 err = ext4_journal_stop(handle); 5194 if (!ret) 5195 ret = err; 5196 return ret; 5197 } 5198 5199 /* 5200 * Turn on quotas during mount time - we need to find 5201 * the quota file and such... 5202 */ 5203 static int ext4_quota_on_mount(struct super_block *sb, int type) 5204 { 5205 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 5206 EXT4_SB(sb)->s_jquota_fmt, type); 5207 } 5208 5209 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5210 { 5211 struct ext4_inode_info *ei = EXT4_I(inode); 5212 5213 /* The first argument of lockdep_set_subclass has to be 5214 * *exactly* the same as the argument to init_rwsem() --- in 5215 * this case, in init_once() --- or lockdep gets unhappy 5216 * because the name of the lock is set using the 5217 * stringification of the argument to init_rwsem(). 5218 */ 5219 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5220 lockdep_set_subclass(&ei->i_data_sem, subclass); 5221 } 5222 5223 /* 5224 * Standard function to be called on quota_on 5225 */ 5226 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5227 struct path *path) 5228 { 5229 int err; 5230 5231 if (!test_opt(sb, QUOTA)) 5232 return -EINVAL; 5233 5234 /* Quotafile not on the same filesystem? */ 5235 if (path->dentry->d_sb != sb) 5236 return -EXDEV; 5237 /* Journaling quota? */ 5238 if (EXT4_SB(sb)->s_qf_names[type]) { 5239 /* Quotafile not in fs root? */ 5240 if (path->dentry->d_parent != sb->s_root) 5241 ext4_msg(sb, KERN_WARNING, 5242 "Quota file not on filesystem root. " 5243 "Journaled quota will not work"); 5244 } 5245 5246 /* 5247 * When we journal data on quota file, we have to flush journal to see 5248 * all updates to the file when we bypass pagecache... 5249 */ 5250 if (EXT4_SB(sb)->s_journal && 5251 ext4_should_journal_data(d_inode(path->dentry))) { 5252 /* 5253 * We don't need to lock updates but journal_flush() could 5254 * otherwise be livelocked... 5255 */ 5256 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5257 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5258 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5259 if (err) 5260 return err; 5261 } 5262 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5263 err = dquot_quota_on(sb, type, format_id, path); 5264 if (err) 5265 lockdep_set_quota_inode(path->dentry->d_inode, 5266 I_DATA_SEM_NORMAL); 5267 return err; 5268 } 5269 5270 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5271 unsigned int flags) 5272 { 5273 int err; 5274 struct inode *qf_inode; 5275 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5276 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5277 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5278 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5279 }; 5280 5281 BUG_ON(!ext4_has_feature_quota(sb)); 5282 5283 if (!qf_inums[type]) 5284 return -EPERM; 5285 5286 qf_inode = ext4_iget(sb, qf_inums[type]); 5287 if (IS_ERR(qf_inode)) { 5288 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5289 return PTR_ERR(qf_inode); 5290 } 5291 5292 /* Don't account quota for quota files to avoid recursion */ 5293 qf_inode->i_flags |= S_NOQUOTA; 5294 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 5295 err = dquot_enable(qf_inode, type, format_id, flags); 5296 iput(qf_inode); 5297 if (err) 5298 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 5299 5300 return err; 5301 } 5302 5303 /* Enable usage tracking for all quota types. */ 5304 static int ext4_enable_quotas(struct super_block *sb) 5305 { 5306 int type, err = 0; 5307 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5308 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5309 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5310 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5311 }; 5312 bool quota_mopt[EXT4_MAXQUOTAS] = { 5313 test_opt(sb, USRQUOTA), 5314 test_opt(sb, GRPQUOTA), 5315 test_opt(sb, PRJQUOTA), 5316 }; 5317 5318 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 5319 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 5320 if (qf_inums[type]) { 5321 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5322 DQUOT_USAGE_ENABLED | 5323 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 5324 if (err) { 5325 ext4_warning(sb, 5326 "Failed to enable quota tracking " 5327 "(type=%d, err=%d). Please run " 5328 "e2fsck to fix.", type, err); 5329 return err; 5330 } 5331 } 5332 } 5333 return 0; 5334 } 5335 5336 static int ext4_quota_off(struct super_block *sb, int type) 5337 { 5338 struct inode *inode = sb_dqopt(sb)->files[type]; 5339 handle_t *handle; 5340 5341 /* Force all delayed allocation blocks to be allocated. 5342 * Caller already holds s_umount sem */ 5343 if (test_opt(sb, DELALLOC)) 5344 sync_filesystem(sb); 5345 5346 if (!inode) 5347 goto out; 5348 5349 /* Update modification times of quota files when userspace can 5350 * start looking at them */ 5351 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5352 if (IS_ERR(handle)) 5353 goto out; 5354 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 5355 ext4_mark_inode_dirty(handle, inode); 5356 ext4_journal_stop(handle); 5357 5358 out: 5359 return dquot_quota_off(sb, type); 5360 } 5361 5362 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5363 * acquiring the locks... As quota files are never truncated and quota code 5364 * itself serializes the operations (and no one else should touch the files) 5365 * we don't have to be afraid of races */ 5366 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5367 size_t len, loff_t off) 5368 { 5369 struct inode *inode = sb_dqopt(sb)->files[type]; 5370 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5371 int offset = off & (sb->s_blocksize - 1); 5372 int tocopy; 5373 size_t toread; 5374 struct buffer_head *bh; 5375 loff_t i_size = i_size_read(inode); 5376 5377 if (off > i_size) 5378 return 0; 5379 if (off+len > i_size) 5380 len = i_size-off; 5381 toread = len; 5382 while (toread > 0) { 5383 tocopy = sb->s_blocksize - offset < toread ? 5384 sb->s_blocksize - offset : toread; 5385 bh = ext4_bread(NULL, inode, blk, 0); 5386 if (IS_ERR(bh)) 5387 return PTR_ERR(bh); 5388 if (!bh) /* A hole? */ 5389 memset(data, 0, tocopy); 5390 else 5391 memcpy(data, bh->b_data+offset, tocopy); 5392 brelse(bh); 5393 offset = 0; 5394 toread -= tocopy; 5395 data += tocopy; 5396 blk++; 5397 } 5398 return len; 5399 } 5400 5401 /* Write to quotafile (we know the transaction is already started and has 5402 * enough credits) */ 5403 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5404 const char *data, size_t len, loff_t off) 5405 { 5406 struct inode *inode = sb_dqopt(sb)->files[type]; 5407 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5408 int err, offset = off & (sb->s_blocksize - 1); 5409 int retries = 0; 5410 struct buffer_head *bh; 5411 handle_t *handle = journal_current_handle(); 5412 5413 if (EXT4_SB(sb)->s_journal && !handle) { 5414 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5415 " cancelled because transaction is not started", 5416 (unsigned long long)off, (unsigned long long)len); 5417 return -EIO; 5418 } 5419 /* 5420 * Since we account only one data block in transaction credits, 5421 * then it is impossible to cross a block boundary. 5422 */ 5423 if (sb->s_blocksize - offset < len) { 5424 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5425 " cancelled because not block aligned", 5426 (unsigned long long)off, (unsigned long long)len); 5427 return -EIO; 5428 } 5429 5430 do { 5431 bh = ext4_bread(handle, inode, blk, 5432 EXT4_GET_BLOCKS_CREATE | 5433 EXT4_GET_BLOCKS_METADATA_NOFAIL); 5434 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 5435 ext4_should_retry_alloc(inode->i_sb, &retries)); 5436 if (IS_ERR(bh)) 5437 return PTR_ERR(bh); 5438 if (!bh) 5439 goto out; 5440 BUFFER_TRACE(bh, "get write access"); 5441 err = ext4_journal_get_write_access(handle, bh); 5442 if (err) { 5443 brelse(bh); 5444 return err; 5445 } 5446 lock_buffer(bh); 5447 memcpy(bh->b_data+offset, data, len); 5448 flush_dcache_page(bh->b_page); 5449 unlock_buffer(bh); 5450 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5451 brelse(bh); 5452 out: 5453 if (inode->i_size < off + len) { 5454 i_size_write(inode, off + len); 5455 EXT4_I(inode)->i_disksize = inode->i_size; 5456 ext4_mark_inode_dirty(handle, inode); 5457 } 5458 return len; 5459 } 5460 5461 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) 5462 { 5463 const struct quota_format_ops *ops; 5464 5465 if (!sb_has_quota_loaded(sb, qid->type)) 5466 return -ESRCH; 5467 ops = sb_dqopt(sb)->ops[qid->type]; 5468 if (!ops || !ops->get_next_id) 5469 return -ENOSYS; 5470 return dquot_get_next_id(sb, qid); 5471 } 5472 #endif 5473 5474 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5475 const char *dev_name, void *data) 5476 { 5477 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5478 } 5479 5480 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 5481 static inline void register_as_ext2(void) 5482 { 5483 int err = register_filesystem(&ext2_fs_type); 5484 if (err) 5485 printk(KERN_WARNING 5486 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 5487 } 5488 5489 static inline void unregister_as_ext2(void) 5490 { 5491 unregister_filesystem(&ext2_fs_type); 5492 } 5493 5494 static inline int ext2_feature_set_ok(struct super_block *sb) 5495 { 5496 if (ext4_has_unknown_ext2_incompat_features(sb)) 5497 return 0; 5498 if (sb->s_flags & MS_RDONLY) 5499 return 1; 5500 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 5501 return 0; 5502 return 1; 5503 } 5504 #else 5505 static inline void register_as_ext2(void) { } 5506 static inline void unregister_as_ext2(void) { } 5507 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 5508 #endif 5509 5510 static inline void register_as_ext3(void) 5511 { 5512 int err = register_filesystem(&ext3_fs_type); 5513 if (err) 5514 printk(KERN_WARNING 5515 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 5516 } 5517 5518 static inline void unregister_as_ext3(void) 5519 { 5520 unregister_filesystem(&ext3_fs_type); 5521 } 5522 5523 static inline int ext3_feature_set_ok(struct super_block *sb) 5524 { 5525 if (ext4_has_unknown_ext3_incompat_features(sb)) 5526 return 0; 5527 if (!ext4_has_feature_journal(sb)) 5528 return 0; 5529 if (sb->s_flags & MS_RDONLY) 5530 return 1; 5531 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 5532 return 0; 5533 return 1; 5534 } 5535 5536 static struct file_system_type ext4_fs_type = { 5537 .owner = THIS_MODULE, 5538 .name = "ext4", 5539 .mount = ext4_mount, 5540 .kill_sb = kill_block_super, 5541 .fs_flags = FS_REQUIRES_DEV, 5542 }; 5543 MODULE_ALIAS_FS("ext4"); 5544 5545 /* Shared across all ext4 file systems */ 5546 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5547 5548 static int __init ext4_init_fs(void) 5549 { 5550 int i, err; 5551 5552 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 5553 ext4_li_info = NULL; 5554 mutex_init(&ext4_li_mtx); 5555 5556 /* Build-time check for flags consistency */ 5557 ext4_check_flag_values(); 5558 5559 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 5560 init_waitqueue_head(&ext4__ioend_wq[i]); 5561 5562 err = ext4_init_es(); 5563 if (err) 5564 return err; 5565 5566 err = ext4_init_pageio(); 5567 if (err) 5568 goto out5; 5569 5570 err = ext4_init_system_zone(); 5571 if (err) 5572 goto out4; 5573 5574 err = ext4_init_sysfs(); 5575 if (err) 5576 goto out3; 5577 5578 err = ext4_init_mballoc(); 5579 if (err) 5580 goto out2; 5581 err = init_inodecache(); 5582 if (err) 5583 goto out1; 5584 register_as_ext3(); 5585 register_as_ext2(); 5586 err = register_filesystem(&ext4_fs_type); 5587 if (err) 5588 goto out; 5589 5590 return 0; 5591 out: 5592 unregister_as_ext2(); 5593 unregister_as_ext3(); 5594 destroy_inodecache(); 5595 out1: 5596 ext4_exit_mballoc(); 5597 out2: 5598 ext4_exit_sysfs(); 5599 out3: 5600 ext4_exit_system_zone(); 5601 out4: 5602 ext4_exit_pageio(); 5603 out5: 5604 ext4_exit_es(); 5605 5606 return err; 5607 } 5608 5609 static void __exit ext4_exit_fs(void) 5610 { 5611 ext4_destroy_lazyinit_thread(); 5612 unregister_as_ext2(); 5613 unregister_as_ext3(); 5614 unregister_filesystem(&ext4_fs_type); 5615 destroy_inodecache(); 5616 ext4_exit_mballoc(); 5617 ext4_exit_sysfs(); 5618 ext4_exit_system_zone(); 5619 ext4_exit_pageio(); 5620 ext4_exit_es(); 5621 } 5622 5623 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 5624 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5625 MODULE_LICENSE("GPL"); 5626 module_init(ext4_init_fs) 5627 module_exit(ext4_exit_fs) 5628