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/jbd2.h> 25 #include <linux/slab.h> 26 #include <linux/init.h> 27 #include <linux/blkdev.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/proc_fs.h> 38 #include <linux/ctype.h> 39 #include <linux/log2.h> 40 #include <linux/crc16.h> 41 #include <linux/cleancache.h> 42 #include <asm/uaccess.h> 43 44 #include <linux/kthread.h> 45 #include <linux/freezer.h> 46 47 #include "ext4.h" 48 #include "ext4_extents.h" /* Needed for trace points definition */ 49 #include "ext4_jbd2.h" 50 #include "xattr.h" 51 #include "acl.h" 52 #include "mballoc.h" 53 54 #define CREATE_TRACE_POINTS 55 #include <trace/events/ext4.h> 56 57 static struct proc_dir_entry *ext4_proc_root; 58 static struct kset *ext4_kset; 59 static struct ext4_lazy_init *ext4_li_info; 60 static struct mutex ext4_li_mtx; 61 static struct ext4_features *ext4_feat; 62 static int ext4_mballoc_ready; 63 64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 65 unsigned long journal_devnum); 66 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 67 static int ext4_commit_super(struct super_block *sb, int sync); 68 static void ext4_mark_recovery_complete(struct super_block *sb, 69 struct ext4_super_block *es); 70 static void ext4_clear_journal_err(struct super_block *sb, 71 struct ext4_super_block *es); 72 static int ext4_sync_fs(struct super_block *sb, int wait); 73 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait); 74 static int ext4_remount(struct super_block *sb, int *flags, char *data); 75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76 static int ext4_unfreeze(struct super_block *sb); 77 static int ext4_freeze(struct super_block *sb); 78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79 const char *dev_name, void *data); 80 static inline int ext2_feature_set_ok(struct super_block *sb); 81 static inline int ext3_feature_set_ok(struct super_block *sb); 82 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83 static void ext4_destroy_lazyinit_thread(void); 84 static void ext4_unregister_li_request(struct super_block *sb); 85 static void ext4_clear_request_list(void); 86 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t); 87 88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 89 static struct file_system_type ext2_fs_type = { 90 .owner = THIS_MODULE, 91 .name = "ext2", 92 .mount = ext4_mount, 93 .kill_sb = kill_block_super, 94 .fs_flags = FS_REQUIRES_DEV, 95 }; 96 MODULE_ALIAS_FS("ext2"); 97 MODULE_ALIAS("ext2"); 98 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 99 #else 100 #define IS_EXT2_SB(sb) (0) 101 #endif 102 103 104 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 105 static struct file_system_type ext3_fs_type = { 106 .owner = THIS_MODULE, 107 .name = "ext3", 108 .mount = ext4_mount, 109 .kill_sb = kill_block_super, 110 .fs_flags = FS_REQUIRES_DEV, 111 }; 112 MODULE_ALIAS_FS("ext3"); 113 MODULE_ALIAS("ext3"); 114 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 115 #else 116 #define IS_EXT3_SB(sb) (0) 117 #endif 118 119 static int ext4_verify_csum_type(struct super_block *sb, 120 struct ext4_super_block *es) 121 { 122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 124 return 1; 125 126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 127 } 128 129 static __le32 ext4_superblock_csum(struct super_block *sb, 130 struct ext4_super_block *es) 131 { 132 struct ext4_sb_info *sbi = EXT4_SB(sb); 133 int offset = offsetof(struct ext4_super_block, s_checksum); 134 __u32 csum; 135 136 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 137 138 return cpu_to_le32(csum); 139 } 140 141 int ext4_superblock_csum_verify(struct super_block *sb, 142 struct ext4_super_block *es) 143 { 144 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 146 return 1; 147 148 return es->s_checksum == ext4_superblock_csum(sb, es); 149 } 150 151 void ext4_superblock_csum_set(struct super_block *sb) 152 { 153 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 154 155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 157 return; 158 159 es->s_checksum = ext4_superblock_csum(sb, es); 160 } 161 162 void *ext4_kvmalloc(size_t size, gfp_t flags) 163 { 164 void *ret; 165 166 ret = kmalloc(size, flags | __GFP_NOWARN); 167 if (!ret) 168 ret = __vmalloc(size, flags, PAGE_KERNEL); 169 return ret; 170 } 171 172 void *ext4_kvzalloc(size_t size, gfp_t flags) 173 { 174 void *ret; 175 176 ret = kzalloc(size, flags | __GFP_NOWARN); 177 if (!ret) 178 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 179 return ret; 180 } 181 182 void ext4_kvfree(void *ptr) 183 { 184 if (is_vmalloc_addr(ptr)) 185 vfree(ptr); 186 else 187 kfree(ptr); 188 189 } 190 191 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 192 struct ext4_group_desc *bg) 193 { 194 return le32_to_cpu(bg->bg_block_bitmap_lo) | 195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 196 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 197 } 198 199 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 200 struct ext4_group_desc *bg) 201 { 202 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 204 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 205 } 206 207 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 208 struct ext4_group_desc *bg) 209 { 210 return le32_to_cpu(bg->bg_inode_table_lo) | 211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 213 } 214 215 __u32 ext4_free_group_clusters(struct super_block *sb, 216 struct ext4_group_desc *bg) 217 { 218 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 220 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 221 } 222 223 __u32 ext4_free_inodes_count(struct super_block *sb, 224 struct ext4_group_desc *bg) 225 { 226 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 228 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 229 } 230 231 __u32 ext4_used_dirs_count(struct super_block *sb, 232 struct ext4_group_desc *bg) 233 { 234 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 236 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 237 } 238 239 __u32 ext4_itable_unused_count(struct super_block *sb, 240 struct ext4_group_desc *bg) 241 { 242 return le16_to_cpu(bg->bg_itable_unused_lo) | 243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 244 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 245 } 246 247 void ext4_block_bitmap_set(struct super_block *sb, 248 struct ext4_group_desc *bg, ext4_fsblk_t blk) 249 { 250 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 252 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 253 } 254 255 void ext4_inode_bitmap_set(struct super_block *sb, 256 struct ext4_group_desc *bg, ext4_fsblk_t blk) 257 { 258 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 260 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 261 } 262 263 void ext4_inode_table_set(struct super_block *sb, 264 struct ext4_group_desc *bg, ext4_fsblk_t blk) 265 { 266 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 268 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 269 } 270 271 void ext4_free_group_clusters_set(struct super_block *sb, 272 struct ext4_group_desc *bg, __u32 count) 273 { 274 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 276 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 277 } 278 279 void ext4_free_inodes_set(struct super_block *sb, 280 struct ext4_group_desc *bg, __u32 count) 281 { 282 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 284 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 285 } 286 287 void ext4_used_dirs_set(struct super_block *sb, 288 struct ext4_group_desc *bg, __u32 count) 289 { 290 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 292 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 293 } 294 295 void ext4_itable_unused_set(struct super_block *sb, 296 struct ext4_group_desc *bg, __u32 count) 297 { 298 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 300 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 301 } 302 303 304 static void __save_error_info(struct super_block *sb, const char *func, 305 unsigned int line) 306 { 307 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 308 309 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 311 es->s_last_error_time = cpu_to_le32(get_seconds()); 312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 313 es->s_last_error_line = cpu_to_le32(line); 314 if (!es->s_first_error_time) { 315 es->s_first_error_time = es->s_last_error_time; 316 strncpy(es->s_first_error_func, func, 317 sizeof(es->s_first_error_func)); 318 es->s_first_error_line = cpu_to_le32(line); 319 es->s_first_error_ino = es->s_last_error_ino; 320 es->s_first_error_block = es->s_last_error_block; 321 } 322 /* 323 * Start the daily error reporting function if it hasn't been 324 * started already 325 */ 326 if (!es->s_error_count) 327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 328 le32_add_cpu(&es->s_error_count, 1); 329 } 330 331 static void save_error_info(struct super_block *sb, const char *func, 332 unsigned int line) 333 { 334 __save_error_info(sb, func, line); 335 ext4_commit_super(sb, 1); 336 } 337 338 /* 339 * The del_gendisk() function uninitializes the disk-specific data 340 * structures, including the bdi structure, without telling anyone 341 * else. Once this happens, any attempt to call mark_buffer_dirty() 342 * (for example, by ext4_commit_super), will cause a kernel OOPS. 343 * This is a kludge to prevent these oops until we can put in a proper 344 * hook in del_gendisk() to inform the VFS and file system layers. 345 */ 346 static int block_device_ejected(struct super_block *sb) 347 { 348 struct inode *bd_inode = sb->s_bdev->bd_inode; 349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info; 350 351 return bdi->dev == NULL; 352 } 353 354 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 355 { 356 struct super_block *sb = journal->j_private; 357 struct ext4_sb_info *sbi = EXT4_SB(sb); 358 int error = is_journal_aborted(journal); 359 struct ext4_journal_cb_entry *jce; 360 361 BUG_ON(txn->t_state == T_FINISHED); 362 spin_lock(&sbi->s_md_lock); 363 while (!list_empty(&txn->t_private_list)) { 364 jce = list_entry(txn->t_private_list.next, 365 struct ext4_journal_cb_entry, jce_list); 366 list_del_init(&jce->jce_list); 367 spin_unlock(&sbi->s_md_lock); 368 jce->jce_func(sb, jce, error); 369 spin_lock(&sbi->s_md_lock); 370 } 371 spin_unlock(&sbi->s_md_lock); 372 } 373 374 /* Deal with the reporting of failure conditions on a filesystem such as 375 * inconsistencies detected or read IO failures. 376 * 377 * On ext2, we can store the error state of the filesystem in the 378 * superblock. That is not possible on ext4, because we may have other 379 * write ordering constraints on the superblock which prevent us from 380 * writing it out straight away; and given that the journal is about to 381 * be aborted, we can't rely on the current, or future, transactions to 382 * write out the superblock safely. 383 * 384 * We'll just use the jbd2_journal_abort() error code to record an error in 385 * the journal instead. On recovery, the journal will complain about 386 * that error until we've noted it down and cleared it. 387 */ 388 389 static void ext4_handle_error(struct super_block *sb) 390 { 391 if (sb->s_flags & MS_RDONLY) 392 return; 393 394 if (!test_opt(sb, ERRORS_CONT)) { 395 journal_t *journal = EXT4_SB(sb)->s_journal; 396 397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 398 if (journal) 399 jbd2_journal_abort(journal, -EIO); 400 } 401 if (test_opt(sb, ERRORS_RO)) { 402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 403 /* 404 * Make sure updated value of ->s_mount_flags will be visible 405 * before ->s_flags update 406 */ 407 smp_wmb(); 408 sb->s_flags |= MS_RDONLY; 409 } 410 if (test_opt(sb, ERRORS_PANIC)) 411 panic("EXT4-fs (device %s): panic forced after error\n", 412 sb->s_id); 413 } 414 415 #define ext4_error_ratelimit(sb) \ 416 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 417 "EXT4-fs error") 418 419 void __ext4_error(struct super_block *sb, const char *function, 420 unsigned int line, const char *fmt, ...) 421 { 422 struct va_format vaf; 423 va_list args; 424 425 if (ext4_error_ratelimit(sb)) { 426 va_start(args, fmt); 427 vaf.fmt = fmt; 428 vaf.va = &args; 429 printk(KERN_CRIT 430 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 431 sb->s_id, function, line, current->comm, &vaf); 432 va_end(args); 433 } 434 save_error_info(sb, function, line); 435 ext4_handle_error(sb); 436 } 437 438 void __ext4_error_inode(struct inode *inode, const char *function, 439 unsigned int line, ext4_fsblk_t block, 440 const char *fmt, ...) 441 { 442 va_list args; 443 struct va_format vaf; 444 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 445 446 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 447 es->s_last_error_block = cpu_to_le64(block); 448 if (ext4_error_ratelimit(inode->i_sb)) { 449 va_start(args, fmt); 450 vaf.fmt = fmt; 451 vaf.va = &args; 452 if (block) 453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 454 "inode #%lu: block %llu: comm %s: %pV\n", 455 inode->i_sb->s_id, function, line, inode->i_ino, 456 block, current->comm, &vaf); 457 else 458 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 459 "inode #%lu: comm %s: %pV\n", 460 inode->i_sb->s_id, function, line, inode->i_ino, 461 current->comm, &vaf); 462 va_end(args); 463 } 464 save_error_info(inode->i_sb, function, line); 465 ext4_handle_error(inode->i_sb); 466 } 467 468 void __ext4_error_file(struct file *file, const char *function, 469 unsigned int line, ext4_fsblk_t block, 470 const char *fmt, ...) 471 { 472 va_list args; 473 struct va_format vaf; 474 struct ext4_super_block *es; 475 struct inode *inode = file_inode(file); 476 char pathname[80], *path; 477 478 es = EXT4_SB(inode->i_sb)->s_es; 479 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 480 if (ext4_error_ratelimit(inode->i_sb)) { 481 path = d_path(&(file->f_path), pathname, sizeof(pathname)); 482 if (IS_ERR(path)) 483 path = "(unknown)"; 484 va_start(args, fmt); 485 vaf.fmt = fmt; 486 vaf.va = &args; 487 if (block) 488 printk(KERN_CRIT 489 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 490 "block %llu: comm %s: path %s: %pV\n", 491 inode->i_sb->s_id, function, line, inode->i_ino, 492 block, current->comm, path, &vaf); 493 else 494 printk(KERN_CRIT 495 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 496 "comm %s: path %s: %pV\n", 497 inode->i_sb->s_id, function, line, inode->i_ino, 498 current->comm, path, &vaf); 499 va_end(args); 500 } 501 save_error_info(inode->i_sb, function, line); 502 ext4_handle_error(inode->i_sb); 503 } 504 505 const char *ext4_decode_error(struct super_block *sb, int errno, 506 char nbuf[16]) 507 { 508 char *errstr = NULL; 509 510 switch (errno) { 511 case -EIO: 512 errstr = "IO failure"; 513 break; 514 case -ENOMEM: 515 errstr = "Out of memory"; 516 break; 517 case -EROFS: 518 if (!sb || (EXT4_SB(sb)->s_journal && 519 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 520 errstr = "Journal has aborted"; 521 else 522 errstr = "Readonly filesystem"; 523 break; 524 default: 525 /* If the caller passed in an extra buffer for unknown 526 * errors, textualise them now. Else we just return 527 * NULL. */ 528 if (nbuf) { 529 /* Check for truncated error codes... */ 530 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 531 errstr = nbuf; 532 } 533 break; 534 } 535 536 return errstr; 537 } 538 539 /* __ext4_std_error decodes expected errors from journaling functions 540 * automatically and invokes the appropriate error response. */ 541 542 void __ext4_std_error(struct super_block *sb, const char *function, 543 unsigned int line, int errno) 544 { 545 char nbuf[16]; 546 const char *errstr; 547 548 /* Special case: if the error is EROFS, and we're not already 549 * inside a transaction, then there's really no point in logging 550 * an error. */ 551 if (errno == -EROFS && journal_current_handle() == NULL && 552 (sb->s_flags & MS_RDONLY)) 553 return; 554 555 if (ext4_error_ratelimit(sb)) { 556 errstr = ext4_decode_error(sb, errno, nbuf); 557 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 558 sb->s_id, function, line, errstr); 559 } 560 561 save_error_info(sb, function, line); 562 ext4_handle_error(sb); 563 } 564 565 /* 566 * ext4_abort is a much stronger failure handler than ext4_error. The 567 * abort function may be used to deal with unrecoverable failures such 568 * as journal IO errors or ENOMEM at a critical moment in log management. 569 * 570 * We unconditionally force the filesystem into an ABORT|READONLY state, 571 * unless the error response on the fs has been set to panic in which 572 * case we take the easy way out and panic immediately. 573 */ 574 575 void __ext4_abort(struct super_block *sb, const char *function, 576 unsigned int line, const char *fmt, ...) 577 { 578 va_list args; 579 580 save_error_info(sb, function, line); 581 va_start(args, fmt); 582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, 583 function, line); 584 vprintk(fmt, args); 585 printk("\n"); 586 va_end(args); 587 588 if ((sb->s_flags & MS_RDONLY) == 0) { 589 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 590 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 591 /* 592 * Make sure updated value of ->s_mount_flags will be visible 593 * before ->s_flags update 594 */ 595 smp_wmb(); 596 sb->s_flags |= MS_RDONLY; 597 if (EXT4_SB(sb)->s_journal) 598 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 599 save_error_info(sb, function, line); 600 } 601 if (test_opt(sb, ERRORS_PANIC)) 602 panic("EXT4-fs panic from previous error\n"); 603 } 604 605 void __ext4_msg(struct super_block *sb, 606 const char *prefix, const char *fmt, ...) 607 { 608 struct va_format vaf; 609 va_list args; 610 611 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 612 return; 613 614 va_start(args, fmt); 615 vaf.fmt = fmt; 616 vaf.va = &args; 617 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 618 va_end(args); 619 } 620 621 void __ext4_warning(struct super_block *sb, const char *function, 622 unsigned int line, const char *fmt, ...) 623 { 624 struct va_format vaf; 625 va_list args; 626 627 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 628 "EXT4-fs warning")) 629 return; 630 631 va_start(args, fmt); 632 vaf.fmt = fmt; 633 vaf.va = &args; 634 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 635 sb->s_id, function, line, &vaf); 636 va_end(args); 637 } 638 639 void __ext4_grp_locked_error(const char *function, unsigned int line, 640 struct super_block *sb, ext4_group_t grp, 641 unsigned long ino, ext4_fsblk_t block, 642 const char *fmt, ...) 643 __releases(bitlock) 644 __acquires(bitlock) 645 { 646 struct va_format vaf; 647 va_list args; 648 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 649 650 es->s_last_error_ino = cpu_to_le32(ino); 651 es->s_last_error_block = cpu_to_le64(block); 652 __save_error_info(sb, function, line); 653 654 if (ext4_error_ratelimit(sb)) { 655 va_start(args, fmt); 656 vaf.fmt = fmt; 657 vaf.va = &args; 658 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 659 sb->s_id, function, line, grp); 660 if (ino) 661 printk(KERN_CONT "inode %lu: ", ino); 662 if (block) 663 printk(KERN_CONT "block %llu:", 664 (unsigned long long) block); 665 printk(KERN_CONT "%pV\n", &vaf); 666 va_end(args); 667 } 668 669 if (test_opt(sb, ERRORS_CONT)) { 670 ext4_commit_super(sb, 0); 671 return; 672 } 673 674 ext4_unlock_group(sb, grp); 675 ext4_handle_error(sb); 676 /* 677 * We only get here in the ERRORS_RO case; relocking the group 678 * may be dangerous, but nothing bad will happen since the 679 * filesystem will have already been marked read/only and the 680 * journal has been aborted. We return 1 as a hint to callers 681 * who might what to use the return value from 682 * ext4_grp_locked_error() to distinguish between the 683 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 684 * aggressively from the ext4 function in question, with a 685 * more appropriate error code. 686 */ 687 ext4_lock_group(sb, grp); 688 return; 689 } 690 691 void ext4_update_dynamic_rev(struct super_block *sb) 692 { 693 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 694 695 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 696 return; 697 698 ext4_warning(sb, 699 "updating to rev %d because of new feature flag, " 700 "running e2fsck is recommended", 701 EXT4_DYNAMIC_REV); 702 703 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 704 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 705 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 706 /* leave es->s_feature_*compat flags alone */ 707 /* es->s_uuid will be set by e2fsck if empty */ 708 709 /* 710 * The rest of the superblock fields should be zero, and if not it 711 * means they are likely already in use, so leave them alone. We 712 * can leave it up to e2fsck to clean up any inconsistencies there. 713 */ 714 } 715 716 /* 717 * Open the external journal device 718 */ 719 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 720 { 721 struct block_device *bdev; 722 char b[BDEVNAME_SIZE]; 723 724 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 725 if (IS_ERR(bdev)) 726 goto fail; 727 return bdev; 728 729 fail: 730 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 731 __bdevname(dev, b), PTR_ERR(bdev)); 732 return NULL; 733 } 734 735 /* 736 * Release the journal device 737 */ 738 static void ext4_blkdev_put(struct block_device *bdev) 739 { 740 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 741 } 742 743 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 744 { 745 struct block_device *bdev; 746 bdev = sbi->journal_bdev; 747 if (bdev) { 748 ext4_blkdev_put(bdev); 749 sbi->journal_bdev = NULL; 750 } 751 } 752 753 static inline struct inode *orphan_list_entry(struct list_head *l) 754 { 755 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 756 } 757 758 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 759 { 760 struct list_head *l; 761 762 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 763 le32_to_cpu(sbi->s_es->s_last_orphan)); 764 765 printk(KERN_ERR "sb_info orphan list:\n"); 766 list_for_each(l, &sbi->s_orphan) { 767 struct inode *inode = orphan_list_entry(l); 768 printk(KERN_ERR " " 769 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 770 inode->i_sb->s_id, inode->i_ino, inode, 771 inode->i_mode, inode->i_nlink, 772 NEXT_ORPHAN(inode)); 773 } 774 } 775 776 static void ext4_put_super(struct super_block *sb) 777 { 778 struct ext4_sb_info *sbi = EXT4_SB(sb); 779 struct ext4_super_block *es = sbi->s_es; 780 int i, err; 781 782 ext4_unregister_li_request(sb); 783 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 784 785 flush_workqueue(sbi->rsv_conversion_wq); 786 destroy_workqueue(sbi->rsv_conversion_wq); 787 788 if (sbi->s_journal) { 789 err = jbd2_journal_destroy(sbi->s_journal); 790 sbi->s_journal = NULL; 791 if (err < 0) 792 ext4_abort(sb, "Couldn't clean up the journal"); 793 } 794 795 ext4_es_unregister_shrinker(sbi); 796 del_timer_sync(&sbi->s_err_report); 797 ext4_release_system_zone(sb); 798 ext4_mb_release(sb); 799 ext4_ext_release(sb); 800 ext4_xattr_put_super(sb); 801 802 if (!(sb->s_flags & MS_RDONLY)) { 803 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 804 es->s_state = cpu_to_le16(sbi->s_mount_state); 805 } 806 if (!(sb->s_flags & MS_RDONLY)) 807 ext4_commit_super(sb, 1); 808 809 if (sbi->s_proc) { 810 remove_proc_entry("options", sbi->s_proc); 811 remove_proc_entry(sb->s_id, ext4_proc_root); 812 } 813 kobject_del(&sbi->s_kobj); 814 815 for (i = 0; i < sbi->s_gdb_count; i++) 816 brelse(sbi->s_group_desc[i]); 817 ext4_kvfree(sbi->s_group_desc); 818 ext4_kvfree(sbi->s_flex_groups); 819 percpu_counter_destroy(&sbi->s_freeclusters_counter); 820 percpu_counter_destroy(&sbi->s_freeinodes_counter); 821 percpu_counter_destroy(&sbi->s_dirs_counter); 822 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 823 percpu_counter_destroy(&sbi->s_extent_cache_cnt); 824 brelse(sbi->s_sbh); 825 #ifdef CONFIG_QUOTA 826 for (i = 0; i < MAXQUOTAS; i++) 827 kfree(sbi->s_qf_names[i]); 828 #endif 829 830 /* Debugging code just in case the in-memory inode orphan list 831 * isn't empty. The on-disk one can be non-empty if we've 832 * detected an error and taken the fs readonly, but the 833 * in-memory list had better be clean by this point. */ 834 if (!list_empty(&sbi->s_orphan)) 835 dump_orphan_list(sb, sbi); 836 J_ASSERT(list_empty(&sbi->s_orphan)); 837 838 invalidate_bdev(sb->s_bdev); 839 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 840 /* 841 * Invalidate the journal device's buffers. We don't want them 842 * floating about in memory - the physical journal device may 843 * hotswapped, and it breaks the `ro-after' testing code. 844 */ 845 sync_blockdev(sbi->journal_bdev); 846 invalidate_bdev(sbi->journal_bdev); 847 ext4_blkdev_remove(sbi); 848 } 849 if (sbi->s_mb_cache) { 850 ext4_xattr_destroy_cache(sbi->s_mb_cache); 851 sbi->s_mb_cache = NULL; 852 } 853 if (sbi->s_mmp_tsk) 854 kthread_stop(sbi->s_mmp_tsk); 855 sb->s_fs_info = NULL; 856 /* 857 * Now that we are completely done shutting down the 858 * superblock, we need to actually destroy the kobject. 859 */ 860 kobject_put(&sbi->s_kobj); 861 wait_for_completion(&sbi->s_kobj_unregister); 862 if (sbi->s_chksum_driver) 863 crypto_free_shash(sbi->s_chksum_driver); 864 kfree(sbi->s_blockgroup_lock); 865 kfree(sbi); 866 } 867 868 static struct kmem_cache *ext4_inode_cachep; 869 870 /* 871 * Called inside transaction, so use GFP_NOFS 872 */ 873 static struct inode *ext4_alloc_inode(struct super_block *sb) 874 { 875 struct ext4_inode_info *ei; 876 877 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 878 if (!ei) 879 return NULL; 880 881 ei->vfs_inode.i_version = 1; 882 INIT_LIST_HEAD(&ei->i_prealloc_list); 883 spin_lock_init(&ei->i_prealloc_lock); 884 ext4_es_init_tree(&ei->i_es_tree); 885 rwlock_init(&ei->i_es_lock); 886 INIT_LIST_HEAD(&ei->i_es_lru); 887 ei->i_es_lru_nr = 0; 888 ei->i_touch_when = 0; 889 ei->i_reserved_data_blocks = 0; 890 ei->i_reserved_meta_blocks = 0; 891 ei->i_allocated_meta_blocks = 0; 892 ei->i_da_metadata_calc_len = 0; 893 ei->i_da_metadata_calc_last_lblock = 0; 894 spin_lock_init(&(ei->i_block_reservation_lock)); 895 #ifdef CONFIG_QUOTA 896 ei->i_reserved_quota = 0; 897 #endif 898 ei->jinode = NULL; 899 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 900 spin_lock_init(&ei->i_completed_io_lock); 901 ei->i_sync_tid = 0; 902 ei->i_datasync_tid = 0; 903 atomic_set(&ei->i_ioend_count, 0); 904 atomic_set(&ei->i_unwritten, 0); 905 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 906 907 return &ei->vfs_inode; 908 } 909 910 static int ext4_drop_inode(struct inode *inode) 911 { 912 int drop = generic_drop_inode(inode); 913 914 trace_ext4_drop_inode(inode, drop); 915 return drop; 916 } 917 918 static void ext4_i_callback(struct rcu_head *head) 919 { 920 struct inode *inode = container_of(head, struct inode, i_rcu); 921 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 922 } 923 924 static void ext4_destroy_inode(struct inode *inode) 925 { 926 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 927 ext4_msg(inode->i_sb, KERN_ERR, 928 "Inode %lu (%p): orphan list check failed!", 929 inode->i_ino, EXT4_I(inode)); 930 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 931 EXT4_I(inode), sizeof(struct ext4_inode_info), 932 true); 933 dump_stack(); 934 } 935 call_rcu(&inode->i_rcu, ext4_i_callback); 936 } 937 938 static void init_once(void *foo) 939 { 940 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 941 942 INIT_LIST_HEAD(&ei->i_orphan); 943 init_rwsem(&ei->xattr_sem); 944 init_rwsem(&ei->i_data_sem); 945 inode_init_once(&ei->vfs_inode); 946 } 947 948 static int __init init_inodecache(void) 949 { 950 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 951 sizeof(struct ext4_inode_info), 952 0, (SLAB_RECLAIM_ACCOUNT| 953 SLAB_MEM_SPREAD), 954 init_once); 955 if (ext4_inode_cachep == NULL) 956 return -ENOMEM; 957 return 0; 958 } 959 960 static void destroy_inodecache(void) 961 { 962 /* 963 * Make sure all delayed rcu free inodes are flushed before we 964 * destroy cache. 965 */ 966 rcu_barrier(); 967 kmem_cache_destroy(ext4_inode_cachep); 968 } 969 970 void ext4_clear_inode(struct inode *inode) 971 { 972 invalidate_inode_buffers(inode); 973 clear_inode(inode); 974 dquot_drop(inode); 975 ext4_discard_preallocations(inode); 976 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 977 ext4_es_lru_del(inode); 978 if (EXT4_I(inode)->jinode) { 979 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 980 EXT4_I(inode)->jinode); 981 jbd2_free_inode(EXT4_I(inode)->jinode); 982 EXT4_I(inode)->jinode = NULL; 983 } 984 } 985 986 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 987 u64 ino, u32 generation) 988 { 989 struct inode *inode; 990 991 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 992 return ERR_PTR(-ESTALE); 993 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 994 return ERR_PTR(-ESTALE); 995 996 /* iget isn't really right if the inode is currently unallocated!! 997 * 998 * ext4_read_inode will return a bad_inode if the inode had been 999 * deleted, so we should be safe. 1000 * 1001 * Currently we don't know the generation for parent directory, so 1002 * a generation of 0 means "accept any" 1003 */ 1004 inode = ext4_iget(sb, ino); 1005 if (IS_ERR(inode)) 1006 return ERR_CAST(inode); 1007 if (generation && inode->i_generation != generation) { 1008 iput(inode); 1009 return ERR_PTR(-ESTALE); 1010 } 1011 1012 return inode; 1013 } 1014 1015 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1016 int fh_len, int fh_type) 1017 { 1018 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1019 ext4_nfs_get_inode); 1020 } 1021 1022 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1023 int fh_len, int fh_type) 1024 { 1025 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1026 ext4_nfs_get_inode); 1027 } 1028 1029 /* 1030 * Try to release metadata pages (indirect blocks, directories) which are 1031 * mapped via the block device. Since these pages could have journal heads 1032 * which would prevent try_to_free_buffers() from freeing them, we must use 1033 * jbd2 layer's try_to_free_buffers() function to release them. 1034 */ 1035 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1036 gfp_t wait) 1037 { 1038 journal_t *journal = EXT4_SB(sb)->s_journal; 1039 1040 WARN_ON(PageChecked(page)); 1041 if (!page_has_buffers(page)) 1042 return 0; 1043 if (journal) 1044 return jbd2_journal_try_to_free_buffers(journal, page, 1045 wait & ~__GFP_WAIT); 1046 return try_to_free_buffers(page); 1047 } 1048 1049 #ifdef CONFIG_QUOTA 1050 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") 1051 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) 1052 1053 static int ext4_write_dquot(struct dquot *dquot); 1054 static int ext4_acquire_dquot(struct dquot *dquot); 1055 static int ext4_release_dquot(struct dquot *dquot); 1056 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1057 static int ext4_write_info(struct super_block *sb, int type); 1058 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1059 struct path *path); 1060 static int ext4_quota_on_sysfile(struct super_block *sb, int type, 1061 int format_id); 1062 static int ext4_quota_off(struct super_block *sb, int type); 1063 static int ext4_quota_off_sysfile(struct super_block *sb, int type); 1064 static int ext4_quota_on_mount(struct super_block *sb, int type); 1065 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1066 size_t len, loff_t off); 1067 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1068 const char *data, size_t len, loff_t off); 1069 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1070 unsigned int flags); 1071 static int ext4_enable_quotas(struct super_block *sb); 1072 1073 static const struct dquot_operations ext4_quota_operations = { 1074 .get_reserved_space = ext4_get_reserved_space, 1075 .write_dquot = ext4_write_dquot, 1076 .acquire_dquot = ext4_acquire_dquot, 1077 .release_dquot = ext4_release_dquot, 1078 .mark_dirty = ext4_mark_dquot_dirty, 1079 .write_info = ext4_write_info, 1080 .alloc_dquot = dquot_alloc, 1081 .destroy_dquot = dquot_destroy, 1082 }; 1083 1084 static const struct quotactl_ops ext4_qctl_operations = { 1085 .quota_on = ext4_quota_on, 1086 .quota_off = ext4_quota_off, 1087 .quota_sync = dquot_quota_sync, 1088 .get_info = dquot_get_dqinfo, 1089 .set_info = dquot_set_dqinfo, 1090 .get_dqblk = dquot_get_dqblk, 1091 .set_dqblk = dquot_set_dqblk 1092 }; 1093 1094 static const struct quotactl_ops ext4_qctl_sysfile_operations = { 1095 .quota_on_meta = ext4_quota_on_sysfile, 1096 .quota_off = ext4_quota_off_sysfile, 1097 .quota_sync = dquot_quota_sync, 1098 .get_info = dquot_get_dqinfo, 1099 .set_info = dquot_set_dqinfo, 1100 .get_dqblk = dquot_get_dqblk, 1101 .set_dqblk = dquot_set_dqblk 1102 }; 1103 #endif 1104 1105 static const struct super_operations ext4_sops = { 1106 .alloc_inode = ext4_alloc_inode, 1107 .destroy_inode = ext4_destroy_inode, 1108 .write_inode = ext4_write_inode, 1109 .dirty_inode = ext4_dirty_inode, 1110 .drop_inode = ext4_drop_inode, 1111 .evict_inode = ext4_evict_inode, 1112 .put_super = ext4_put_super, 1113 .sync_fs = ext4_sync_fs, 1114 .freeze_fs = ext4_freeze, 1115 .unfreeze_fs = ext4_unfreeze, 1116 .statfs = ext4_statfs, 1117 .remount_fs = ext4_remount, 1118 .show_options = ext4_show_options, 1119 #ifdef CONFIG_QUOTA 1120 .quota_read = ext4_quota_read, 1121 .quota_write = ext4_quota_write, 1122 #endif 1123 .bdev_try_to_free_page = bdev_try_to_free_page, 1124 }; 1125 1126 static const struct super_operations ext4_nojournal_sops = { 1127 .alloc_inode = ext4_alloc_inode, 1128 .destroy_inode = ext4_destroy_inode, 1129 .write_inode = ext4_write_inode, 1130 .dirty_inode = ext4_dirty_inode, 1131 .drop_inode = ext4_drop_inode, 1132 .evict_inode = ext4_evict_inode, 1133 .sync_fs = ext4_sync_fs_nojournal, 1134 .put_super = ext4_put_super, 1135 .statfs = ext4_statfs, 1136 .remount_fs = ext4_remount, 1137 .show_options = ext4_show_options, 1138 #ifdef CONFIG_QUOTA 1139 .quota_read = ext4_quota_read, 1140 .quota_write = ext4_quota_write, 1141 #endif 1142 .bdev_try_to_free_page = bdev_try_to_free_page, 1143 }; 1144 1145 static const struct export_operations ext4_export_ops = { 1146 .fh_to_dentry = ext4_fh_to_dentry, 1147 .fh_to_parent = ext4_fh_to_parent, 1148 .get_parent = ext4_get_parent, 1149 }; 1150 1151 enum { 1152 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1153 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1154 Opt_nouid32, Opt_debug, Opt_removed, 1155 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1156 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1157 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1158 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1159 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1160 Opt_data_err_abort, Opt_data_err_ignore, 1161 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1162 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1163 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1164 Opt_usrquota, Opt_grpquota, Opt_i_version, 1165 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 1166 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1167 Opt_inode_readahead_blks, Opt_journal_ioprio, 1168 Opt_dioread_nolock, Opt_dioread_lock, 1169 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1170 Opt_max_dir_size_kb, 1171 }; 1172 1173 static const match_table_t tokens = { 1174 {Opt_bsd_df, "bsddf"}, 1175 {Opt_minix_df, "minixdf"}, 1176 {Opt_grpid, "grpid"}, 1177 {Opt_grpid, "bsdgroups"}, 1178 {Opt_nogrpid, "nogrpid"}, 1179 {Opt_nogrpid, "sysvgroups"}, 1180 {Opt_resgid, "resgid=%u"}, 1181 {Opt_resuid, "resuid=%u"}, 1182 {Opt_sb, "sb=%u"}, 1183 {Opt_err_cont, "errors=continue"}, 1184 {Opt_err_panic, "errors=panic"}, 1185 {Opt_err_ro, "errors=remount-ro"}, 1186 {Opt_nouid32, "nouid32"}, 1187 {Opt_debug, "debug"}, 1188 {Opt_removed, "oldalloc"}, 1189 {Opt_removed, "orlov"}, 1190 {Opt_user_xattr, "user_xattr"}, 1191 {Opt_nouser_xattr, "nouser_xattr"}, 1192 {Opt_acl, "acl"}, 1193 {Opt_noacl, "noacl"}, 1194 {Opt_noload, "norecovery"}, 1195 {Opt_noload, "noload"}, 1196 {Opt_removed, "nobh"}, 1197 {Opt_removed, "bh"}, 1198 {Opt_commit, "commit=%u"}, 1199 {Opt_min_batch_time, "min_batch_time=%u"}, 1200 {Opt_max_batch_time, "max_batch_time=%u"}, 1201 {Opt_journal_dev, "journal_dev=%u"}, 1202 {Opt_journal_path, "journal_path=%s"}, 1203 {Opt_journal_checksum, "journal_checksum"}, 1204 {Opt_journal_async_commit, "journal_async_commit"}, 1205 {Opt_abort, "abort"}, 1206 {Opt_data_journal, "data=journal"}, 1207 {Opt_data_ordered, "data=ordered"}, 1208 {Opt_data_writeback, "data=writeback"}, 1209 {Opt_data_err_abort, "data_err=abort"}, 1210 {Opt_data_err_ignore, "data_err=ignore"}, 1211 {Opt_offusrjquota, "usrjquota="}, 1212 {Opt_usrjquota, "usrjquota=%s"}, 1213 {Opt_offgrpjquota, "grpjquota="}, 1214 {Opt_grpjquota, "grpjquota=%s"}, 1215 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1216 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1217 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1218 {Opt_grpquota, "grpquota"}, 1219 {Opt_noquota, "noquota"}, 1220 {Opt_quota, "quota"}, 1221 {Opt_usrquota, "usrquota"}, 1222 {Opt_barrier, "barrier=%u"}, 1223 {Opt_barrier, "barrier"}, 1224 {Opt_nobarrier, "nobarrier"}, 1225 {Opt_i_version, "i_version"}, 1226 {Opt_stripe, "stripe=%u"}, 1227 {Opt_delalloc, "delalloc"}, 1228 {Opt_nodelalloc, "nodelalloc"}, 1229 {Opt_removed, "mblk_io_submit"}, 1230 {Opt_removed, "nomblk_io_submit"}, 1231 {Opt_block_validity, "block_validity"}, 1232 {Opt_noblock_validity, "noblock_validity"}, 1233 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1234 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1235 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1236 {Opt_auto_da_alloc, "auto_da_alloc"}, 1237 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1238 {Opt_dioread_nolock, "dioread_nolock"}, 1239 {Opt_dioread_lock, "dioread_lock"}, 1240 {Opt_discard, "discard"}, 1241 {Opt_nodiscard, "nodiscard"}, 1242 {Opt_init_itable, "init_itable=%u"}, 1243 {Opt_init_itable, "init_itable"}, 1244 {Opt_noinit_itable, "noinit_itable"}, 1245 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1246 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1247 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1248 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1249 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1250 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1251 {Opt_err, NULL}, 1252 }; 1253 1254 static ext4_fsblk_t get_sb_block(void **data) 1255 { 1256 ext4_fsblk_t sb_block; 1257 char *options = (char *) *data; 1258 1259 if (!options || strncmp(options, "sb=", 3) != 0) 1260 return 1; /* Default location */ 1261 1262 options += 3; 1263 /* TODO: use simple_strtoll with >32bit ext4 */ 1264 sb_block = simple_strtoul(options, &options, 0); 1265 if (*options && *options != ',') { 1266 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1267 (char *) *data); 1268 return 1; 1269 } 1270 if (*options == ',') 1271 options++; 1272 *data = (void *) options; 1273 1274 return sb_block; 1275 } 1276 1277 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1278 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1279 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1280 1281 #ifdef CONFIG_QUOTA 1282 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1283 { 1284 struct ext4_sb_info *sbi = EXT4_SB(sb); 1285 char *qname; 1286 int ret = -1; 1287 1288 if (sb_any_quota_loaded(sb) && 1289 !sbi->s_qf_names[qtype]) { 1290 ext4_msg(sb, KERN_ERR, 1291 "Cannot change journaled " 1292 "quota options when quota turned on"); 1293 return -1; 1294 } 1295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) { 1296 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options " 1297 "when QUOTA feature is enabled"); 1298 return -1; 1299 } 1300 qname = match_strdup(args); 1301 if (!qname) { 1302 ext4_msg(sb, KERN_ERR, 1303 "Not enough memory for storing quotafile name"); 1304 return -1; 1305 } 1306 if (sbi->s_qf_names[qtype]) { 1307 if (strcmp(sbi->s_qf_names[qtype], qname) == 0) 1308 ret = 1; 1309 else 1310 ext4_msg(sb, KERN_ERR, 1311 "%s quota file already specified", 1312 QTYPE2NAME(qtype)); 1313 goto errout; 1314 } 1315 if (strchr(qname, '/')) { 1316 ext4_msg(sb, KERN_ERR, 1317 "quotafile must be on filesystem root"); 1318 goto errout; 1319 } 1320 sbi->s_qf_names[qtype] = qname; 1321 set_opt(sb, QUOTA); 1322 return 1; 1323 errout: 1324 kfree(qname); 1325 return ret; 1326 } 1327 1328 static int clear_qf_name(struct super_block *sb, int qtype) 1329 { 1330 1331 struct ext4_sb_info *sbi = EXT4_SB(sb); 1332 1333 if (sb_any_quota_loaded(sb) && 1334 sbi->s_qf_names[qtype]) { 1335 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1336 " when quota turned on"); 1337 return -1; 1338 } 1339 kfree(sbi->s_qf_names[qtype]); 1340 sbi->s_qf_names[qtype] = NULL; 1341 return 1; 1342 } 1343 #endif 1344 1345 #define MOPT_SET 0x0001 1346 #define MOPT_CLEAR 0x0002 1347 #define MOPT_NOSUPPORT 0x0004 1348 #define MOPT_EXPLICIT 0x0008 1349 #define MOPT_CLEAR_ERR 0x0010 1350 #define MOPT_GTE0 0x0020 1351 #ifdef CONFIG_QUOTA 1352 #define MOPT_Q 0 1353 #define MOPT_QFMT 0x0040 1354 #else 1355 #define MOPT_Q MOPT_NOSUPPORT 1356 #define MOPT_QFMT MOPT_NOSUPPORT 1357 #endif 1358 #define MOPT_DATAJ 0x0080 1359 #define MOPT_NO_EXT2 0x0100 1360 #define MOPT_NO_EXT3 0x0200 1361 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1362 #define MOPT_STRING 0x0400 1363 1364 static const struct mount_opts { 1365 int token; 1366 int mount_opt; 1367 int flags; 1368 } ext4_mount_opts[] = { 1369 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1370 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1371 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1372 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1373 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1374 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1375 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1376 MOPT_EXT4_ONLY | MOPT_SET}, 1377 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1378 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1379 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1380 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1381 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1382 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1383 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1384 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1385 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1386 MOPT_EXT4_ONLY | MOPT_SET}, 1387 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1388 EXT4_MOUNT_JOURNAL_CHECKSUM), 1389 MOPT_EXT4_ONLY | MOPT_SET}, 1390 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1391 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1392 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1393 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1394 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1395 MOPT_NO_EXT2 | MOPT_SET}, 1396 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1397 MOPT_NO_EXT2 | MOPT_CLEAR}, 1398 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1399 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1400 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1401 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1402 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1403 {Opt_commit, 0, MOPT_GTE0}, 1404 {Opt_max_batch_time, 0, MOPT_GTE0}, 1405 {Opt_min_batch_time, 0, MOPT_GTE0}, 1406 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1407 {Opt_init_itable, 0, MOPT_GTE0}, 1408 {Opt_stripe, 0, MOPT_GTE0}, 1409 {Opt_resuid, 0, MOPT_GTE0}, 1410 {Opt_resgid, 0, MOPT_GTE0}, 1411 {Opt_journal_dev, 0, MOPT_GTE0}, 1412 {Opt_journal_path, 0, MOPT_STRING}, 1413 {Opt_journal_ioprio, 0, MOPT_GTE0}, 1414 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1415 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1416 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1417 MOPT_NO_EXT2 | MOPT_DATAJ}, 1418 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1419 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1420 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1421 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1422 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1423 #else 1424 {Opt_acl, 0, MOPT_NOSUPPORT}, 1425 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1426 #endif 1427 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1428 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1429 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1430 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1431 MOPT_SET | MOPT_Q}, 1432 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1433 MOPT_SET | MOPT_Q}, 1434 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1435 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q}, 1436 {Opt_usrjquota, 0, MOPT_Q}, 1437 {Opt_grpjquota, 0, MOPT_Q}, 1438 {Opt_offusrjquota, 0, MOPT_Q}, 1439 {Opt_offgrpjquota, 0, MOPT_Q}, 1440 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1441 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1442 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1443 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1444 {Opt_err, 0, 0} 1445 }; 1446 1447 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1448 substring_t *args, unsigned long *journal_devnum, 1449 unsigned int *journal_ioprio, int is_remount) 1450 { 1451 struct ext4_sb_info *sbi = EXT4_SB(sb); 1452 const struct mount_opts *m; 1453 kuid_t uid; 1454 kgid_t gid; 1455 int arg = 0; 1456 1457 #ifdef CONFIG_QUOTA 1458 if (token == Opt_usrjquota) 1459 return set_qf_name(sb, USRQUOTA, &args[0]); 1460 else if (token == Opt_grpjquota) 1461 return set_qf_name(sb, GRPQUOTA, &args[0]); 1462 else if (token == Opt_offusrjquota) 1463 return clear_qf_name(sb, USRQUOTA); 1464 else if (token == Opt_offgrpjquota) 1465 return clear_qf_name(sb, GRPQUOTA); 1466 #endif 1467 switch (token) { 1468 case Opt_noacl: 1469 case Opt_nouser_xattr: 1470 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1471 break; 1472 case Opt_sb: 1473 return 1; /* handled by get_sb_block() */ 1474 case Opt_removed: 1475 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1476 return 1; 1477 case Opt_abort: 1478 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1479 return 1; 1480 case Opt_i_version: 1481 sb->s_flags |= MS_I_VERSION; 1482 return 1; 1483 } 1484 1485 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1486 if (token == m->token) 1487 break; 1488 1489 if (m->token == Opt_err) { 1490 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1491 "or missing value", opt); 1492 return -1; 1493 } 1494 1495 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1496 ext4_msg(sb, KERN_ERR, 1497 "Mount option \"%s\" incompatible with ext2", opt); 1498 return -1; 1499 } 1500 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1501 ext4_msg(sb, KERN_ERR, 1502 "Mount option \"%s\" incompatible with ext3", opt); 1503 return -1; 1504 } 1505 1506 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1507 return -1; 1508 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1509 return -1; 1510 if (m->flags & MOPT_EXPLICIT) 1511 set_opt2(sb, EXPLICIT_DELALLOC); 1512 if (m->flags & MOPT_CLEAR_ERR) 1513 clear_opt(sb, ERRORS_MASK); 1514 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1515 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1516 "options when quota turned on"); 1517 return -1; 1518 } 1519 1520 if (m->flags & MOPT_NOSUPPORT) { 1521 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1522 } else if (token == Opt_commit) { 1523 if (arg == 0) 1524 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1525 sbi->s_commit_interval = HZ * arg; 1526 } else if (token == Opt_max_batch_time) { 1527 if (arg == 0) 1528 arg = EXT4_DEF_MAX_BATCH_TIME; 1529 sbi->s_max_batch_time = arg; 1530 } else if (token == Opt_min_batch_time) { 1531 sbi->s_min_batch_time = arg; 1532 } else if (token == Opt_inode_readahead_blks) { 1533 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1534 ext4_msg(sb, KERN_ERR, 1535 "EXT4-fs: inode_readahead_blks must be " 1536 "0 or a power of 2 smaller than 2^31"); 1537 return -1; 1538 } 1539 sbi->s_inode_readahead_blks = arg; 1540 } else if (token == Opt_init_itable) { 1541 set_opt(sb, INIT_INODE_TABLE); 1542 if (!args->from) 1543 arg = EXT4_DEF_LI_WAIT_MULT; 1544 sbi->s_li_wait_mult = arg; 1545 } else if (token == Opt_max_dir_size_kb) { 1546 sbi->s_max_dir_size_kb = arg; 1547 } else if (token == Opt_stripe) { 1548 sbi->s_stripe = arg; 1549 } else if (token == Opt_resuid) { 1550 uid = make_kuid(current_user_ns(), arg); 1551 if (!uid_valid(uid)) { 1552 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1553 return -1; 1554 } 1555 sbi->s_resuid = uid; 1556 } else if (token == Opt_resgid) { 1557 gid = make_kgid(current_user_ns(), arg); 1558 if (!gid_valid(gid)) { 1559 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1560 return -1; 1561 } 1562 sbi->s_resgid = gid; 1563 } else if (token == Opt_journal_dev) { 1564 if (is_remount) { 1565 ext4_msg(sb, KERN_ERR, 1566 "Cannot specify journal on remount"); 1567 return -1; 1568 } 1569 *journal_devnum = arg; 1570 } else if (token == Opt_journal_path) { 1571 char *journal_path; 1572 struct inode *journal_inode; 1573 struct path path; 1574 int error; 1575 1576 if (is_remount) { 1577 ext4_msg(sb, KERN_ERR, 1578 "Cannot specify journal on remount"); 1579 return -1; 1580 } 1581 journal_path = match_strdup(&args[0]); 1582 if (!journal_path) { 1583 ext4_msg(sb, KERN_ERR, "error: could not dup " 1584 "journal device string"); 1585 return -1; 1586 } 1587 1588 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1589 if (error) { 1590 ext4_msg(sb, KERN_ERR, "error: could not find " 1591 "journal device path: error %d", error); 1592 kfree(journal_path); 1593 return -1; 1594 } 1595 1596 journal_inode = path.dentry->d_inode; 1597 if (!S_ISBLK(journal_inode->i_mode)) { 1598 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1599 "is not a block device", journal_path); 1600 path_put(&path); 1601 kfree(journal_path); 1602 return -1; 1603 } 1604 1605 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1606 path_put(&path); 1607 kfree(journal_path); 1608 } else if (token == Opt_journal_ioprio) { 1609 if (arg > 7) { 1610 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1611 " (must be 0-7)"); 1612 return -1; 1613 } 1614 *journal_ioprio = 1615 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1616 } else if (m->flags & MOPT_DATAJ) { 1617 if (is_remount) { 1618 if (!sbi->s_journal) 1619 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1620 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 1621 ext4_msg(sb, KERN_ERR, 1622 "Cannot change data mode on remount"); 1623 return -1; 1624 } 1625 } else { 1626 clear_opt(sb, DATA_FLAGS); 1627 sbi->s_mount_opt |= m->mount_opt; 1628 } 1629 #ifdef CONFIG_QUOTA 1630 } else if (m->flags & MOPT_QFMT) { 1631 if (sb_any_quota_loaded(sb) && 1632 sbi->s_jquota_fmt != m->mount_opt) { 1633 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 1634 "quota options when quota turned on"); 1635 return -1; 1636 } 1637 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 1638 EXT4_FEATURE_RO_COMPAT_QUOTA)) { 1639 ext4_msg(sb, KERN_ERR, 1640 "Cannot set journaled quota options " 1641 "when QUOTA feature is enabled"); 1642 return -1; 1643 } 1644 sbi->s_jquota_fmt = m->mount_opt; 1645 #endif 1646 } else { 1647 if (!args->from) 1648 arg = 1; 1649 if (m->flags & MOPT_CLEAR) 1650 arg = !arg; 1651 else if (unlikely(!(m->flags & MOPT_SET))) { 1652 ext4_msg(sb, KERN_WARNING, 1653 "buggy handling of option %s", opt); 1654 WARN_ON(1); 1655 return -1; 1656 } 1657 if (arg != 0) 1658 sbi->s_mount_opt |= m->mount_opt; 1659 else 1660 sbi->s_mount_opt &= ~m->mount_opt; 1661 } 1662 return 1; 1663 } 1664 1665 static int parse_options(char *options, struct super_block *sb, 1666 unsigned long *journal_devnum, 1667 unsigned int *journal_ioprio, 1668 int is_remount) 1669 { 1670 struct ext4_sb_info *sbi = EXT4_SB(sb); 1671 char *p; 1672 substring_t args[MAX_OPT_ARGS]; 1673 int token; 1674 1675 if (!options) 1676 return 1; 1677 1678 while ((p = strsep(&options, ",")) != NULL) { 1679 if (!*p) 1680 continue; 1681 /* 1682 * Initialize args struct so we know whether arg was 1683 * found; some options take optional arguments. 1684 */ 1685 args[0].to = args[0].from = NULL; 1686 token = match_token(p, tokens, args); 1687 if (handle_mount_opt(sb, p, token, args, journal_devnum, 1688 journal_ioprio, is_remount) < 0) 1689 return 0; 1690 } 1691 #ifdef CONFIG_QUOTA 1692 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 1693 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) { 1694 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA " 1695 "feature is enabled"); 1696 return 0; 1697 } 1698 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1699 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1700 clear_opt(sb, USRQUOTA); 1701 1702 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1703 clear_opt(sb, GRPQUOTA); 1704 1705 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1706 ext4_msg(sb, KERN_ERR, "old and new quota " 1707 "format mixing"); 1708 return 0; 1709 } 1710 1711 if (!sbi->s_jquota_fmt) { 1712 ext4_msg(sb, KERN_ERR, "journaled quota format " 1713 "not specified"); 1714 return 0; 1715 } 1716 } else { 1717 if (sbi->s_jquota_fmt) { 1718 ext4_msg(sb, KERN_ERR, "journaled quota format " 1719 "specified with no journaling " 1720 "enabled"); 1721 return 0; 1722 } 1723 } 1724 #endif 1725 if (test_opt(sb, DIOREAD_NOLOCK)) { 1726 int blocksize = 1727 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 1728 1729 if (blocksize < PAGE_CACHE_SIZE) { 1730 ext4_msg(sb, KERN_ERR, "can't mount with " 1731 "dioread_nolock if block size != PAGE_SIZE"); 1732 return 0; 1733 } 1734 } 1735 return 1; 1736 } 1737 1738 static inline void ext4_show_quota_options(struct seq_file *seq, 1739 struct super_block *sb) 1740 { 1741 #if defined(CONFIG_QUOTA) 1742 struct ext4_sb_info *sbi = EXT4_SB(sb); 1743 1744 if (sbi->s_jquota_fmt) { 1745 char *fmtname = ""; 1746 1747 switch (sbi->s_jquota_fmt) { 1748 case QFMT_VFS_OLD: 1749 fmtname = "vfsold"; 1750 break; 1751 case QFMT_VFS_V0: 1752 fmtname = "vfsv0"; 1753 break; 1754 case QFMT_VFS_V1: 1755 fmtname = "vfsv1"; 1756 break; 1757 } 1758 seq_printf(seq, ",jqfmt=%s", fmtname); 1759 } 1760 1761 if (sbi->s_qf_names[USRQUOTA]) 1762 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); 1763 1764 if (sbi->s_qf_names[GRPQUOTA]) 1765 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); 1766 #endif 1767 } 1768 1769 static const char *token2str(int token) 1770 { 1771 const struct match_token *t; 1772 1773 for (t = tokens; t->token != Opt_err; t++) 1774 if (t->token == token && !strchr(t->pattern, '=')) 1775 break; 1776 return t->pattern; 1777 } 1778 1779 /* 1780 * Show an option if 1781 * - it's set to a non-default value OR 1782 * - if the per-sb default is different from the global default 1783 */ 1784 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 1785 int nodefs) 1786 { 1787 struct ext4_sb_info *sbi = EXT4_SB(sb); 1788 struct ext4_super_block *es = sbi->s_es; 1789 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; 1790 const struct mount_opts *m; 1791 char sep = nodefs ? '\n' : ','; 1792 1793 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 1794 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 1795 1796 if (sbi->s_sb_block != 1) 1797 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 1798 1799 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 1800 int want_set = m->flags & MOPT_SET; 1801 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 1802 (m->flags & MOPT_CLEAR_ERR)) 1803 continue; 1804 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 1805 continue; /* skip if same as the default */ 1806 if ((want_set && 1807 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 1808 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 1809 continue; /* select Opt_noFoo vs Opt_Foo */ 1810 SEQ_OPTS_PRINT("%s", token2str(m->token)); 1811 } 1812 1813 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 1814 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 1815 SEQ_OPTS_PRINT("resuid=%u", 1816 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 1817 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 1818 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 1819 SEQ_OPTS_PRINT("resgid=%u", 1820 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 1821 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 1822 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 1823 SEQ_OPTS_PUTS("errors=remount-ro"); 1824 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 1825 SEQ_OPTS_PUTS("errors=continue"); 1826 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 1827 SEQ_OPTS_PUTS("errors=panic"); 1828 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 1829 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 1830 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 1831 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 1832 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 1833 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 1834 if (sb->s_flags & MS_I_VERSION) 1835 SEQ_OPTS_PUTS("i_version"); 1836 if (nodefs || sbi->s_stripe) 1837 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 1838 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { 1839 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 1840 SEQ_OPTS_PUTS("data=journal"); 1841 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 1842 SEQ_OPTS_PUTS("data=ordered"); 1843 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 1844 SEQ_OPTS_PUTS("data=writeback"); 1845 } 1846 if (nodefs || 1847 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 1848 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 1849 sbi->s_inode_readahead_blks); 1850 1851 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && 1852 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 1853 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 1854 if (nodefs || sbi->s_max_dir_size_kb) 1855 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 1856 1857 ext4_show_quota_options(seq, sb); 1858 return 0; 1859 } 1860 1861 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 1862 { 1863 return _ext4_show_options(seq, root->d_sb, 0); 1864 } 1865 1866 static int options_seq_show(struct seq_file *seq, void *offset) 1867 { 1868 struct super_block *sb = seq->private; 1869 int rc; 1870 1871 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw"); 1872 rc = _ext4_show_options(seq, sb, 1); 1873 seq_puts(seq, "\n"); 1874 return rc; 1875 } 1876 1877 static int options_open_fs(struct inode *inode, struct file *file) 1878 { 1879 return single_open(file, options_seq_show, PDE_DATA(inode)); 1880 } 1881 1882 static const struct file_operations ext4_seq_options_fops = { 1883 .owner = THIS_MODULE, 1884 .open = options_open_fs, 1885 .read = seq_read, 1886 .llseek = seq_lseek, 1887 .release = single_release, 1888 }; 1889 1890 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 1891 int read_only) 1892 { 1893 struct ext4_sb_info *sbi = EXT4_SB(sb); 1894 int res = 0; 1895 1896 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 1897 ext4_msg(sb, KERN_ERR, "revision level too high, " 1898 "forcing read-only mode"); 1899 res = MS_RDONLY; 1900 } 1901 if (read_only) 1902 goto done; 1903 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1904 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1905 "running e2fsck is recommended"); 1906 else if ((sbi->s_mount_state & EXT4_ERROR_FS)) 1907 ext4_msg(sb, KERN_WARNING, 1908 "warning: mounting fs with errors, " 1909 "running e2fsck is recommended"); 1910 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 1911 le16_to_cpu(es->s_mnt_count) >= 1912 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 1913 ext4_msg(sb, KERN_WARNING, 1914 "warning: maximal mount count reached, " 1915 "running e2fsck is recommended"); 1916 else if (le32_to_cpu(es->s_checkinterval) && 1917 (le32_to_cpu(es->s_lastcheck) + 1918 le32_to_cpu(es->s_checkinterval) <= get_seconds())) 1919 ext4_msg(sb, KERN_WARNING, 1920 "warning: checktime reached, " 1921 "running e2fsck is recommended"); 1922 if (!sbi->s_journal) 1923 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 1924 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1925 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1926 le16_add_cpu(&es->s_mnt_count, 1); 1927 es->s_mtime = cpu_to_le32(get_seconds()); 1928 ext4_update_dynamic_rev(sb); 1929 if (sbi->s_journal) 1930 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1931 1932 ext4_commit_super(sb, 1); 1933 done: 1934 if (test_opt(sb, DEBUG)) 1935 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1936 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1937 sb->s_blocksize, 1938 sbi->s_groups_count, 1939 EXT4_BLOCKS_PER_GROUP(sb), 1940 EXT4_INODES_PER_GROUP(sb), 1941 sbi->s_mount_opt, sbi->s_mount_opt2); 1942 1943 cleancache_init_fs(sb); 1944 return res; 1945 } 1946 1947 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 1948 { 1949 struct ext4_sb_info *sbi = EXT4_SB(sb); 1950 struct flex_groups *new_groups; 1951 int size; 1952 1953 if (!sbi->s_log_groups_per_flex) 1954 return 0; 1955 1956 size = ext4_flex_group(sbi, ngroup - 1) + 1; 1957 if (size <= sbi->s_flex_groups_allocated) 1958 return 0; 1959 1960 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 1961 new_groups = ext4_kvzalloc(size, GFP_KERNEL); 1962 if (!new_groups) { 1963 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 1964 size / (int) sizeof(struct flex_groups)); 1965 return -ENOMEM; 1966 } 1967 1968 if (sbi->s_flex_groups) { 1969 memcpy(new_groups, sbi->s_flex_groups, 1970 (sbi->s_flex_groups_allocated * 1971 sizeof(struct flex_groups))); 1972 ext4_kvfree(sbi->s_flex_groups); 1973 } 1974 sbi->s_flex_groups = new_groups; 1975 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 1976 return 0; 1977 } 1978 1979 static int ext4_fill_flex_info(struct super_block *sb) 1980 { 1981 struct ext4_sb_info *sbi = EXT4_SB(sb); 1982 struct ext4_group_desc *gdp = NULL; 1983 ext4_group_t flex_group; 1984 int i, err; 1985 1986 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 1987 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 1988 sbi->s_log_groups_per_flex = 0; 1989 return 1; 1990 } 1991 1992 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 1993 if (err) 1994 goto failed; 1995 1996 for (i = 0; i < sbi->s_groups_count; i++) { 1997 gdp = ext4_get_group_desc(sb, i, NULL); 1998 1999 flex_group = ext4_flex_group(sbi, i); 2000 atomic_add(ext4_free_inodes_count(sb, gdp), 2001 &sbi->s_flex_groups[flex_group].free_inodes); 2002 atomic64_add(ext4_free_group_clusters(sb, gdp), 2003 &sbi->s_flex_groups[flex_group].free_clusters); 2004 atomic_add(ext4_used_dirs_count(sb, gdp), 2005 &sbi->s_flex_groups[flex_group].used_dirs); 2006 } 2007 2008 return 1; 2009 failed: 2010 return 0; 2011 } 2012 2013 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, 2014 struct ext4_group_desc *gdp) 2015 { 2016 int offset; 2017 __u16 crc = 0; 2018 __le32 le_group = cpu_to_le32(block_group); 2019 2020 if ((sbi->s_es->s_feature_ro_compat & 2021 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) { 2022 /* Use new metadata_csum algorithm */ 2023 __le16 save_csum; 2024 __u32 csum32; 2025 2026 save_csum = gdp->bg_checksum; 2027 gdp->bg_checksum = 0; 2028 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2029 sizeof(le_group)); 2030 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, 2031 sbi->s_desc_size); 2032 gdp->bg_checksum = save_csum; 2033 2034 crc = csum32 & 0xFFFF; 2035 goto out; 2036 } 2037 2038 /* old crc16 code */ 2039 offset = offsetof(struct ext4_group_desc, bg_checksum); 2040 2041 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2042 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2043 crc = crc16(crc, (__u8 *)gdp, offset); 2044 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2045 /* for checksum of struct ext4_group_desc do the rest...*/ 2046 if ((sbi->s_es->s_feature_incompat & 2047 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && 2048 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2049 crc = crc16(crc, (__u8 *)gdp + offset, 2050 le16_to_cpu(sbi->s_es->s_desc_size) - 2051 offset); 2052 2053 out: 2054 return cpu_to_le16(crc); 2055 } 2056 2057 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2058 struct ext4_group_desc *gdp) 2059 { 2060 if (ext4_has_group_desc_csum(sb) && 2061 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb), 2062 block_group, gdp))) 2063 return 0; 2064 2065 return 1; 2066 } 2067 2068 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2069 struct ext4_group_desc *gdp) 2070 { 2071 if (!ext4_has_group_desc_csum(sb)) 2072 return; 2073 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp); 2074 } 2075 2076 /* Called at mount-time, super-block is locked */ 2077 static int ext4_check_descriptors(struct super_block *sb, 2078 ext4_group_t *first_not_zeroed) 2079 { 2080 struct ext4_sb_info *sbi = EXT4_SB(sb); 2081 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2082 ext4_fsblk_t last_block; 2083 ext4_fsblk_t block_bitmap; 2084 ext4_fsblk_t inode_bitmap; 2085 ext4_fsblk_t inode_table; 2086 int flexbg_flag = 0; 2087 ext4_group_t i, grp = sbi->s_groups_count; 2088 2089 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 2090 flexbg_flag = 1; 2091 2092 ext4_debug("Checking group descriptors"); 2093 2094 for (i = 0; i < sbi->s_groups_count; i++) { 2095 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2096 2097 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2098 last_block = ext4_blocks_count(sbi->s_es) - 1; 2099 else 2100 last_block = first_block + 2101 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2102 2103 if ((grp == sbi->s_groups_count) && 2104 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2105 grp = i; 2106 2107 block_bitmap = ext4_block_bitmap(sb, gdp); 2108 if (block_bitmap < first_block || block_bitmap > last_block) { 2109 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2110 "Block bitmap for group %u not in group " 2111 "(block %llu)!", i, block_bitmap); 2112 return 0; 2113 } 2114 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2115 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2116 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2117 "Inode bitmap for group %u not in group " 2118 "(block %llu)!", i, inode_bitmap); 2119 return 0; 2120 } 2121 inode_table = ext4_inode_table(sb, gdp); 2122 if (inode_table < first_block || 2123 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2124 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2125 "Inode table for group %u not in group " 2126 "(block %llu)!", i, inode_table); 2127 return 0; 2128 } 2129 ext4_lock_group(sb, i); 2130 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2131 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2132 "Checksum for group %u failed (%u!=%u)", 2133 i, le16_to_cpu(ext4_group_desc_csum(sbi, i, 2134 gdp)), le16_to_cpu(gdp->bg_checksum)); 2135 if (!(sb->s_flags & MS_RDONLY)) { 2136 ext4_unlock_group(sb, i); 2137 return 0; 2138 } 2139 } 2140 ext4_unlock_group(sb, i); 2141 if (!flexbg_flag) 2142 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2143 } 2144 if (NULL != first_not_zeroed) 2145 *first_not_zeroed = grp; 2146 2147 ext4_free_blocks_count_set(sbi->s_es, 2148 EXT4_C2B(sbi, ext4_count_free_clusters(sb))); 2149 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2150 return 1; 2151 } 2152 2153 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2154 * the superblock) which were deleted from all directories, but held open by 2155 * a process at the time of a crash. We walk the list and try to delete these 2156 * inodes at recovery time (only with a read-write filesystem). 2157 * 2158 * In order to keep the orphan inode chain consistent during traversal (in 2159 * case of crash during recovery), we link each inode into the superblock 2160 * orphan list_head and handle it the same way as an inode deletion during 2161 * normal operation (which journals the operations for us). 2162 * 2163 * We only do an iget() and an iput() on each inode, which is very safe if we 2164 * accidentally point at an in-use or already deleted inode. The worst that 2165 * can happen in this case is that we get a "bit already cleared" message from 2166 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2167 * e2fsck was run on this filesystem, and it must have already done the orphan 2168 * inode cleanup for us, so we can safely abort without any further action. 2169 */ 2170 static void ext4_orphan_cleanup(struct super_block *sb, 2171 struct ext4_super_block *es) 2172 { 2173 unsigned int s_flags = sb->s_flags; 2174 int nr_orphans = 0, nr_truncates = 0; 2175 #ifdef CONFIG_QUOTA 2176 int i; 2177 #endif 2178 if (!es->s_last_orphan) { 2179 jbd_debug(4, "no orphan inodes to clean up\n"); 2180 return; 2181 } 2182 2183 if (bdev_read_only(sb->s_bdev)) { 2184 ext4_msg(sb, KERN_ERR, "write access " 2185 "unavailable, skipping orphan cleanup"); 2186 return; 2187 } 2188 2189 /* Check if feature set would not allow a r/w mount */ 2190 if (!ext4_feature_set_ok(sb, 0)) { 2191 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2192 "unknown ROCOMPAT features"); 2193 return; 2194 } 2195 2196 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2197 /* don't clear list on RO mount w/ errors */ 2198 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { 2199 jbd_debug(1, "Errors on filesystem, " 2200 "clearing orphan list.\n"); 2201 es->s_last_orphan = 0; 2202 } 2203 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2204 return; 2205 } 2206 2207 if (s_flags & MS_RDONLY) { 2208 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2209 sb->s_flags &= ~MS_RDONLY; 2210 } 2211 #ifdef CONFIG_QUOTA 2212 /* Needed for iput() to work correctly and not trash data */ 2213 sb->s_flags |= MS_ACTIVE; 2214 /* Turn on quotas so that they are updated correctly */ 2215 for (i = 0; i < MAXQUOTAS; i++) { 2216 if (EXT4_SB(sb)->s_qf_names[i]) { 2217 int ret = ext4_quota_on_mount(sb, i); 2218 if (ret < 0) 2219 ext4_msg(sb, KERN_ERR, 2220 "Cannot turn on journaled " 2221 "quota: error %d", ret); 2222 } 2223 } 2224 #endif 2225 2226 while (es->s_last_orphan) { 2227 struct inode *inode; 2228 2229 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2230 if (IS_ERR(inode)) { 2231 es->s_last_orphan = 0; 2232 break; 2233 } 2234 2235 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2236 dquot_initialize(inode); 2237 if (inode->i_nlink) { 2238 if (test_opt(sb, DEBUG)) 2239 ext4_msg(sb, KERN_DEBUG, 2240 "%s: truncating inode %lu to %lld bytes", 2241 __func__, inode->i_ino, inode->i_size); 2242 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2243 inode->i_ino, inode->i_size); 2244 mutex_lock(&inode->i_mutex); 2245 truncate_inode_pages(inode->i_mapping, inode->i_size); 2246 ext4_truncate(inode); 2247 mutex_unlock(&inode->i_mutex); 2248 nr_truncates++; 2249 } else { 2250 if (test_opt(sb, DEBUG)) 2251 ext4_msg(sb, KERN_DEBUG, 2252 "%s: deleting unreferenced inode %lu", 2253 __func__, inode->i_ino); 2254 jbd_debug(2, "deleting unreferenced inode %lu\n", 2255 inode->i_ino); 2256 nr_orphans++; 2257 } 2258 iput(inode); /* The delete magic happens here! */ 2259 } 2260 2261 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2262 2263 if (nr_orphans) 2264 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2265 PLURAL(nr_orphans)); 2266 if (nr_truncates) 2267 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2268 PLURAL(nr_truncates)); 2269 #ifdef CONFIG_QUOTA 2270 /* Turn quotas off */ 2271 for (i = 0; i < MAXQUOTAS; i++) { 2272 if (sb_dqopt(sb)->files[i]) 2273 dquot_quota_off(sb, i); 2274 } 2275 #endif 2276 sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2277 } 2278 2279 /* 2280 * Maximal extent format file size. 2281 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2282 * extent format containers, within a sector_t, and within i_blocks 2283 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2284 * so that won't be a limiting factor. 2285 * 2286 * However there is other limiting factor. We do store extents in the form 2287 * of starting block and length, hence the resulting length of the extent 2288 * covering maximum file size must fit into on-disk format containers as 2289 * well. Given that length is always by 1 unit bigger than max unit (because 2290 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2291 * 2292 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2293 */ 2294 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2295 { 2296 loff_t res; 2297 loff_t upper_limit = MAX_LFS_FILESIZE; 2298 2299 /* small i_blocks in vfs inode? */ 2300 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2301 /* 2302 * CONFIG_LBDAF is not enabled implies the inode 2303 * i_block represent total blocks in 512 bytes 2304 * 32 == size of vfs inode i_blocks * 8 2305 */ 2306 upper_limit = (1LL << 32) - 1; 2307 2308 /* total blocks in file system block size */ 2309 upper_limit >>= (blkbits - 9); 2310 upper_limit <<= blkbits; 2311 } 2312 2313 /* 2314 * 32-bit extent-start container, ee_block. We lower the maxbytes 2315 * by one fs block, so ee_len can cover the extent of maximum file 2316 * size 2317 */ 2318 res = (1LL << 32) - 1; 2319 res <<= blkbits; 2320 2321 /* Sanity check against vm- & vfs- imposed limits */ 2322 if (res > upper_limit) 2323 res = upper_limit; 2324 2325 return res; 2326 } 2327 2328 /* 2329 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2330 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2331 * We need to be 1 filesystem block less than the 2^48 sector limit. 2332 */ 2333 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2334 { 2335 loff_t res = EXT4_NDIR_BLOCKS; 2336 int meta_blocks; 2337 loff_t upper_limit; 2338 /* This is calculated to be the largest file size for a dense, block 2339 * mapped file such that the file's total number of 512-byte sectors, 2340 * including data and all indirect blocks, does not exceed (2^48 - 1). 2341 * 2342 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2343 * number of 512-byte sectors of the file. 2344 */ 2345 2346 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2347 /* 2348 * !has_huge_files or CONFIG_LBDAF not enabled implies that 2349 * the inode i_block field represents total file blocks in 2350 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 2351 */ 2352 upper_limit = (1LL << 32) - 1; 2353 2354 /* total blocks in file system block size */ 2355 upper_limit >>= (bits - 9); 2356 2357 } else { 2358 /* 2359 * We use 48 bit ext4_inode i_blocks 2360 * With EXT4_HUGE_FILE_FL set the i_blocks 2361 * represent total number of blocks in 2362 * file system block size 2363 */ 2364 upper_limit = (1LL << 48) - 1; 2365 2366 } 2367 2368 /* indirect blocks */ 2369 meta_blocks = 1; 2370 /* double indirect blocks */ 2371 meta_blocks += 1 + (1LL << (bits-2)); 2372 /* tripple indirect blocks */ 2373 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2374 2375 upper_limit -= meta_blocks; 2376 upper_limit <<= bits; 2377 2378 res += 1LL << (bits-2); 2379 res += 1LL << (2*(bits-2)); 2380 res += 1LL << (3*(bits-2)); 2381 res <<= bits; 2382 if (res > upper_limit) 2383 res = upper_limit; 2384 2385 if (res > MAX_LFS_FILESIZE) 2386 res = MAX_LFS_FILESIZE; 2387 2388 return res; 2389 } 2390 2391 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2392 ext4_fsblk_t logical_sb_block, int nr) 2393 { 2394 struct ext4_sb_info *sbi = EXT4_SB(sb); 2395 ext4_group_t bg, first_meta_bg; 2396 int has_super = 0; 2397 2398 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2399 2400 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || 2401 nr < first_meta_bg) 2402 return logical_sb_block + nr + 1; 2403 bg = sbi->s_desc_per_block * nr; 2404 if (ext4_bg_has_super(sb, bg)) 2405 has_super = 1; 2406 2407 return (has_super + ext4_group_first_block_no(sb, bg)); 2408 } 2409 2410 /** 2411 * ext4_get_stripe_size: Get the stripe size. 2412 * @sbi: In memory super block info 2413 * 2414 * If we have specified it via mount option, then 2415 * use the mount option value. If the value specified at mount time is 2416 * greater than the blocks per group use the super block value. 2417 * If the super block value is greater than blocks per group return 0. 2418 * Allocator needs it be less than blocks per group. 2419 * 2420 */ 2421 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2422 { 2423 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2424 unsigned long stripe_width = 2425 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2426 int ret; 2427 2428 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2429 ret = sbi->s_stripe; 2430 else if (stripe_width <= sbi->s_blocks_per_group) 2431 ret = stripe_width; 2432 else if (stride <= sbi->s_blocks_per_group) 2433 ret = stride; 2434 else 2435 ret = 0; 2436 2437 /* 2438 * If the stripe width is 1, this makes no sense and 2439 * we set it to 0 to turn off stripe handling code. 2440 */ 2441 if (ret <= 1) 2442 ret = 0; 2443 2444 return ret; 2445 } 2446 2447 /* sysfs supprt */ 2448 2449 struct ext4_attr { 2450 struct attribute attr; 2451 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); 2452 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, 2453 const char *, size_t); 2454 union { 2455 int offset; 2456 int deprecated_val; 2457 } u; 2458 }; 2459 2460 static int parse_strtoull(const char *buf, 2461 unsigned long long max, unsigned long long *value) 2462 { 2463 int ret; 2464 2465 ret = kstrtoull(skip_spaces(buf), 0, value); 2466 if (!ret && *value > max) 2467 ret = -EINVAL; 2468 return ret; 2469 } 2470 2471 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, 2472 struct ext4_sb_info *sbi, 2473 char *buf) 2474 { 2475 return snprintf(buf, PAGE_SIZE, "%llu\n", 2476 (s64) EXT4_C2B(sbi, 2477 percpu_counter_sum(&sbi->s_dirtyclusters_counter))); 2478 } 2479 2480 static ssize_t session_write_kbytes_show(struct ext4_attr *a, 2481 struct ext4_sb_info *sbi, char *buf) 2482 { 2483 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2484 2485 if (!sb->s_bdev->bd_part) 2486 return snprintf(buf, PAGE_SIZE, "0\n"); 2487 return snprintf(buf, PAGE_SIZE, "%lu\n", 2488 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2489 sbi->s_sectors_written_start) >> 1); 2490 } 2491 2492 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, 2493 struct ext4_sb_info *sbi, char *buf) 2494 { 2495 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2496 2497 if (!sb->s_bdev->bd_part) 2498 return snprintf(buf, PAGE_SIZE, "0\n"); 2499 return snprintf(buf, PAGE_SIZE, "%llu\n", 2500 (unsigned long long)(sbi->s_kbytes_written + 2501 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2502 EXT4_SB(sb)->s_sectors_written_start) >> 1))); 2503 } 2504 2505 static ssize_t inode_readahead_blks_store(struct ext4_attr *a, 2506 struct ext4_sb_info *sbi, 2507 const char *buf, size_t count) 2508 { 2509 unsigned long t; 2510 int ret; 2511 2512 ret = kstrtoul(skip_spaces(buf), 0, &t); 2513 if (ret) 2514 return ret; 2515 2516 if (t && (!is_power_of_2(t) || t > 0x40000000)) 2517 return -EINVAL; 2518 2519 sbi->s_inode_readahead_blks = t; 2520 return count; 2521 } 2522 2523 static ssize_t sbi_ui_show(struct ext4_attr *a, 2524 struct ext4_sb_info *sbi, char *buf) 2525 { 2526 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset); 2527 2528 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 2529 } 2530 2531 static ssize_t sbi_ui_store(struct ext4_attr *a, 2532 struct ext4_sb_info *sbi, 2533 const char *buf, size_t count) 2534 { 2535 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset); 2536 unsigned long t; 2537 int ret; 2538 2539 ret = kstrtoul(skip_spaces(buf), 0, &t); 2540 if (ret) 2541 return ret; 2542 *ui = t; 2543 return count; 2544 } 2545 2546 static ssize_t reserved_clusters_show(struct ext4_attr *a, 2547 struct ext4_sb_info *sbi, char *buf) 2548 { 2549 return snprintf(buf, PAGE_SIZE, "%llu\n", 2550 (unsigned long long) atomic64_read(&sbi->s_resv_clusters)); 2551 } 2552 2553 static ssize_t reserved_clusters_store(struct ext4_attr *a, 2554 struct ext4_sb_info *sbi, 2555 const char *buf, size_t count) 2556 { 2557 unsigned long long val; 2558 int ret; 2559 2560 if (parse_strtoull(buf, -1ULL, &val)) 2561 return -EINVAL; 2562 ret = ext4_reserve_clusters(sbi, val); 2563 2564 return ret ? ret : count; 2565 } 2566 2567 static ssize_t trigger_test_error(struct ext4_attr *a, 2568 struct ext4_sb_info *sbi, 2569 const char *buf, size_t count) 2570 { 2571 int len = count; 2572 2573 if (!capable(CAP_SYS_ADMIN)) 2574 return -EPERM; 2575 2576 if (len && buf[len-1] == '\n') 2577 len--; 2578 2579 if (len) 2580 ext4_error(sbi->s_sb, "%.*s", len, buf); 2581 return count; 2582 } 2583 2584 static ssize_t sbi_deprecated_show(struct ext4_attr *a, 2585 struct ext4_sb_info *sbi, char *buf) 2586 { 2587 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val); 2588 } 2589 2590 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ 2591 static struct ext4_attr ext4_attr_##_name = { \ 2592 .attr = {.name = __stringify(_name), .mode = _mode }, \ 2593 .show = _show, \ 2594 .store = _store, \ 2595 .u = { \ 2596 .offset = offsetof(struct ext4_sb_info, _elname),\ 2597 }, \ 2598 } 2599 #define EXT4_ATTR(name, mode, show, store) \ 2600 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) 2601 2602 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) 2603 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) 2604 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) 2605 #define EXT4_RW_ATTR_SBI_UI(name, elname) \ 2606 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) 2607 #define ATTR_LIST(name) &ext4_attr_##name.attr 2608 #define EXT4_DEPRECATED_ATTR(_name, _val) \ 2609 static struct ext4_attr ext4_attr_##_name = { \ 2610 .attr = {.name = __stringify(_name), .mode = 0444 }, \ 2611 .show = sbi_deprecated_show, \ 2612 .u = { \ 2613 .deprecated_val = _val, \ 2614 }, \ 2615 } 2616 2617 EXT4_RO_ATTR(delayed_allocation_blocks); 2618 EXT4_RO_ATTR(session_write_kbytes); 2619 EXT4_RO_ATTR(lifetime_write_kbytes); 2620 EXT4_RW_ATTR(reserved_clusters); 2621 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, 2622 inode_readahead_blks_store, s_inode_readahead_blks); 2623 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); 2624 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); 2625 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); 2626 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); 2627 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); 2628 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); 2629 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); 2630 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128); 2631 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb); 2632 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error); 2633 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval); 2634 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst); 2635 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval); 2636 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst); 2637 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval); 2638 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst); 2639 2640 static struct attribute *ext4_attrs[] = { 2641 ATTR_LIST(delayed_allocation_blocks), 2642 ATTR_LIST(session_write_kbytes), 2643 ATTR_LIST(lifetime_write_kbytes), 2644 ATTR_LIST(reserved_clusters), 2645 ATTR_LIST(inode_readahead_blks), 2646 ATTR_LIST(inode_goal), 2647 ATTR_LIST(mb_stats), 2648 ATTR_LIST(mb_max_to_scan), 2649 ATTR_LIST(mb_min_to_scan), 2650 ATTR_LIST(mb_order2_req), 2651 ATTR_LIST(mb_stream_req), 2652 ATTR_LIST(mb_group_prealloc), 2653 ATTR_LIST(max_writeback_mb_bump), 2654 ATTR_LIST(extent_max_zeroout_kb), 2655 ATTR_LIST(trigger_fs_error), 2656 ATTR_LIST(err_ratelimit_interval_ms), 2657 ATTR_LIST(err_ratelimit_burst), 2658 ATTR_LIST(warning_ratelimit_interval_ms), 2659 ATTR_LIST(warning_ratelimit_burst), 2660 ATTR_LIST(msg_ratelimit_interval_ms), 2661 ATTR_LIST(msg_ratelimit_burst), 2662 NULL, 2663 }; 2664 2665 /* Features this copy of ext4 supports */ 2666 EXT4_INFO_ATTR(lazy_itable_init); 2667 EXT4_INFO_ATTR(batched_discard); 2668 EXT4_INFO_ATTR(meta_bg_resize); 2669 2670 static struct attribute *ext4_feat_attrs[] = { 2671 ATTR_LIST(lazy_itable_init), 2672 ATTR_LIST(batched_discard), 2673 ATTR_LIST(meta_bg_resize), 2674 NULL, 2675 }; 2676 2677 static ssize_t ext4_attr_show(struct kobject *kobj, 2678 struct attribute *attr, char *buf) 2679 { 2680 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2681 s_kobj); 2682 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2683 2684 return a->show ? a->show(a, sbi, buf) : 0; 2685 } 2686 2687 static ssize_t ext4_attr_store(struct kobject *kobj, 2688 struct attribute *attr, 2689 const char *buf, size_t len) 2690 { 2691 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2692 s_kobj); 2693 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2694 2695 return a->store ? a->store(a, sbi, buf, len) : 0; 2696 } 2697 2698 static void ext4_sb_release(struct kobject *kobj) 2699 { 2700 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2701 s_kobj); 2702 complete(&sbi->s_kobj_unregister); 2703 } 2704 2705 static const struct sysfs_ops ext4_attr_ops = { 2706 .show = ext4_attr_show, 2707 .store = ext4_attr_store, 2708 }; 2709 2710 static struct kobj_type ext4_ktype = { 2711 .default_attrs = ext4_attrs, 2712 .sysfs_ops = &ext4_attr_ops, 2713 .release = ext4_sb_release, 2714 }; 2715 2716 static void ext4_feat_release(struct kobject *kobj) 2717 { 2718 complete(&ext4_feat->f_kobj_unregister); 2719 } 2720 2721 static struct kobj_type ext4_feat_ktype = { 2722 .default_attrs = ext4_feat_attrs, 2723 .sysfs_ops = &ext4_attr_ops, 2724 .release = ext4_feat_release, 2725 }; 2726 2727 /* 2728 * Check whether this filesystem can be mounted based on 2729 * the features present and the RDONLY/RDWR mount requested. 2730 * Returns 1 if this filesystem can be mounted as requested, 2731 * 0 if it cannot be. 2732 */ 2733 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2734 { 2735 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { 2736 ext4_msg(sb, KERN_ERR, 2737 "Couldn't mount because of " 2738 "unsupported optional features (%x)", 2739 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2740 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2741 return 0; 2742 } 2743 2744 if (readonly) 2745 return 1; 2746 2747 /* Check that feature set is OK for a read-write mount */ 2748 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { 2749 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2750 "unsupported optional features (%x)", 2751 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2752 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2753 return 0; 2754 } 2755 /* 2756 * Large file size enabled file system can only be mounted 2757 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2758 */ 2759 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { 2760 if (sizeof(blkcnt_t) < sizeof(u64)) { 2761 ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2762 "cannot be mounted RDWR without " 2763 "CONFIG_LBDAF"); 2764 return 0; 2765 } 2766 } 2767 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) && 2768 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { 2769 ext4_msg(sb, KERN_ERR, 2770 "Can't support bigalloc feature without " 2771 "extents feature\n"); 2772 return 0; 2773 } 2774 2775 #ifndef CONFIG_QUOTA 2776 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 2777 !readonly) { 2778 ext4_msg(sb, KERN_ERR, 2779 "Filesystem with quota feature cannot be mounted RDWR " 2780 "without CONFIG_QUOTA"); 2781 return 0; 2782 } 2783 #endif /* CONFIG_QUOTA */ 2784 return 1; 2785 } 2786 2787 /* 2788 * This function is called once a day if we have errors logged 2789 * on the file system 2790 */ 2791 static void print_daily_error_info(unsigned long arg) 2792 { 2793 struct super_block *sb = (struct super_block *) arg; 2794 struct ext4_sb_info *sbi; 2795 struct ext4_super_block *es; 2796 2797 sbi = EXT4_SB(sb); 2798 es = sbi->s_es; 2799 2800 if (es->s_error_count) 2801 ext4_msg(sb, KERN_NOTICE, "error count: %u", 2802 le32_to_cpu(es->s_error_count)); 2803 if (es->s_first_error_time) { 2804 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", 2805 sb->s_id, le32_to_cpu(es->s_first_error_time), 2806 (int) sizeof(es->s_first_error_func), 2807 es->s_first_error_func, 2808 le32_to_cpu(es->s_first_error_line)); 2809 if (es->s_first_error_ino) 2810 printk(": inode %u", 2811 le32_to_cpu(es->s_first_error_ino)); 2812 if (es->s_first_error_block) 2813 printk(": block %llu", (unsigned long long) 2814 le64_to_cpu(es->s_first_error_block)); 2815 printk("\n"); 2816 } 2817 if (es->s_last_error_time) { 2818 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", 2819 sb->s_id, le32_to_cpu(es->s_last_error_time), 2820 (int) sizeof(es->s_last_error_func), 2821 es->s_last_error_func, 2822 le32_to_cpu(es->s_last_error_line)); 2823 if (es->s_last_error_ino) 2824 printk(": inode %u", 2825 le32_to_cpu(es->s_last_error_ino)); 2826 if (es->s_last_error_block) 2827 printk(": block %llu", (unsigned long long) 2828 le64_to_cpu(es->s_last_error_block)); 2829 printk("\n"); 2830 } 2831 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 2832 } 2833 2834 /* Find next suitable group and run ext4_init_inode_table */ 2835 static int ext4_run_li_request(struct ext4_li_request *elr) 2836 { 2837 struct ext4_group_desc *gdp = NULL; 2838 ext4_group_t group, ngroups; 2839 struct super_block *sb; 2840 unsigned long timeout = 0; 2841 int ret = 0; 2842 2843 sb = elr->lr_super; 2844 ngroups = EXT4_SB(sb)->s_groups_count; 2845 2846 sb_start_write(sb); 2847 for (group = elr->lr_next_group; group < ngroups; group++) { 2848 gdp = ext4_get_group_desc(sb, group, NULL); 2849 if (!gdp) { 2850 ret = 1; 2851 break; 2852 } 2853 2854 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2855 break; 2856 } 2857 2858 if (group >= ngroups) 2859 ret = 1; 2860 2861 if (!ret) { 2862 timeout = jiffies; 2863 ret = ext4_init_inode_table(sb, group, 2864 elr->lr_timeout ? 0 : 1); 2865 if (elr->lr_timeout == 0) { 2866 timeout = (jiffies - timeout) * 2867 elr->lr_sbi->s_li_wait_mult; 2868 elr->lr_timeout = timeout; 2869 } 2870 elr->lr_next_sched = jiffies + elr->lr_timeout; 2871 elr->lr_next_group = group + 1; 2872 } 2873 sb_end_write(sb); 2874 2875 return ret; 2876 } 2877 2878 /* 2879 * Remove lr_request from the list_request and free the 2880 * request structure. Should be called with li_list_mtx held 2881 */ 2882 static void ext4_remove_li_request(struct ext4_li_request *elr) 2883 { 2884 struct ext4_sb_info *sbi; 2885 2886 if (!elr) 2887 return; 2888 2889 sbi = elr->lr_sbi; 2890 2891 list_del(&elr->lr_request); 2892 sbi->s_li_request = NULL; 2893 kfree(elr); 2894 } 2895 2896 static void ext4_unregister_li_request(struct super_block *sb) 2897 { 2898 mutex_lock(&ext4_li_mtx); 2899 if (!ext4_li_info) { 2900 mutex_unlock(&ext4_li_mtx); 2901 return; 2902 } 2903 2904 mutex_lock(&ext4_li_info->li_list_mtx); 2905 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2906 mutex_unlock(&ext4_li_info->li_list_mtx); 2907 mutex_unlock(&ext4_li_mtx); 2908 } 2909 2910 static struct task_struct *ext4_lazyinit_task; 2911 2912 /* 2913 * This is the function where ext4lazyinit thread lives. It walks 2914 * through the request list searching for next scheduled filesystem. 2915 * When such a fs is found, run the lazy initialization request 2916 * (ext4_rn_li_request) and keep track of the time spend in this 2917 * function. Based on that time we compute next schedule time of 2918 * the request. When walking through the list is complete, compute 2919 * next waking time and put itself into sleep. 2920 */ 2921 static int ext4_lazyinit_thread(void *arg) 2922 { 2923 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2924 struct list_head *pos, *n; 2925 struct ext4_li_request *elr; 2926 unsigned long next_wakeup, cur; 2927 2928 BUG_ON(NULL == eli); 2929 2930 cont_thread: 2931 while (true) { 2932 next_wakeup = MAX_JIFFY_OFFSET; 2933 2934 mutex_lock(&eli->li_list_mtx); 2935 if (list_empty(&eli->li_request_list)) { 2936 mutex_unlock(&eli->li_list_mtx); 2937 goto exit_thread; 2938 } 2939 2940 list_for_each_safe(pos, n, &eli->li_request_list) { 2941 elr = list_entry(pos, struct ext4_li_request, 2942 lr_request); 2943 2944 if (time_after_eq(jiffies, elr->lr_next_sched)) { 2945 if (ext4_run_li_request(elr) != 0) { 2946 /* error, remove the lazy_init job */ 2947 ext4_remove_li_request(elr); 2948 continue; 2949 } 2950 } 2951 2952 if (time_before(elr->lr_next_sched, next_wakeup)) 2953 next_wakeup = elr->lr_next_sched; 2954 } 2955 mutex_unlock(&eli->li_list_mtx); 2956 2957 try_to_freeze(); 2958 2959 cur = jiffies; 2960 if ((time_after_eq(cur, next_wakeup)) || 2961 (MAX_JIFFY_OFFSET == next_wakeup)) { 2962 cond_resched(); 2963 continue; 2964 } 2965 2966 schedule_timeout_interruptible(next_wakeup - cur); 2967 2968 if (kthread_should_stop()) { 2969 ext4_clear_request_list(); 2970 goto exit_thread; 2971 } 2972 } 2973 2974 exit_thread: 2975 /* 2976 * It looks like the request list is empty, but we need 2977 * to check it under the li_list_mtx lock, to prevent any 2978 * additions into it, and of course we should lock ext4_li_mtx 2979 * to atomically free the list and ext4_li_info, because at 2980 * this point another ext4 filesystem could be registering 2981 * new one. 2982 */ 2983 mutex_lock(&ext4_li_mtx); 2984 mutex_lock(&eli->li_list_mtx); 2985 if (!list_empty(&eli->li_request_list)) { 2986 mutex_unlock(&eli->li_list_mtx); 2987 mutex_unlock(&ext4_li_mtx); 2988 goto cont_thread; 2989 } 2990 mutex_unlock(&eli->li_list_mtx); 2991 kfree(ext4_li_info); 2992 ext4_li_info = NULL; 2993 mutex_unlock(&ext4_li_mtx); 2994 2995 return 0; 2996 } 2997 2998 static void ext4_clear_request_list(void) 2999 { 3000 struct list_head *pos, *n; 3001 struct ext4_li_request *elr; 3002 3003 mutex_lock(&ext4_li_info->li_list_mtx); 3004 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3005 elr = list_entry(pos, struct ext4_li_request, 3006 lr_request); 3007 ext4_remove_li_request(elr); 3008 } 3009 mutex_unlock(&ext4_li_info->li_list_mtx); 3010 } 3011 3012 static int ext4_run_lazyinit_thread(void) 3013 { 3014 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3015 ext4_li_info, "ext4lazyinit"); 3016 if (IS_ERR(ext4_lazyinit_task)) { 3017 int err = PTR_ERR(ext4_lazyinit_task); 3018 ext4_clear_request_list(); 3019 kfree(ext4_li_info); 3020 ext4_li_info = NULL; 3021 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3022 "initialization thread\n", 3023 err); 3024 return err; 3025 } 3026 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3027 return 0; 3028 } 3029 3030 /* 3031 * Check whether it make sense to run itable init. thread or not. 3032 * If there is at least one uninitialized inode table, return 3033 * corresponding group number, else the loop goes through all 3034 * groups and return total number of groups. 3035 */ 3036 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3037 { 3038 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3039 struct ext4_group_desc *gdp = NULL; 3040 3041 for (group = 0; group < ngroups; group++) { 3042 gdp = ext4_get_group_desc(sb, group, NULL); 3043 if (!gdp) 3044 continue; 3045 3046 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3047 break; 3048 } 3049 3050 return group; 3051 } 3052 3053 static int ext4_li_info_new(void) 3054 { 3055 struct ext4_lazy_init *eli = NULL; 3056 3057 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3058 if (!eli) 3059 return -ENOMEM; 3060 3061 INIT_LIST_HEAD(&eli->li_request_list); 3062 mutex_init(&eli->li_list_mtx); 3063 3064 eli->li_state |= EXT4_LAZYINIT_QUIT; 3065 3066 ext4_li_info = eli; 3067 3068 return 0; 3069 } 3070 3071 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3072 ext4_group_t start) 3073 { 3074 struct ext4_sb_info *sbi = EXT4_SB(sb); 3075 struct ext4_li_request *elr; 3076 3077 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3078 if (!elr) 3079 return NULL; 3080 3081 elr->lr_super = sb; 3082 elr->lr_sbi = sbi; 3083 elr->lr_next_group = start; 3084 3085 /* 3086 * Randomize first schedule time of the request to 3087 * spread the inode table initialization requests 3088 * better. 3089 */ 3090 elr->lr_next_sched = jiffies + (prandom_u32() % 3091 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3092 return elr; 3093 } 3094 3095 int ext4_register_li_request(struct super_block *sb, 3096 ext4_group_t first_not_zeroed) 3097 { 3098 struct ext4_sb_info *sbi = EXT4_SB(sb); 3099 struct ext4_li_request *elr = NULL; 3100 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3101 int ret = 0; 3102 3103 mutex_lock(&ext4_li_mtx); 3104 if (sbi->s_li_request != NULL) { 3105 /* 3106 * Reset timeout so it can be computed again, because 3107 * s_li_wait_mult might have changed. 3108 */ 3109 sbi->s_li_request->lr_timeout = 0; 3110 goto out; 3111 } 3112 3113 if (first_not_zeroed == ngroups || 3114 (sb->s_flags & MS_RDONLY) || 3115 !test_opt(sb, INIT_INODE_TABLE)) 3116 goto out; 3117 3118 elr = ext4_li_request_new(sb, first_not_zeroed); 3119 if (!elr) { 3120 ret = -ENOMEM; 3121 goto out; 3122 } 3123 3124 if (NULL == ext4_li_info) { 3125 ret = ext4_li_info_new(); 3126 if (ret) 3127 goto out; 3128 } 3129 3130 mutex_lock(&ext4_li_info->li_list_mtx); 3131 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3132 mutex_unlock(&ext4_li_info->li_list_mtx); 3133 3134 sbi->s_li_request = elr; 3135 /* 3136 * set elr to NULL here since it has been inserted to 3137 * the request_list and the removal and free of it is 3138 * handled by ext4_clear_request_list from now on. 3139 */ 3140 elr = NULL; 3141 3142 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3143 ret = ext4_run_lazyinit_thread(); 3144 if (ret) 3145 goto out; 3146 } 3147 out: 3148 mutex_unlock(&ext4_li_mtx); 3149 if (ret) 3150 kfree(elr); 3151 return ret; 3152 } 3153 3154 /* 3155 * We do not need to lock anything since this is called on 3156 * module unload. 3157 */ 3158 static void ext4_destroy_lazyinit_thread(void) 3159 { 3160 /* 3161 * If thread exited earlier 3162 * there's nothing to be done. 3163 */ 3164 if (!ext4_li_info || !ext4_lazyinit_task) 3165 return; 3166 3167 kthread_stop(ext4_lazyinit_task); 3168 } 3169 3170 static int set_journal_csum_feature_set(struct super_block *sb) 3171 { 3172 int ret = 1; 3173 int compat, incompat; 3174 struct ext4_sb_info *sbi = EXT4_SB(sb); 3175 3176 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3177 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 3178 /* journal checksum v2 */ 3179 compat = 0; 3180 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2; 3181 } else { 3182 /* journal checksum v1 */ 3183 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3184 incompat = 0; 3185 } 3186 3187 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3188 ret = jbd2_journal_set_features(sbi->s_journal, 3189 compat, 0, 3190 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3191 incompat); 3192 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3193 ret = jbd2_journal_set_features(sbi->s_journal, 3194 compat, 0, 3195 incompat); 3196 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3197 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3198 } else { 3199 jbd2_journal_clear_features(sbi->s_journal, 3200 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3201 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3202 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3203 } 3204 3205 return ret; 3206 } 3207 3208 /* 3209 * Note: calculating the overhead so we can be compatible with 3210 * historical BSD practice is quite difficult in the face of 3211 * clusters/bigalloc. This is because multiple metadata blocks from 3212 * different block group can end up in the same allocation cluster. 3213 * Calculating the exact overhead in the face of clustered allocation 3214 * requires either O(all block bitmaps) in memory or O(number of block 3215 * groups**2) in time. We will still calculate the superblock for 3216 * older file systems --- and if we come across with a bigalloc file 3217 * system with zero in s_overhead_clusters the estimate will be close to 3218 * correct especially for very large cluster sizes --- but for newer 3219 * file systems, it's better to calculate this figure once at mkfs 3220 * time, and store it in the superblock. If the superblock value is 3221 * present (even for non-bigalloc file systems), we will use it. 3222 */ 3223 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3224 char *buf) 3225 { 3226 struct ext4_sb_info *sbi = EXT4_SB(sb); 3227 struct ext4_group_desc *gdp; 3228 ext4_fsblk_t first_block, last_block, b; 3229 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3230 int s, j, count = 0; 3231 3232 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC)) 3233 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3234 sbi->s_itb_per_group + 2); 3235 3236 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3237 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3238 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3239 for (i = 0; i < ngroups; i++) { 3240 gdp = ext4_get_group_desc(sb, i, NULL); 3241 b = ext4_block_bitmap(sb, gdp); 3242 if (b >= first_block && b <= last_block) { 3243 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3244 count++; 3245 } 3246 b = ext4_inode_bitmap(sb, gdp); 3247 if (b >= first_block && b <= last_block) { 3248 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3249 count++; 3250 } 3251 b = ext4_inode_table(sb, gdp); 3252 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3253 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3254 int c = EXT4_B2C(sbi, b - first_block); 3255 ext4_set_bit(c, buf); 3256 count++; 3257 } 3258 if (i != grp) 3259 continue; 3260 s = 0; 3261 if (ext4_bg_has_super(sb, grp)) { 3262 ext4_set_bit(s++, buf); 3263 count++; 3264 } 3265 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { 3266 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3267 count++; 3268 } 3269 } 3270 if (!count) 3271 return 0; 3272 return EXT4_CLUSTERS_PER_GROUP(sb) - 3273 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3274 } 3275 3276 /* 3277 * Compute the overhead and stash it in sbi->s_overhead 3278 */ 3279 int ext4_calculate_overhead(struct super_block *sb) 3280 { 3281 struct ext4_sb_info *sbi = EXT4_SB(sb); 3282 struct ext4_super_block *es = sbi->s_es; 3283 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3284 ext4_fsblk_t overhead = 0; 3285 char *buf = (char *) get_zeroed_page(GFP_KERNEL); 3286 3287 if (!buf) 3288 return -ENOMEM; 3289 3290 /* 3291 * Compute the overhead (FS structures). This is constant 3292 * for a given filesystem unless the number of block groups 3293 * changes so we cache the previous value until it does. 3294 */ 3295 3296 /* 3297 * All of the blocks before first_data_block are overhead 3298 */ 3299 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3300 3301 /* 3302 * Add the overhead found in each block group 3303 */ 3304 for (i = 0; i < ngroups; i++) { 3305 int blks; 3306 3307 blks = count_overhead(sb, i, buf); 3308 overhead += blks; 3309 if (blks) 3310 memset(buf, 0, PAGE_SIZE); 3311 cond_resched(); 3312 } 3313 /* Add the journal blocks as well */ 3314 if (sbi->s_journal) 3315 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3316 3317 sbi->s_overhead = overhead; 3318 smp_wmb(); 3319 free_page((unsigned long) buf); 3320 return 0; 3321 } 3322 3323 3324 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb) 3325 { 3326 ext4_fsblk_t resv_clusters; 3327 3328 /* 3329 * There's no need to reserve anything when we aren't using extents. 3330 * The space estimates are exact, there are no unwritten extents, 3331 * hole punching doesn't need new metadata... This is needed especially 3332 * to keep ext2/3 backward compatibility. 3333 */ 3334 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) 3335 return 0; 3336 /* 3337 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3338 * This should cover the situations where we can not afford to run 3339 * out of space like for example punch hole, or converting 3340 * uninitialized extents in delalloc path. In most cases such 3341 * allocation would require 1, or 2 blocks, higher numbers are 3342 * very rare. 3343 */ 3344 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >> 3345 EXT4_SB(sb)->s_cluster_bits; 3346 3347 do_div(resv_clusters, 50); 3348 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3349 3350 return resv_clusters; 3351 } 3352 3353 3354 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count) 3355 { 3356 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >> 3357 sbi->s_cluster_bits; 3358 3359 if (count >= clusters) 3360 return -EINVAL; 3361 3362 atomic64_set(&sbi->s_resv_clusters, count); 3363 return 0; 3364 } 3365 3366 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3367 { 3368 char *orig_data = kstrdup(data, GFP_KERNEL); 3369 struct buffer_head *bh; 3370 struct ext4_super_block *es = NULL; 3371 struct ext4_sb_info *sbi; 3372 ext4_fsblk_t block; 3373 ext4_fsblk_t sb_block = get_sb_block(&data); 3374 ext4_fsblk_t logical_sb_block; 3375 unsigned long offset = 0; 3376 unsigned long journal_devnum = 0; 3377 unsigned long def_mount_opts; 3378 struct inode *root; 3379 char *cp; 3380 const char *descr; 3381 int ret = -ENOMEM; 3382 int blocksize, clustersize; 3383 unsigned int db_count; 3384 unsigned int i; 3385 int needs_recovery, has_huge_files, has_bigalloc; 3386 __u64 blocks_count; 3387 int err = 0; 3388 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3389 ext4_group_t first_not_zeroed; 3390 3391 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3392 if (!sbi) 3393 goto out_free_orig; 3394 3395 sbi->s_blockgroup_lock = 3396 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3397 if (!sbi->s_blockgroup_lock) { 3398 kfree(sbi); 3399 goto out_free_orig; 3400 } 3401 sb->s_fs_info = sbi; 3402 sbi->s_sb = sb; 3403 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3404 sbi->s_sb_block = sb_block; 3405 if (sb->s_bdev->bd_part) 3406 sbi->s_sectors_written_start = 3407 part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3408 3409 /* Cleanup superblock name */ 3410 for (cp = sb->s_id; (cp = strchr(cp, '/'));) 3411 *cp = '!'; 3412 3413 /* -EINVAL is default */ 3414 ret = -EINVAL; 3415 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3416 if (!blocksize) { 3417 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3418 goto out_fail; 3419 } 3420 3421 /* 3422 * The ext4 superblock will not be buffer aligned for other than 1kB 3423 * block sizes. We need to calculate the offset from buffer start. 3424 */ 3425 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3426 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3427 offset = do_div(logical_sb_block, blocksize); 3428 } else { 3429 logical_sb_block = sb_block; 3430 } 3431 3432 if (!(bh = sb_bread(sb, logical_sb_block))) { 3433 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3434 goto out_fail; 3435 } 3436 /* 3437 * Note: s_es must be initialized as soon as possible because 3438 * some ext4 macro-instructions depend on its value 3439 */ 3440 es = (struct ext4_super_block *) (bh->b_data + offset); 3441 sbi->s_es = es; 3442 sb->s_magic = le16_to_cpu(es->s_magic); 3443 if (sb->s_magic != EXT4_SUPER_MAGIC) 3444 goto cantfind_ext4; 3445 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3446 3447 /* Warn if metadata_csum and gdt_csum are both set. */ 3448 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3449 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 3450 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) 3451 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are " 3452 "redundant flags; please run fsck."); 3453 3454 /* Check for a known checksum algorithm */ 3455 if (!ext4_verify_csum_type(sb, es)) { 3456 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3457 "unknown checksum algorithm."); 3458 silent = 1; 3459 goto cantfind_ext4; 3460 } 3461 3462 /* Load the checksum driver */ 3463 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3464 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 3465 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3466 if (IS_ERR(sbi->s_chksum_driver)) { 3467 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3468 ret = PTR_ERR(sbi->s_chksum_driver); 3469 sbi->s_chksum_driver = NULL; 3470 goto failed_mount; 3471 } 3472 } 3473 3474 /* Check superblock checksum */ 3475 if (!ext4_superblock_csum_verify(sb, es)) { 3476 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3477 "invalid superblock checksum. Run e2fsck?"); 3478 silent = 1; 3479 goto cantfind_ext4; 3480 } 3481 3482 /* Precompute checksum seed for all metadata */ 3483 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3484 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 3485 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3486 sizeof(es->s_uuid)); 3487 3488 /* Set defaults before we parse the mount options */ 3489 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3490 set_opt(sb, INIT_INODE_TABLE); 3491 if (def_mount_opts & EXT4_DEFM_DEBUG) 3492 set_opt(sb, DEBUG); 3493 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3494 set_opt(sb, GRPID); 3495 if (def_mount_opts & EXT4_DEFM_UID16) 3496 set_opt(sb, NO_UID32); 3497 /* xattr user namespace & acls are now defaulted on */ 3498 set_opt(sb, XATTR_USER); 3499 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3500 set_opt(sb, POSIX_ACL); 3501 #endif 3502 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3503 set_opt(sb, JOURNAL_DATA); 3504 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3505 set_opt(sb, ORDERED_DATA); 3506 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3507 set_opt(sb, WRITEBACK_DATA); 3508 3509 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3510 set_opt(sb, ERRORS_PANIC); 3511 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3512 set_opt(sb, ERRORS_CONT); 3513 else 3514 set_opt(sb, ERRORS_RO); 3515 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) 3516 set_opt(sb, BLOCK_VALIDITY); 3517 if (def_mount_opts & EXT4_DEFM_DISCARD) 3518 set_opt(sb, DISCARD); 3519 3520 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3521 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3522 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3523 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3524 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3525 3526 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3527 set_opt(sb, BARRIER); 3528 3529 /* 3530 * enable delayed allocation by default 3531 * Use -o nodelalloc to turn it off 3532 */ 3533 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3534 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3535 set_opt(sb, DELALLOC); 3536 3537 /* 3538 * set default s_li_wait_mult for lazyinit, for the case there is 3539 * no mount option specified. 3540 */ 3541 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3542 3543 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3544 &journal_devnum, &journal_ioprio, 0)) { 3545 ext4_msg(sb, KERN_WARNING, 3546 "failed to parse options in superblock: %s", 3547 sbi->s_es->s_mount_opts); 3548 } 3549 sbi->s_def_mount_opt = sbi->s_mount_opt; 3550 if (!parse_options((char *) data, sb, &journal_devnum, 3551 &journal_ioprio, 0)) 3552 goto failed_mount; 3553 3554 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3555 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3556 "with data=journal disables delayed " 3557 "allocation and O_DIRECT support!\n"); 3558 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3559 ext4_msg(sb, KERN_ERR, "can't mount with " 3560 "both data=journal and delalloc"); 3561 goto failed_mount; 3562 } 3563 if (test_opt(sb, DIOREAD_NOLOCK)) { 3564 ext4_msg(sb, KERN_ERR, "can't mount with " 3565 "both data=journal and dioread_nolock"); 3566 goto failed_mount; 3567 } 3568 if (test_opt(sb, DELALLOC)) 3569 clear_opt(sb, DELALLOC); 3570 } 3571 3572 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3573 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3574 3575 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3576 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || 3577 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || 3578 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) 3579 ext4_msg(sb, KERN_WARNING, 3580 "feature flags set on rev 0 fs, " 3581 "running e2fsck is recommended"); 3582 3583 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3584 set_opt2(sb, HURD_COMPAT); 3585 if (EXT4_HAS_INCOMPAT_FEATURE(sb, 3586 EXT4_FEATURE_INCOMPAT_64BIT)) { 3587 ext4_msg(sb, KERN_ERR, 3588 "The Hurd can't support 64-bit file systems"); 3589 goto failed_mount; 3590 } 3591 } 3592 3593 if (IS_EXT2_SB(sb)) { 3594 if (ext2_feature_set_ok(sb)) 3595 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3596 "using the ext4 subsystem"); 3597 else { 3598 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3599 "to feature incompatibilities"); 3600 goto failed_mount; 3601 } 3602 } 3603 3604 if (IS_EXT3_SB(sb)) { 3605 if (ext3_feature_set_ok(sb)) 3606 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3607 "using the ext4 subsystem"); 3608 else { 3609 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3610 "to feature incompatibilities"); 3611 goto failed_mount; 3612 } 3613 } 3614 3615 /* 3616 * Check feature flags regardless of the revision level, since we 3617 * previously didn't change the revision level when setting the flags, 3618 * so there is a chance incompat flags are set on a rev 0 filesystem. 3619 */ 3620 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3621 goto failed_mount; 3622 3623 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3624 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3625 blocksize > EXT4_MAX_BLOCK_SIZE) { 3626 ext4_msg(sb, KERN_ERR, 3627 "Unsupported filesystem blocksize %d", blocksize); 3628 goto failed_mount; 3629 } 3630 3631 if (sb->s_blocksize != blocksize) { 3632 /* Validate the filesystem blocksize */ 3633 if (!sb_set_blocksize(sb, blocksize)) { 3634 ext4_msg(sb, KERN_ERR, "bad block size %d", 3635 blocksize); 3636 goto failed_mount; 3637 } 3638 3639 brelse(bh); 3640 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3641 offset = do_div(logical_sb_block, blocksize); 3642 bh = sb_bread(sb, logical_sb_block); 3643 if (!bh) { 3644 ext4_msg(sb, KERN_ERR, 3645 "Can't read superblock on 2nd try"); 3646 goto failed_mount; 3647 } 3648 es = (struct ext4_super_block *)(bh->b_data + offset); 3649 sbi->s_es = es; 3650 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3651 ext4_msg(sb, KERN_ERR, 3652 "Magic mismatch, very weird!"); 3653 goto failed_mount; 3654 } 3655 } 3656 3657 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3658 EXT4_FEATURE_RO_COMPAT_HUGE_FILE); 3659 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3660 has_huge_files); 3661 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3662 3663 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3664 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3665 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3666 } else { 3667 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3668 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3669 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3670 (!is_power_of_2(sbi->s_inode_size)) || 3671 (sbi->s_inode_size > blocksize)) { 3672 ext4_msg(sb, KERN_ERR, 3673 "unsupported inode size: %d", 3674 sbi->s_inode_size); 3675 goto failed_mount; 3676 } 3677 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3678 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3679 } 3680 3681 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3682 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { 3683 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 3684 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3685 !is_power_of_2(sbi->s_desc_size)) { 3686 ext4_msg(sb, KERN_ERR, 3687 "unsupported descriptor size %lu", 3688 sbi->s_desc_size); 3689 goto failed_mount; 3690 } 3691 } else 3692 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 3693 3694 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3695 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3696 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3697 goto cantfind_ext4; 3698 3699 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3700 if (sbi->s_inodes_per_block == 0) 3701 goto cantfind_ext4; 3702 sbi->s_itb_per_group = sbi->s_inodes_per_group / 3703 sbi->s_inodes_per_block; 3704 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3705 sbi->s_sbh = bh; 3706 sbi->s_mount_state = le16_to_cpu(es->s_state); 3707 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3708 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 3709 3710 for (i = 0; i < 4; i++) 3711 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3712 sbi->s_def_hash_version = es->s_def_hash_version; 3713 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) { 3714 i = le32_to_cpu(es->s_flags); 3715 if (i & EXT2_FLAGS_UNSIGNED_HASH) 3716 sbi->s_hash_unsigned = 3; 3717 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3718 #ifdef __CHAR_UNSIGNED__ 3719 if (!(sb->s_flags & MS_RDONLY)) 3720 es->s_flags |= 3721 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3722 sbi->s_hash_unsigned = 3; 3723 #else 3724 if (!(sb->s_flags & MS_RDONLY)) 3725 es->s_flags |= 3726 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3727 #endif 3728 } 3729 } 3730 3731 /* Handle clustersize */ 3732 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3733 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3734 EXT4_FEATURE_RO_COMPAT_BIGALLOC); 3735 if (has_bigalloc) { 3736 if (clustersize < blocksize) { 3737 ext4_msg(sb, KERN_ERR, 3738 "cluster size (%d) smaller than " 3739 "block size (%d)", clustersize, blocksize); 3740 goto failed_mount; 3741 } 3742 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3743 le32_to_cpu(es->s_log_block_size); 3744 sbi->s_clusters_per_group = 3745 le32_to_cpu(es->s_clusters_per_group); 3746 if (sbi->s_clusters_per_group > blocksize * 8) { 3747 ext4_msg(sb, KERN_ERR, 3748 "#clusters per group too big: %lu", 3749 sbi->s_clusters_per_group); 3750 goto failed_mount; 3751 } 3752 if (sbi->s_blocks_per_group != 3753 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3754 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3755 "clusters per group (%lu) inconsistent", 3756 sbi->s_blocks_per_group, 3757 sbi->s_clusters_per_group); 3758 goto failed_mount; 3759 } 3760 } else { 3761 if (clustersize != blocksize) { 3762 ext4_warning(sb, "fragment/cluster size (%d) != " 3763 "block size (%d)", clustersize, 3764 blocksize); 3765 clustersize = blocksize; 3766 } 3767 if (sbi->s_blocks_per_group > blocksize * 8) { 3768 ext4_msg(sb, KERN_ERR, 3769 "#blocks per group too big: %lu", 3770 sbi->s_blocks_per_group); 3771 goto failed_mount; 3772 } 3773 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3774 sbi->s_cluster_bits = 0; 3775 } 3776 sbi->s_cluster_ratio = clustersize / blocksize; 3777 3778 if (sbi->s_inodes_per_group > blocksize * 8) { 3779 ext4_msg(sb, KERN_ERR, 3780 "#inodes per group too big: %lu", 3781 sbi->s_inodes_per_group); 3782 goto failed_mount; 3783 } 3784 3785 /* Do we have standard group size of clustersize * 8 blocks ? */ 3786 if (sbi->s_blocks_per_group == clustersize << 3) 3787 set_opt2(sb, STD_GROUP_SIZE); 3788 3789 /* 3790 * Test whether we have more sectors than will fit in sector_t, 3791 * and whether the max offset is addressable by the page cache. 3792 */ 3793 err = generic_check_addressable(sb->s_blocksize_bits, 3794 ext4_blocks_count(es)); 3795 if (err) { 3796 ext4_msg(sb, KERN_ERR, "filesystem" 3797 " too large to mount safely on this system"); 3798 if (sizeof(sector_t) < 8) 3799 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3800 goto failed_mount; 3801 } 3802 3803 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3804 goto cantfind_ext4; 3805 3806 /* check blocks count against device size */ 3807 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 3808 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3809 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3810 "exceeds size of device (%llu blocks)", 3811 ext4_blocks_count(es), blocks_count); 3812 goto failed_mount; 3813 } 3814 3815 /* 3816 * It makes no sense for the first data block to be beyond the end 3817 * of the filesystem. 3818 */ 3819 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3820 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 3821 "block %u is beyond end of filesystem (%llu)", 3822 le32_to_cpu(es->s_first_data_block), 3823 ext4_blocks_count(es)); 3824 goto failed_mount; 3825 } 3826 blocks_count = (ext4_blocks_count(es) - 3827 le32_to_cpu(es->s_first_data_block) + 3828 EXT4_BLOCKS_PER_GROUP(sb) - 1); 3829 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 3830 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3831 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 3832 "(block count %llu, first data block %u, " 3833 "blocks per group %lu)", sbi->s_groups_count, 3834 ext4_blocks_count(es), 3835 le32_to_cpu(es->s_first_data_block), 3836 EXT4_BLOCKS_PER_GROUP(sb)); 3837 goto failed_mount; 3838 } 3839 sbi->s_groups_count = blocks_count; 3840 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3841 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3842 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3843 EXT4_DESC_PER_BLOCK(sb); 3844 sbi->s_group_desc = ext4_kvmalloc(db_count * 3845 sizeof(struct buffer_head *), 3846 GFP_KERNEL); 3847 if (sbi->s_group_desc == NULL) { 3848 ext4_msg(sb, KERN_ERR, "not enough memory"); 3849 ret = -ENOMEM; 3850 goto failed_mount; 3851 } 3852 3853 if (ext4_proc_root) 3854 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); 3855 3856 if (sbi->s_proc) 3857 proc_create_data("options", S_IRUGO, sbi->s_proc, 3858 &ext4_seq_options_fops, sb); 3859 3860 bgl_lock_init(sbi->s_blockgroup_lock); 3861 3862 for (i = 0; i < db_count; i++) { 3863 block = descriptor_loc(sb, logical_sb_block, i); 3864 sbi->s_group_desc[i] = sb_bread(sb, block); 3865 if (!sbi->s_group_desc[i]) { 3866 ext4_msg(sb, KERN_ERR, 3867 "can't read group descriptor %d", i); 3868 db_count = i; 3869 goto failed_mount2; 3870 } 3871 } 3872 3873 /* 3874 * set up enough so that it can read an inode, 3875 * and create new inode for buddy allocator 3876 */ 3877 sbi->s_gdb_count = db_count; 3878 if (!test_opt(sb, NOLOAD) && 3879 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 3880 sb->s_op = &ext4_sops; 3881 else 3882 sb->s_op = &ext4_nojournal_sops; 3883 3884 ext4_ext_init(sb); 3885 err = ext4_mb_init(sb); 3886 if (err) { 3887 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 3888 err); 3889 goto failed_mount2; 3890 } 3891 3892 if (!ext4_check_descriptors(sb, &first_not_zeroed)) { 3893 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3894 goto failed_mount2a; 3895 } 3896 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 3897 if (!ext4_fill_flex_info(sb)) { 3898 ext4_msg(sb, KERN_ERR, 3899 "unable to initialize " 3900 "flex_bg meta info!"); 3901 goto failed_mount2a; 3902 } 3903 3904 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3905 spin_lock_init(&sbi->s_next_gen_lock); 3906 3907 init_timer(&sbi->s_err_report); 3908 sbi->s_err_report.function = print_daily_error_info; 3909 sbi->s_err_report.data = (unsigned long) sb; 3910 3911 /* Register extent status tree shrinker */ 3912 ext4_es_register_shrinker(sbi); 3913 3914 err = percpu_counter_init(&sbi->s_freeclusters_counter, 3915 ext4_count_free_clusters(sb)); 3916 if (!err) { 3917 err = percpu_counter_init(&sbi->s_freeinodes_counter, 3918 ext4_count_free_inodes(sb)); 3919 } 3920 if (!err) { 3921 err = percpu_counter_init(&sbi->s_dirs_counter, 3922 ext4_count_dirs(sb)); 3923 } 3924 if (!err) { 3925 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0); 3926 } 3927 if (!err) { 3928 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0); 3929 } 3930 if (err) { 3931 ext4_msg(sb, KERN_ERR, "insufficient memory"); 3932 goto failed_mount3; 3933 } 3934 3935 sbi->s_stripe = ext4_get_stripe_size(sbi); 3936 sbi->s_extent_max_zeroout_kb = 32; 3937 3938 sb->s_export_op = &ext4_export_ops; 3939 sb->s_xattr = ext4_xattr_handlers; 3940 #ifdef CONFIG_QUOTA 3941 sb->dq_op = &ext4_quota_operations; 3942 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 3943 sb->s_qcop = &ext4_qctl_sysfile_operations; 3944 else 3945 sb->s_qcop = &ext4_qctl_operations; 3946 #endif 3947 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3948 3949 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 3950 mutex_init(&sbi->s_orphan_lock); 3951 3952 sb->s_root = NULL; 3953 3954 needs_recovery = (es->s_last_orphan != 0 || 3955 EXT4_HAS_INCOMPAT_FEATURE(sb, 3956 EXT4_FEATURE_INCOMPAT_RECOVER)); 3957 3958 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) && 3959 !(sb->s_flags & MS_RDONLY)) 3960 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 3961 goto failed_mount3; 3962 3963 /* 3964 * The first inode we look at is the journal inode. Don't try 3965 * root first: it may be modified in the journal! 3966 */ 3967 if (!test_opt(sb, NOLOAD) && 3968 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 3969 if (ext4_load_journal(sb, es, journal_devnum)) 3970 goto failed_mount3; 3971 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 3972 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 3973 ext4_msg(sb, KERN_ERR, "required journal recovery " 3974 "suppressed and not mounted read-only"); 3975 goto failed_mount_wq; 3976 } else { 3977 clear_opt(sb, DATA_FLAGS); 3978 sbi->s_journal = NULL; 3979 needs_recovery = 0; 3980 goto no_journal; 3981 } 3982 3983 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) && 3984 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3985 JBD2_FEATURE_INCOMPAT_64BIT)) { 3986 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3987 goto failed_mount_wq; 3988 } 3989 3990 if (!set_journal_csum_feature_set(sb)) { 3991 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 3992 "feature set"); 3993 goto failed_mount_wq; 3994 } 3995 3996 /* We have now updated the journal if required, so we can 3997 * validate the data journaling mode. */ 3998 switch (test_opt(sb, DATA_FLAGS)) { 3999 case 0: 4000 /* No mode set, assume a default based on the journal 4001 * capabilities: ORDERED_DATA if the journal can 4002 * cope, else JOURNAL_DATA 4003 */ 4004 if (jbd2_journal_check_available_features 4005 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 4006 set_opt(sb, ORDERED_DATA); 4007 else 4008 set_opt(sb, JOURNAL_DATA); 4009 break; 4010 4011 case EXT4_MOUNT_ORDERED_DATA: 4012 case EXT4_MOUNT_WRITEBACK_DATA: 4013 if (!jbd2_journal_check_available_features 4014 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4015 ext4_msg(sb, KERN_ERR, "Journal does not support " 4016 "requested data journaling mode"); 4017 goto failed_mount_wq; 4018 } 4019 default: 4020 break; 4021 } 4022 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4023 4024 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4025 4026 /* 4027 * The journal may have updated the bg summary counts, so we 4028 * need to update the global counters. 4029 */ 4030 percpu_counter_set(&sbi->s_freeclusters_counter, 4031 ext4_count_free_clusters(sb)); 4032 percpu_counter_set(&sbi->s_freeinodes_counter, 4033 ext4_count_free_inodes(sb)); 4034 percpu_counter_set(&sbi->s_dirs_counter, 4035 ext4_count_dirs(sb)); 4036 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0); 4037 4038 no_journal: 4039 if (ext4_mballoc_ready) { 4040 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id); 4041 if (!sbi->s_mb_cache) { 4042 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache"); 4043 goto failed_mount_wq; 4044 } 4045 } 4046 4047 /* 4048 * Get the # of file system overhead blocks from the 4049 * superblock if present. 4050 */ 4051 if (es->s_overhead_clusters) 4052 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4053 else { 4054 err = ext4_calculate_overhead(sb); 4055 if (err) 4056 goto failed_mount_wq; 4057 } 4058 4059 /* 4060 * The maximum number of concurrent works can be high and 4061 * concurrency isn't really necessary. Limit it to 1. 4062 */ 4063 EXT4_SB(sb)->rsv_conversion_wq = 4064 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4065 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4066 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4067 ret = -ENOMEM; 4068 goto failed_mount4; 4069 } 4070 4071 /* 4072 * The jbd2_journal_load will have done any necessary log recovery, 4073 * so we can safely mount the rest of the filesystem now. 4074 */ 4075 4076 root = ext4_iget(sb, EXT4_ROOT_INO); 4077 if (IS_ERR(root)) { 4078 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4079 ret = PTR_ERR(root); 4080 root = NULL; 4081 goto failed_mount4; 4082 } 4083 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4084 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4085 iput(root); 4086 goto failed_mount4; 4087 } 4088 sb->s_root = d_make_root(root); 4089 if (!sb->s_root) { 4090 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4091 ret = -ENOMEM; 4092 goto failed_mount4; 4093 } 4094 4095 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY)) 4096 sb->s_flags |= MS_RDONLY; 4097 4098 /* determine the minimum size of new large inodes, if present */ 4099 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4100 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4101 EXT4_GOOD_OLD_INODE_SIZE; 4102 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 4103 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { 4104 if (sbi->s_want_extra_isize < 4105 le16_to_cpu(es->s_want_extra_isize)) 4106 sbi->s_want_extra_isize = 4107 le16_to_cpu(es->s_want_extra_isize); 4108 if (sbi->s_want_extra_isize < 4109 le16_to_cpu(es->s_min_extra_isize)) 4110 sbi->s_want_extra_isize = 4111 le16_to_cpu(es->s_min_extra_isize); 4112 } 4113 } 4114 /* Check if enough inode space is available */ 4115 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 4116 sbi->s_inode_size) { 4117 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4118 EXT4_GOOD_OLD_INODE_SIZE; 4119 ext4_msg(sb, KERN_INFO, "required extra inode space not" 4120 "available"); 4121 } 4122 4123 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb)); 4124 if (err) { 4125 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for " 4126 "reserved pool", ext4_calculate_resv_clusters(sb)); 4127 goto failed_mount5; 4128 } 4129 4130 err = ext4_setup_system_zone(sb); 4131 if (err) { 4132 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4133 "zone (%d)", err); 4134 goto failed_mount5; 4135 } 4136 4137 err = ext4_register_li_request(sb, first_not_zeroed); 4138 if (err) 4139 goto failed_mount6; 4140 4141 sbi->s_kobj.kset = ext4_kset; 4142 init_completion(&sbi->s_kobj_unregister); 4143 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 4144 "%s", sb->s_id); 4145 if (err) 4146 goto failed_mount7; 4147 4148 #ifdef CONFIG_QUOTA 4149 /* Enable quota usage during mount. */ 4150 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 4151 !(sb->s_flags & MS_RDONLY)) { 4152 err = ext4_enable_quotas(sb); 4153 if (err) 4154 goto failed_mount8; 4155 } 4156 #endif /* CONFIG_QUOTA */ 4157 4158 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4159 ext4_orphan_cleanup(sb, es); 4160 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4161 if (needs_recovery) { 4162 ext4_msg(sb, KERN_INFO, "recovery complete"); 4163 ext4_mark_recovery_complete(sb, es); 4164 } 4165 if (EXT4_SB(sb)->s_journal) { 4166 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4167 descr = " journalled data mode"; 4168 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4169 descr = " ordered data mode"; 4170 else 4171 descr = " writeback data mode"; 4172 } else 4173 descr = "out journal"; 4174 4175 if (test_opt(sb, DISCARD)) { 4176 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4177 if (!blk_queue_discard(q)) 4178 ext4_msg(sb, KERN_WARNING, 4179 "mounting with \"discard\" option, but " 4180 "the device does not support discard"); 4181 } 4182 4183 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4184 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 4185 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4186 4187 if (es->s_error_count) 4188 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4189 4190 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4191 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4192 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4193 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4194 4195 kfree(orig_data); 4196 return 0; 4197 4198 cantfind_ext4: 4199 if (!silent) 4200 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4201 goto failed_mount; 4202 4203 #ifdef CONFIG_QUOTA 4204 failed_mount8: 4205 kobject_del(&sbi->s_kobj); 4206 #endif 4207 failed_mount7: 4208 ext4_unregister_li_request(sb); 4209 failed_mount6: 4210 ext4_release_system_zone(sb); 4211 failed_mount5: 4212 dput(sb->s_root); 4213 sb->s_root = NULL; 4214 failed_mount4: 4215 ext4_msg(sb, KERN_ERR, "mount failed"); 4216 if (EXT4_SB(sb)->rsv_conversion_wq) 4217 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4218 failed_mount_wq: 4219 if (sbi->s_journal) { 4220 jbd2_journal_destroy(sbi->s_journal); 4221 sbi->s_journal = NULL; 4222 } 4223 failed_mount3: 4224 ext4_es_unregister_shrinker(sbi); 4225 del_timer_sync(&sbi->s_err_report); 4226 if (sbi->s_flex_groups) 4227 ext4_kvfree(sbi->s_flex_groups); 4228 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4229 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4230 percpu_counter_destroy(&sbi->s_dirs_counter); 4231 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4232 percpu_counter_destroy(&sbi->s_extent_cache_cnt); 4233 if (sbi->s_mmp_tsk) 4234 kthread_stop(sbi->s_mmp_tsk); 4235 failed_mount2a: 4236 ext4_mb_release(sb); 4237 failed_mount2: 4238 for (i = 0; i < db_count; i++) 4239 brelse(sbi->s_group_desc[i]); 4240 ext4_kvfree(sbi->s_group_desc); 4241 failed_mount: 4242 ext4_ext_release(sb); 4243 if (sbi->s_chksum_driver) 4244 crypto_free_shash(sbi->s_chksum_driver); 4245 if (sbi->s_proc) { 4246 remove_proc_entry("options", sbi->s_proc); 4247 remove_proc_entry(sb->s_id, ext4_proc_root); 4248 } 4249 #ifdef CONFIG_QUOTA 4250 for (i = 0; i < MAXQUOTAS; i++) 4251 kfree(sbi->s_qf_names[i]); 4252 #endif 4253 ext4_blkdev_remove(sbi); 4254 brelse(bh); 4255 out_fail: 4256 sb->s_fs_info = NULL; 4257 kfree(sbi->s_blockgroup_lock); 4258 kfree(sbi); 4259 out_free_orig: 4260 kfree(orig_data); 4261 return err ? err : ret; 4262 } 4263 4264 /* 4265 * Setup any per-fs journal parameters now. We'll do this both on 4266 * initial mount, once the journal has been initialised but before we've 4267 * done any recovery; and again on any subsequent remount. 4268 */ 4269 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4270 { 4271 struct ext4_sb_info *sbi = EXT4_SB(sb); 4272 4273 journal->j_commit_interval = sbi->s_commit_interval; 4274 journal->j_min_batch_time = sbi->s_min_batch_time; 4275 journal->j_max_batch_time = sbi->s_max_batch_time; 4276 4277 write_lock(&journal->j_state_lock); 4278 if (test_opt(sb, BARRIER)) 4279 journal->j_flags |= JBD2_BARRIER; 4280 else 4281 journal->j_flags &= ~JBD2_BARRIER; 4282 if (test_opt(sb, DATA_ERR_ABORT)) 4283 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4284 else 4285 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4286 write_unlock(&journal->j_state_lock); 4287 } 4288 4289 static journal_t *ext4_get_journal(struct super_block *sb, 4290 unsigned int journal_inum) 4291 { 4292 struct inode *journal_inode; 4293 journal_t *journal; 4294 4295 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4296 4297 /* First, test for the existence of a valid inode on disk. Bad 4298 * things happen if we iget() an unused inode, as the subsequent 4299 * iput() will try to delete it. */ 4300 4301 journal_inode = ext4_iget(sb, journal_inum); 4302 if (IS_ERR(journal_inode)) { 4303 ext4_msg(sb, KERN_ERR, "no journal found"); 4304 return NULL; 4305 } 4306 if (!journal_inode->i_nlink) { 4307 make_bad_inode(journal_inode); 4308 iput(journal_inode); 4309 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4310 return NULL; 4311 } 4312 4313 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4314 journal_inode, journal_inode->i_size); 4315 if (!S_ISREG(journal_inode->i_mode)) { 4316 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4317 iput(journal_inode); 4318 return NULL; 4319 } 4320 4321 journal = jbd2_journal_init_inode(journal_inode); 4322 if (!journal) { 4323 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4324 iput(journal_inode); 4325 return NULL; 4326 } 4327 journal->j_private = sb; 4328 ext4_init_journal_params(sb, journal); 4329 return journal; 4330 } 4331 4332 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4333 dev_t j_dev) 4334 { 4335 struct buffer_head *bh; 4336 journal_t *journal; 4337 ext4_fsblk_t start; 4338 ext4_fsblk_t len; 4339 int hblock, blocksize; 4340 ext4_fsblk_t sb_block; 4341 unsigned long offset; 4342 struct ext4_super_block *es; 4343 struct block_device *bdev; 4344 4345 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4346 4347 bdev = ext4_blkdev_get(j_dev, sb); 4348 if (bdev == NULL) 4349 return NULL; 4350 4351 blocksize = sb->s_blocksize; 4352 hblock = bdev_logical_block_size(bdev); 4353 if (blocksize < hblock) { 4354 ext4_msg(sb, KERN_ERR, 4355 "blocksize too small for journal device"); 4356 goto out_bdev; 4357 } 4358 4359 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4360 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4361 set_blocksize(bdev, blocksize); 4362 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4363 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4364 "external journal"); 4365 goto out_bdev; 4366 } 4367 4368 es = (struct ext4_super_block *) (bh->b_data + offset); 4369 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4370 !(le32_to_cpu(es->s_feature_incompat) & 4371 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4372 ext4_msg(sb, KERN_ERR, "external journal has " 4373 "bad superblock"); 4374 brelse(bh); 4375 goto out_bdev; 4376 } 4377 4378 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4379 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4380 brelse(bh); 4381 goto out_bdev; 4382 } 4383 4384 len = ext4_blocks_count(es); 4385 start = sb_block + 1; 4386 brelse(bh); /* we're done with the superblock */ 4387 4388 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4389 start, len, blocksize); 4390 if (!journal) { 4391 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4392 goto out_bdev; 4393 } 4394 journal->j_private = sb; 4395 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4396 wait_on_buffer(journal->j_sb_buffer); 4397 if (!buffer_uptodate(journal->j_sb_buffer)) { 4398 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4399 goto out_journal; 4400 } 4401 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4402 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4403 "user (unsupported) - %d", 4404 be32_to_cpu(journal->j_superblock->s_nr_users)); 4405 goto out_journal; 4406 } 4407 EXT4_SB(sb)->journal_bdev = bdev; 4408 ext4_init_journal_params(sb, journal); 4409 return journal; 4410 4411 out_journal: 4412 jbd2_journal_destroy(journal); 4413 out_bdev: 4414 ext4_blkdev_put(bdev); 4415 return NULL; 4416 } 4417 4418 static int ext4_load_journal(struct super_block *sb, 4419 struct ext4_super_block *es, 4420 unsigned long journal_devnum) 4421 { 4422 journal_t *journal; 4423 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4424 dev_t journal_dev; 4425 int err = 0; 4426 int really_read_only; 4427 4428 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4429 4430 if (journal_devnum && 4431 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4432 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4433 "numbers have changed"); 4434 journal_dev = new_decode_dev(journal_devnum); 4435 } else 4436 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4437 4438 really_read_only = bdev_read_only(sb->s_bdev); 4439 4440 /* 4441 * Are we loading a blank journal or performing recovery after a 4442 * crash? For recovery, we need to check in advance whether we 4443 * can get read-write access to the device. 4444 */ 4445 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 4446 if (sb->s_flags & MS_RDONLY) { 4447 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4448 "required on readonly filesystem"); 4449 if (really_read_only) { 4450 ext4_msg(sb, KERN_ERR, "write access " 4451 "unavailable, cannot proceed"); 4452 return -EROFS; 4453 } 4454 ext4_msg(sb, KERN_INFO, "write access will " 4455 "be enabled during recovery"); 4456 } 4457 } 4458 4459 if (journal_inum && journal_dev) { 4460 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4461 "and inode journals!"); 4462 return -EINVAL; 4463 } 4464 4465 if (journal_inum) { 4466 if (!(journal = ext4_get_journal(sb, journal_inum))) 4467 return -EINVAL; 4468 } else { 4469 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4470 return -EINVAL; 4471 } 4472 4473 if (!(journal->j_flags & JBD2_BARRIER)) 4474 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4475 4476 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) 4477 err = jbd2_journal_wipe(journal, !really_read_only); 4478 if (!err) { 4479 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4480 if (save) 4481 memcpy(save, ((char *) es) + 4482 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4483 err = jbd2_journal_load(journal); 4484 if (save) 4485 memcpy(((char *) es) + EXT4_S_ERR_START, 4486 save, EXT4_S_ERR_LEN); 4487 kfree(save); 4488 } 4489 4490 if (err) { 4491 ext4_msg(sb, KERN_ERR, "error loading journal"); 4492 jbd2_journal_destroy(journal); 4493 return err; 4494 } 4495 4496 EXT4_SB(sb)->s_journal = journal; 4497 ext4_clear_journal_err(sb, es); 4498 4499 if (!really_read_only && journal_devnum && 4500 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4501 es->s_journal_dev = cpu_to_le32(journal_devnum); 4502 4503 /* Make sure we flush the recovery flag to disk. */ 4504 ext4_commit_super(sb, 1); 4505 } 4506 4507 return 0; 4508 } 4509 4510 static int ext4_commit_super(struct super_block *sb, int sync) 4511 { 4512 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4513 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4514 int error = 0; 4515 4516 if (!sbh || block_device_ejected(sb)) 4517 return error; 4518 if (buffer_write_io_error(sbh)) { 4519 /* 4520 * Oh, dear. A previous attempt to write the 4521 * superblock failed. This could happen because the 4522 * USB device was yanked out. Or it could happen to 4523 * be a transient write error and maybe the block will 4524 * be remapped. Nothing we can do but to retry the 4525 * write and hope for the best. 4526 */ 4527 ext4_msg(sb, KERN_ERR, "previous I/O error to " 4528 "superblock detected"); 4529 clear_buffer_write_io_error(sbh); 4530 set_buffer_uptodate(sbh); 4531 } 4532 /* 4533 * If the file system is mounted read-only, don't update the 4534 * superblock write time. This avoids updating the superblock 4535 * write time when we are mounting the root file system 4536 * read/only but we need to replay the journal; at that point, 4537 * for people who are east of GMT and who make their clock 4538 * tick in localtime for Windows bug-for-bug compatibility, 4539 * the clock is set in the future, and this will cause e2fsck 4540 * to complain and force a full file system check. 4541 */ 4542 if (!(sb->s_flags & MS_RDONLY)) 4543 es->s_wtime = cpu_to_le32(get_seconds()); 4544 if (sb->s_bdev->bd_part) 4545 es->s_kbytes_written = 4546 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4547 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4548 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4549 else 4550 es->s_kbytes_written = 4551 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4552 ext4_free_blocks_count_set(es, 4553 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4554 &EXT4_SB(sb)->s_freeclusters_counter))); 4555 es->s_free_inodes_count = 4556 cpu_to_le32(percpu_counter_sum_positive( 4557 &EXT4_SB(sb)->s_freeinodes_counter)); 4558 BUFFER_TRACE(sbh, "marking dirty"); 4559 ext4_superblock_csum_set(sb); 4560 mark_buffer_dirty(sbh); 4561 if (sync) { 4562 error = sync_dirty_buffer(sbh); 4563 if (error) 4564 return error; 4565 4566 error = buffer_write_io_error(sbh); 4567 if (error) { 4568 ext4_msg(sb, KERN_ERR, "I/O error while writing " 4569 "superblock"); 4570 clear_buffer_write_io_error(sbh); 4571 set_buffer_uptodate(sbh); 4572 } 4573 } 4574 return error; 4575 } 4576 4577 /* 4578 * Have we just finished recovery? If so, and if we are mounting (or 4579 * remounting) the filesystem readonly, then we will end up with a 4580 * consistent fs on disk. Record that fact. 4581 */ 4582 static void ext4_mark_recovery_complete(struct super_block *sb, 4583 struct ext4_super_block *es) 4584 { 4585 journal_t *journal = EXT4_SB(sb)->s_journal; 4586 4587 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 4588 BUG_ON(journal != NULL); 4589 return; 4590 } 4591 jbd2_journal_lock_updates(journal); 4592 if (jbd2_journal_flush(journal) < 0) 4593 goto out; 4594 4595 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && 4596 sb->s_flags & MS_RDONLY) { 4597 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4598 ext4_commit_super(sb, 1); 4599 } 4600 4601 out: 4602 jbd2_journal_unlock_updates(journal); 4603 } 4604 4605 /* 4606 * If we are mounting (or read-write remounting) a filesystem whose journal 4607 * has recorded an error from a previous lifetime, move that error to the 4608 * main filesystem now. 4609 */ 4610 static void ext4_clear_journal_err(struct super_block *sb, 4611 struct ext4_super_block *es) 4612 { 4613 journal_t *journal; 4614 int j_errno; 4615 const char *errstr; 4616 4617 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4618 4619 journal = EXT4_SB(sb)->s_journal; 4620 4621 /* 4622 * Now check for any error status which may have been recorded in the 4623 * journal by a prior ext4_error() or ext4_abort() 4624 */ 4625 4626 j_errno = jbd2_journal_errno(journal); 4627 if (j_errno) { 4628 char nbuf[16]; 4629 4630 errstr = ext4_decode_error(sb, j_errno, nbuf); 4631 ext4_warning(sb, "Filesystem error recorded " 4632 "from previous mount: %s", errstr); 4633 ext4_warning(sb, "Marking fs in need of filesystem check."); 4634 4635 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4636 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4637 ext4_commit_super(sb, 1); 4638 4639 jbd2_journal_clear_err(journal); 4640 jbd2_journal_update_sb_errno(journal); 4641 } 4642 } 4643 4644 /* 4645 * Force the running and committing transactions to commit, 4646 * and wait on the commit. 4647 */ 4648 int ext4_force_commit(struct super_block *sb) 4649 { 4650 journal_t *journal; 4651 4652 if (sb->s_flags & MS_RDONLY) 4653 return 0; 4654 4655 journal = EXT4_SB(sb)->s_journal; 4656 return ext4_journal_force_commit(journal); 4657 } 4658 4659 static int ext4_sync_fs(struct super_block *sb, int wait) 4660 { 4661 int ret = 0; 4662 tid_t target; 4663 bool needs_barrier = false; 4664 struct ext4_sb_info *sbi = EXT4_SB(sb); 4665 4666 trace_ext4_sync_fs(sb, wait); 4667 flush_workqueue(sbi->rsv_conversion_wq); 4668 /* 4669 * Writeback quota in non-journalled quota case - journalled quota has 4670 * no dirty dquots 4671 */ 4672 dquot_writeback_dquots(sb, -1); 4673 /* 4674 * Data writeback is possible w/o journal transaction, so barrier must 4675 * being sent at the end of the function. But we can skip it if 4676 * transaction_commit will do it for us. 4677 */ 4678 target = jbd2_get_latest_transaction(sbi->s_journal); 4679 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 4680 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 4681 needs_barrier = true; 4682 4683 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4684 if (wait) 4685 ret = jbd2_log_wait_commit(sbi->s_journal, target); 4686 } 4687 if (needs_barrier) { 4688 int err; 4689 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 4690 if (!ret) 4691 ret = err; 4692 } 4693 4694 return ret; 4695 } 4696 4697 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait) 4698 { 4699 int ret = 0; 4700 4701 trace_ext4_sync_fs(sb, wait); 4702 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4703 dquot_writeback_dquots(sb, -1); 4704 if (wait && test_opt(sb, BARRIER)) 4705 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 4706 4707 return ret; 4708 } 4709 4710 /* 4711 * LVM calls this function before a (read-only) snapshot is created. This 4712 * gives us a chance to flush the journal completely and mark the fs clean. 4713 * 4714 * Note that only this function cannot bring a filesystem to be in a clean 4715 * state independently. It relies on upper layer to stop all data & metadata 4716 * modifications. 4717 */ 4718 static int ext4_freeze(struct super_block *sb) 4719 { 4720 int error = 0; 4721 journal_t *journal; 4722 4723 if (sb->s_flags & MS_RDONLY) 4724 return 0; 4725 4726 journal = EXT4_SB(sb)->s_journal; 4727 4728 /* Now we set up the journal barrier. */ 4729 jbd2_journal_lock_updates(journal); 4730 4731 /* 4732 * Don't clear the needs_recovery flag if we failed to flush 4733 * the journal. 4734 */ 4735 error = jbd2_journal_flush(journal); 4736 if (error < 0) 4737 goto out; 4738 4739 /* Journal blocked and flushed, clear needs_recovery flag. */ 4740 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4741 error = ext4_commit_super(sb, 1); 4742 out: 4743 /* we rely on upper layer to stop further updates */ 4744 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 4745 return error; 4746 } 4747 4748 /* 4749 * Called by LVM after the snapshot is done. We need to reset the RECOVER 4750 * flag here, even though the filesystem is not technically dirty yet. 4751 */ 4752 static int ext4_unfreeze(struct super_block *sb) 4753 { 4754 if (sb->s_flags & MS_RDONLY) 4755 return 0; 4756 4757 /* Reset the needs_recovery flag before the fs is unlocked. */ 4758 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4759 ext4_commit_super(sb, 1); 4760 return 0; 4761 } 4762 4763 /* 4764 * Structure to save mount options for ext4_remount's benefit 4765 */ 4766 struct ext4_mount_options { 4767 unsigned long s_mount_opt; 4768 unsigned long s_mount_opt2; 4769 kuid_t s_resuid; 4770 kgid_t s_resgid; 4771 unsigned long s_commit_interval; 4772 u32 s_min_batch_time, s_max_batch_time; 4773 #ifdef CONFIG_QUOTA 4774 int s_jquota_fmt; 4775 char *s_qf_names[MAXQUOTAS]; 4776 #endif 4777 }; 4778 4779 static int ext4_remount(struct super_block *sb, int *flags, char *data) 4780 { 4781 struct ext4_super_block *es; 4782 struct ext4_sb_info *sbi = EXT4_SB(sb); 4783 unsigned long old_sb_flags; 4784 struct ext4_mount_options old_opts; 4785 int enable_quota = 0; 4786 ext4_group_t g; 4787 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4788 int err = 0; 4789 #ifdef CONFIG_QUOTA 4790 int i, j; 4791 #endif 4792 char *orig_data = kstrdup(data, GFP_KERNEL); 4793 4794 /* Store the original options */ 4795 old_sb_flags = sb->s_flags; 4796 old_opts.s_mount_opt = sbi->s_mount_opt; 4797 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4798 old_opts.s_resuid = sbi->s_resuid; 4799 old_opts.s_resgid = sbi->s_resgid; 4800 old_opts.s_commit_interval = sbi->s_commit_interval; 4801 old_opts.s_min_batch_time = sbi->s_min_batch_time; 4802 old_opts.s_max_batch_time = sbi->s_max_batch_time; 4803 #ifdef CONFIG_QUOTA 4804 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4805 for (i = 0; i < MAXQUOTAS; i++) 4806 if (sbi->s_qf_names[i]) { 4807 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], 4808 GFP_KERNEL); 4809 if (!old_opts.s_qf_names[i]) { 4810 for (j = 0; j < i; j++) 4811 kfree(old_opts.s_qf_names[j]); 4812 kfree(orig_data); 4813 return -ENOMEM; 4814 } 4815 } else 4816 old_opts.s_qf_names[i] = NULL; 4817 #endif 4818 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4819 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4820 4821 /* 4822 * Allow the "check" option to be passed as a remount option. 4823 */ 4824 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 4825 err = -EINVAL; 4826 goto restore_opts; 4827 } 4828 4829 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4830 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4831 ext4_msg(sb, KERN_ERR, "can't mount with " 4832 "both data=journal and delalloc"); 4833 err = -EINVAL; 4834 goto restore_opts; 4835 } 4836 if (test_opt(sb, DIOREAD_NOLOCK)) { 4837 ext4_msg(sb, KERN_ERR, "can't mount with " 4838 "both data=journal and dioread_nolock"); 4839 err = -EINVAL; 4840 goto restore_opts; 4841 } 4842 } 4843 4844 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4845 ext4_abort(sb, "Abort forced by user"); 4846 4847 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4848 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4849 4850 es = sbi->s_es; 4851 4852 if (sbi->s_journal) { 4853 ext4_init_journal_params(sb, sbi->s_journal); 4854 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4855 } 4856 4857 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 4858 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4859 err = -EROFS; 4860 goto restore_opts; 4861 } 4862 4863 if (*flags & MS_RDONLY) { 4864 err = sync_filesystem(sb); 4865 if (err < 0) 4866 goto restore_opts; 4867 err = dquot_suspend(sb, -1); 4868 if (err < 0) 4869 goto restore_opts; 4870 4871 /* 4872 * First of all, the unconditional stuff we have to do 4873 * to disable replay of the journal when we next remount 4874 */ 4875 sb->s_flags |= MS_RDONLY; 4876 4877 /* 4878 * OK, test if we are remounting a valid rw partition 4879 * readonly, and if so set the rdonly flag and then 4880 * mark the partition as valid again. 4881 */ 4882 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4883 (sbi->s_mount_state & EXT4_VALID_FS)) 4884 es->s_state = cpu_to_le16(sbi->s_mount_state); 4885 4886 if (sbi->s_journal) 4887 ext4_mark_recovery_complete(sb, es); 4888 } else { 4889 /* Make sure we can mount this feature set readwrite */ 4890 if (!ext4_feature_set_ok(sb, 0)) { 4891 err = -EROFS; 4892 goto restore_opts; 4893 } 4894 /* 4895 * Make sure the group descriptor checksums 4896 * are sane. If they aren't, refuse to remount r/w. 4897 */ 4898 for (g = 0; g < sbi->s_groups_count; g++) { 4899 struct ext4_group_desc *gdp = 4900 ext4_get_group_desc(sb, g, NULL); 4901 4902 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 4903 ext4_msg(sb, KERN_ERR, 4904 "ext4_remount: Checksum for group %u failed (%u!=%u)", 4905 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), 4906 le16_to_cpu(gdp->bg_checksum)); 4907 err = -EINVAL; 4908 goto restore_opts; 4909 } 4910 } 4911 4912 /* 4913 * If we have an unprocessed orphan list hanging 4914 * around from a previously readonly bdev mount, 4915 * require a full umount/remount for now. 4916 */ 4917 if (es->s_last_orphan) { 4918 ext4_msg(sb, KERN_WARNING, "Couldn't " 4919 "remount RDWR because of unprocessed " 4920 "orphan inode list. Please " 4921 "umount/remount instead"); 4922 err = -EINVAL; 4923 goto restore_opts; 4924 } 4925 4926 /* 4927 * Mounting a RDONLY partition read-write, so reread 4928 * and store the current valid flag. (It may have 4929 * been changed by e2fsck since we originally mounted 4930 * the partition.) 4931 */ 4932 if (sbi->s_journal) 4933 ext4_clear_journal_err(sb, es); 4934 sbi->s_mount_state = le16_to_cpu(es->s_state); 4935 if (!ext4_setup_super(sb, es, 0)) 4936 sb->s_flags &= ~MS_RDONLY; 4937 if (EXT4_HAS_INCOMPAT_FEATURE(sb, 4938 EXT4_FEATURE_INCOMPAT_MMP)) 4939 if (ext4_multi_mount_protect(sb, 4940 le64_to_cpu(es->s_mmp_block))) { 4941 err = -EROFS; 4942 goto restore_opts; 4943 } 4944 enable_quota = 1; 4945 } 4946 } 4947 4948 /* 4949 * Reinitialize lazy itable initialization thread based on 4950 * current settings 4951 */ 4952 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4953 ext4_unregister_li_request(sb); 4954 else { 4955 ext4_group_t first_not_zeroed; 4956 first_not_zeroed = ext4_has_uninit_itable(sb); 4957 ext4_register_li_request(sb, first_not_zeroed); 4958 } 4959 4960 ext4_setup_system_zone(sb); 4961 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY)) 4962 ext4_commit_super(sb, 1); 4963 4964 #ifdef CONFIG_QUOTA 4965 /* Release old quota file names */ 4966 for (i = 0; i < MAXQUOTAS; i++) 4967 kfree(old_opts.s_qf_names[i]); 4968 if (enable_quota) { 4969 if (sb_any_quota_suspended(sb)) 4970 dquot_resume(sb, -1); 4971 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 4972 EXT4_FEATURE_RO_COMPAT_QUOTA)) { 4973 err = ext4_enable_quotas(sb); 4974 if (err) 4975 goto restore_opts; 4976 } 4977 } 4978 #endif 4979 4980 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4981 kfree(orig_data); 4982 return 0; 4983 4984 restore_opts: 4985 sb->s_flags = old_sb_flags; 4986 sbi->s_mount_opt = old_opts.s_mount_opt; 4987 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 4988 sbi->s_resuid = old_opts.s_resuid; 4989 sbi->s_resgid = old_opts.s_resgid; 4990 sbi->s_commit_interval = old_opts.s_commit_interval; 4991 sbi->s_min_batch_time = old_opts.s_min_batch_time; 4992 sbi->s_max_batch_time = old_opts.s_max_batch_time; 4993 #ifdef CONFIG_QUOTA 4994 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 4995 for (i = 0; i < MAXQUOTAS; i++) { 4996 kfree(sbi->s_qf_names[i]); 4997 sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 4998 } 4999 #endif 5000 kfree(orig_data); 5001 return err; 5002 } 5003 5004 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5005 { 5006 struct super_block *sb = dentry->d_sb; 5007 struct ext4_sb_info *sbi = EXT4_SB(sb); 5008 struct ext4_super_block *es = sbi->s_es; 5009 ext4_fsblk_t overhead = 0, resv_blocks; 5010 u64 fsid; 5011 s64 bfree; 5012 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5013 5014 if (!test_opt(sb, MINIX_DF)) 5015 overhead = sbi->s_overhead; 5016 5017 buf->f_type = EXT4_SUPER_MAGIC; 5018 buf->f_bsize = sb->s_blocksize; 5019 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5020 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5021 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5022 /* prevent underflow in case that few free space is available */ 5023 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5024 buf->f_bavail = buf->f_bfree - 5025 (ext4_r_blocks_count(es) + resv_blocks); 5026 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5027 buf->f_bavail = 0; 5028 buf->f_files = le32_to_cpu(es->s_inodes_count); 5029 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5030 buf->f_namelen = EXT4_NAME_LEN; 5031 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5032 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5033 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5034 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5035 5036 return 0; 5037 } 5038 5039 /* Helper function for writing quotas on sync - we need to start transaction 5040 * before quota file is locked for write. Otherwise the are possible deadlocks: 5041 * Process 1 Process 2 5042 * ext4_create() quota_sync() 5043 * jbd2_journal_start() write_dquot() 5044 * dquot_initialize() down(dqio_mutex) 5045 * down(dqio_mutex) jbd2_journal_start() 5046 * 5047 */ 5048 5049 #ifdef CONFIG_QUOTA 5050 5051 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5052 { 5053 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5054 } 5055 5056 static int ext4_write_dquot(struct dquot *dquot) 5057 { 5058 int ret, err; 5059 handle_t *handle; 5060 struct inode *inode; 5061 5062 inode = dquot_to_inode(dquot); 5063 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5064 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5065 if (IS_ERR(handle)) 5066 return PTR_ERR(handle); 5067 ret = dquot_commit(dquot); 5068 err = ext4_journal_stop(handle); 5069 if (!ret) 5070 ret = err; 5071 return ret; 5072 } 5073 5074 static int ext4_acquire_dquot(struct dquot *dquot) 5075 { 5076 int ret, err; 5077 handle_t *handle; 5078 5079 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5080 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5081 if (IS_ERR(handle)) 5082 return PTR_ERR(handle); 5083 ret = dquot_acquire(dquot); 5084 err = ext4_journal_stop(handle); 5085 if (!ret) 5086 ret = err; 5087 return ret; 5088 } 5089 5090 static int ext4_release_dquot(struct dquot *dquot) 5091 { 5092 int ret, err; 5093 handle_t *handle; 5094 5095 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5096 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5097 if (IS_ERR(handle)) { 5098 /* Release dquot anyway to avoid endless cycle in dqput() */ 5099 dquot_release(dquot); 5100 return PTR_ERR(handle); 5101 } 5102 ret = dquot_release(dquot); 5103 err = ext4_journal_stop(handle); 5104 if (!ret) 5105 ret = err; 5106 return ret; 5107 } 5108 5109 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5110 { 5111 struct super_block *sb = dquot->dq_sb; 5112 struct ext4_sb_info *sbi = EXT4_SB(sb); 5113 5114 /* Are we journaling quotas? */ 5115 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) || 5116 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5117 dquot_mark_dquot_dirty(dquot); 5118 return ext4_write_dquot(dquot); 5119 } else { 5120 return dquot_mark_dquot_dirty(dquot); 5121 } 5122 } 5123 5124 static int ext4_write_info(struct super_block *sb, int type) 5125 { 5126 int ret, err; 5127 handle_t *handle; 5128 5129 /* Data block + inode block */ 5130 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2); 5131 if (IS_ERR(handle)) 5132 return PTR_ERR(handle); 5133 ret = dquot_commit_info(sb, type); 5134 err = ext4_journal_stop(handle); 5135 if (!ret) 5136 ret = err; 5137 return ret; 5138 } 5139 5140 /* 5141 * Turn on quotas during mount time - we need to find 5142 * the quota file and such... 5143 */ 5144 static int ext4_quota_on_mount(struct super_block *sb, int type) 5145 { 5146 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 5147 EXT4_SB(sb)->s_jquota_fmt, type); 5148 } 5149 5150 /* 5151 * Standard function to be called on quota_on 5152 */ 5153 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5154 struct path *path) 5155 { 5156 int err; 5157 5158 if (!test_opt(sb, QUOTA)) 5159 return -EINVAL; 5160 5161 /* Quotafile not on the same filesystem? */ 5162 if (path->dentry->d_sb != sb) 5163 return -EXDEV; 5164 /* Journaling quota? */ 5165 if (EXT4_SB(sb)->s_qf_names[type]) { 5166 /* Quotafile not in fs root? */ 5167 if (path->dentry->d_parent != sb->s_root) 5168 ext4_msg(sb, KERN_WARNING, 5169 "Quota file not on filesystem root. " 5170 "Journaled quota will not work"); 5171 } 5172 5173 /* 5174 * When we journal data on quota file, we have to flush journal to see 5175 * all updates to the file when we bypass pagecache... 5176 */ 5177 if (EXT4_SB(sb)->s_journal && 5178 ext4_should_journal_data(path->dentry->d_inode)) { 5179 /* 5180 * We don't need to lock updates but journal_flush() could 5181 * otherwise be livelocked... 5182 */ 5183 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5184 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5185 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5186 if (err) 5187 return err; 5188 } 5189 5190 return dquot_quota_on(sb, type, format_id, path); 5191 } 5192 5193 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5194 unsigned int flags) 5195 { 5196 int err; 5197 struct inode *qf_inode; 5198 unsigned long qf_inums[MAXQUOTAS] = { 5199 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5200 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 5201 }; 5202 5203 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)); 5204 5205 if (!qf_inums[type]) 5206 return -EPERM; 5207 5208 qf_inode = ext4_iget(sb, qf_inums[type]); 5209 if (IS_ERR(qf_inode)) { 5210 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5211 return PTR_ERR(qf_inode); 5212 } 5213 5214 /* Don't account quota for quota files to avoid recursion */ 5215 qf_inode->i_flags |= S_NOQUOTA; 5216 err = dquot_enable(qf_inode, type, format_id, flags); 5217 iput(qf_inode); 5218 5219 return err; 5220 } 5221 5222 /* Enable usage tracking for all quota types. */ 5223 static int ext4_enable_quotas(struct super_block *sb) 5224 { 5225 int type, err = 0; 5226 unsigned long qf_inums[MAXQUOTAS] = { 5227 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5228 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 5229 }; 5230 5231 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 5232 for (type = 0; type < MAXQUOTAS; type++) { 5233 if (qf_inums[type]) { 5234 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5235 DQUOT_USAGE_ENABLED); 5236 if (err) { 5237 ext4_warning(sb, 5238 "Failed to enable quota tracking " 5239 "(type=%d, err=%d). Please run " 5240 "e2fsck to fix.", type, err); 5241 return err; 5242 } 5243 } 5244 } 5245 return 0; 5246 } 5247 5248 /* 5249 * quota_on function that is used when QUOTA feature is set. 5250 */ 5251 static int ext4_quota_on_sysfile(struct super_block *sb, int type, 5252 int format_id) 5253 { 5254 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 5255 return -EINVAL; 5256 5257 /* 5258 * USAGE was enabled at mount time. Only need to enable LIMITS now. 5259 */ 5260 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED); 5261 } 5262 5263 static int ext4_quota_off(struct super_block *sb, int type) 5264 { 5265 struct inode *inode = sb_dqopt(sb)->files[type]; 5266 handle_t *handle; 5267 5268 /* Force all delayed allocation blocks to be allocated. 5269 * Caller already holds s_umount sem */ 5270 if (test_opt(sb, DELALLOC)) 5271 sync_filesystem(sb); 5272 5273 if (!inode) 5274 goto out; 5275 5276 /* Update modification times of quota files when userspace can 5277 * start looking at them */ 5278 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5279 if (IS_ERR(handle)) 5280 goto out; 5281 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 5282 ext4_mark_inode_dirty(handle, inode); 5283 ext4_journal_stop(handle); 5284 5285 out: 5286 return dquot_quota_off(sb, type); 5287 } 5288 5289 /* 5290 * quota_off function that is used when QUOTA feature is set. 5291 */ 5292 static int ext4_quota_off_sysfile(struct super_block *sb, int type) 5293 { 5294 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 5295 return -EINVAL; 5296 5297 /* Disable only the limits. */ 5298 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED); 5299 } 5300 5301 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5302 * acquiring the locks... As quota files are never truncated and quota code 5303 * itself serializes the operations (and no one else should touch the files) 5304 * we don't have to be afraid of races */ 5305 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5306 size_t len, loff_t off) 5307 { 5308 struct inode *inode = sb_dqopt(sb)->files[type]; 5309 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5310 int err = 0; 5311 int offset = off & (sb->s_blocksize - 1); 5312 int tocopy; 5313 size_t toread; 5314 struct buffer_head *bh; 5315 loff_t i_size = i_size_read(inode); 5316 5317 if (off > i_size) 5318 return 0; 5319 if (off+len > i_size) 5320 len = i_size-off; 5321 toread = len; 5322 while (toread > 0) { 5323 tocopy = sb->s_blocksize - offset < toread ? 5324 sb->s_blocksize - offset : toread; 5325 bh = ext4_bread(NULL, inode, blk, 0, &err); 5326 if (err) 5327 return err; 5328 if (!bh) /* A hole? */ 5329 memset(data, 0, tocopy); 5330 else 5331 memcpy(data, bh->b_data+offset, tocopy); 5332 brelse(bh); 5333 offset = 0; 5334 toread -= tocopy; 5335 data += tocopy; 5336 blk++; 5337 } 5338 return len; 5339 } 5340 5341 /* Write to quotafile (we know the transaction is already started and has 5342 * enough credits) */ 5343 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5344 const char *data, size_t len, loff_t off) 5345 { 5346 struct inode *inode = sb_dqopt(sb)->files[type]; 5347 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5348 int err = 0; 5349 int offset = off & (sb->s_blocksize - 1); 5350 struct buffer_head *bh; 5351 handle_t *handle = journal_current_handle(); 5352 5353 if (EXT4_SB(sb)->s_journal && !handle) { 5354 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5355 " cancelled because transaction is not started", 5356 (unsigned long long)off, (unsigned long long)len); 5357 return -EIO; 5358 } 5359 /* 5360 * Since we account only one data block in transaction credits, 5361 * then it is impossible to cross a block boundary. 5362 */ 5363 if (sb->s_blocksize - offset < len) { 5364 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5365 " cancelled because not block aligned", 5366 (unsigned long long)off, (unsigned long long)len); 5367 return -EIO; 5368 } 5369 5370 bh = ext4_bread(handle, inode, blk, 1, &err); 5371 if (!bh) 5372 goto out; 5373 err = ext4_journal_get_write_access(handle, bh); 5374 if (err) { 5375 brelse(bh); 5376 goto out; 5377 } 5378 lock_buffer(bh); 5379 memcpy(bh->b_data+offset, data, len); 5380 flush_dcache_page(bh->b_page); 5381 unlock_buffer(bh); 5382 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5383 brelse(bh); 5384 out: 5385 if (err) 5386 return err; 5387 if (inode->i_size < off + len) { 5388 i_size_write(inode, off + len); 5389 EXT4_I(inode)->i_disksize = inode->i_size; 5390 ext4_mark_inode_dirty(handle, inode); 5391 } 5392 return len; 5393 } 5394 5395 #endif 5396 5397 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5398 const char *dev_name, void *data) 5399 { 5400 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5401 } 5402 5403 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 5404 static inline void register_as_ext2(void) 5405 { 5406 int err = register_filesystem(&ext2_fs_type); 5407 if (err) 5408 printk(KERN_WARNING 5409 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 5410 } 5411 5412 static inline void unregister_as_ext2(void) 5413 { 5414 unregister_filesystem(&ext2_fs_type); 5415 } 5416 5417 static inline int ext2_feature_set_ok(struct super_block *sb) 5418 { 5419 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) 5420 return 0; 5421 if (sb->s_flags & MS_RDONLY) 5422 return 1; 5423 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) 5424 return 0; 5425 return 1; 5426 } 5427 #else 5428 static inline void register_as_ext2(void) { } 5429 static inline void unregister_as_ext2(void) { } 5430 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 5431 #endif 5432 5433 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 5434 static inline void register_as_ext3(void) 5435 { 5436 int err = register_filesystem(&ext3_fs_type); 5437 if (err) 5438 printk(KERN_WARNING 5439 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 5440 } 5441 5442 static inline void unregister_as_ext3(void) 5443 { 5444 unregister_filesystem(&ext3_fs_type); 5445 } 5446 5447 static inline int ext3_feature_set_ok(struct super_block *sb) 5448 { 5449 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) 5450 return 0; 5451 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 5452 return 0; 5453 if (sb->s_flags & MS_RDONLY) 5454 return 1; 5455 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) 5456 return 0; 5457 return 1; 5458 } 5459 #else 5460 static inline void register_as_ext3(void) { } 5461 static inline void unregister_as_ext3(void) { } 5462 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } 5463 #endif 5464 5465 static struct file_system_type ext4_fs_type = { 5466 .owner = THIS_MODULE, 5467 .name = "ext4", 5468 .mount = ext4_mount, 5469 .kill_sb = kill_block_super, 5470 .fs_flags = FS_REQUIRES_DEV, 5471 }; 5472 MODULE_ALIAS_FS("ext4"); 5473 5474 static int __init ext4_init_feat_adverts(void) 5475 { 5476 struct ext4_features *ef; 5477 int ret = -ENOMEM; 5478 5479 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); 5480 if (!ef) 5481 goto out; 5482 5483 ef->f_kobj.kset = ext4_kset; 5484 init_completion(&ef->f_kobj_unregister); 5485 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, 5486 "features"); 5487 if (ret) { 5488 kfree(ef); 5489 goto out; 5490 } 5491 5492 ext4_feat = ef; 5493 ret = 0; 5494 out: 5495 return ret; 5496 } 5497 5498 static void ext4_exit_feat_adverts(void) 5499 { 5500 kobject_put(&ext4_feat->f_kobj); 5501 wait_for_completion(&ext4_feat->f_kobj_unregister); 5502 kfree(ext4_feat); 5503 } 5504 5505 /* Shared across all ext4 file systems */ 5506 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5507 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 5508 5509 static int __init ext4_init_fs(void) 5510 { 5511 int i, err; 5512 5513 ext4_li_info = NULL; 5514 mutex_init(&ext4_li_mtx); 5515 5516 /* Build-time check for flags consistency */ 5517 ext4_check_flag_values(); 5518 5519 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { 5520 mutex_init(&ext4__aio_mutex[i]); 5521 init_waitqueue_head(&ext4__ioend_wq[i]); 5522 } 5523 5524 err = ext4_init_es(); 5525 if (err) 5526 return err; 5527 5528 err = ext4_init_pageio(); 5529 if (err) 5530 goto out7; 5531 5532 err = ext4_init_system_zone(); 5533 if (err) 5534 goto out6; 5535 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 5536 if (!ext4_kset) { 5537 err = -ENOMEM; 5538 goto out5; 5539 } 5540 ext4_proc_root = proc_mkdir("fs/ext4", NULL); 5541 5542 err = ext4_init_feat_adverts(); 5543 if (err) 5544 goto out4; 5545 5546 err = ext4_init_mballoc(); 5547 if (err) 5548 goto out2; 5549 else 5550 ext4_mballoc_ready = 1; 5551 err = init_inodecache(); 5552 if (err) 5553 goto out1; 5554 register_as_ext3(); 5555 register_as_ext2(); 5556 err = register_filesystem(&ext4_fs_type); 5557 if (err) 5558 goto out; 5559 5560 return 0; 5561 out: 5562 unregister_as_ext2(); 5563 unregister_as_ext3(); 5564 destroy_inodecache(); 5565 out1: 5566 ext4_mballoc_ready = 0; 5567 ext4_exit_mballoc(); 5568 out2: 5569 ext4_exit_feat_adverts(); 5570 out4: 5571 if (ext4_proc_root) 5572 remove_proc_entry("fs/ext4", NULL); 5573 kset_unregister(ext4_kset); 5574 out5: 5575 ext4_exit_system_zone(); 5576 out6: 5577 ext4_exit_pageio(); 5578 out7: 5579 ext4_exit_es(); 5580 5581 return err; 5582 } 5583 5584 static void __exit ext4_exit_fs(void) 5585 { 5586 ext4_destroy_lazyinit_thread(); 5587 unregister_as_ext2(); 5588 unregister_as_ext3(); 5589 unregister_filesystem(&ext4_fs_type); 5590 destroy_inodecache(); 5591 ext4_exit_mballoc(); 5592 ext4_exit_feat_adverts(); 5593 remove_proc_entry("fs/ext4", NULL); 5594 kset_unregister(ext4_kset); 5595 ext4_exit_system_zone(); 5596 ext4_exit_pageio(); 5597 ext4_exit_es(); 5598 } 5599 5600 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 5601 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5602 MODULE_LICENSE("GPL"); 5603 module_init(ext4_init_fs) 5604 module_exit(ext4_exit_fs) 5605