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 63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 64 unsigned long journal_devnum); 65 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 66 static int ext4_commit_super(struct super_block *sb, int sync); 67 static void ext4_mark_recovery_complete(struct super_block *sb, 68 struct ext4_super_block *es); 69 static void ext4_clear_journal_err(struct super_block *sb, 70 struct ext4_super_block *es); 71 static int ext4_sync_fs(struct super_block *sb, int wait); 72 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait); 73 static int ext4_remount(struct super_block *sb, int *flags, char *data); 74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 75 static int ext4_unfreeze(struct super_block *sb); 76 static int ext4_freeze(struct super_block *sb); 77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 78 const char *dev_name, void *data); 79 static inline int ext2_feature_set_ok(struct super_block *sb); 80 static inline int ext3_feature_set_ok(struct super_block *sb); 81 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 82 static void ext4_destroy_lazyinit_thread(void); 83 static void ext4_unregister_li_request(struct super_block *sb); 84 static void ext4_clear_request_list(void); 85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t); 86 87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 88 static struct file_system_type ext2_fs_type = { 89 .owner = THIS_MODULE, 90 .name = "ext2", 91 .mount = ext4_mount, 92 .kill_sb = kill_block_super, 93 .fs_flags = FS_REQUIRES_DEV, 94 }; 95 MODULE_ALIAS_FS("ext2"); 96 MODULE_ALIAS("ext2"); 97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 98 #else 99 #define IS_EXT2_SB(sb) (0) 100 #endif 101 102 103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 104 static struct file_system_type ext3_fs_type = { 105 .owner = THIS_MODULE, 106 .name = "ext3", 107 .mount = ext4_mount, 108 .kill_sb = kill_block_super, 109 .fs_flags = FS_REQUIRES_DEV, 110 }; 111 MODULE_ALIAS_FS("ext3"); 112 MODULE_ALIAS("ext3"); 113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 114 #else 115 #define IS_EXT3_SB(sb) (0) 116 #endif 117 118 static int ext4_verify_csum_type(struct super_block *sb, 119 struct ext4_super_block *es) 120 { 121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 123 return 1; 124 125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 126 } 127 128 static __le32 ext4_superblock_csum(struct super_block *sb, 129 struct ext4_super_block *es) 130 { 131 struct ext4_sb_info *sbi = EXT4_SB(sb); 132 int offset = offsetof(struct ext4_super_block, s_checksum); 133 __u32 csum; 134 135 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 136 137 return cpu_to_le32(csum); 138 } 139 140 int ext4_superblock_csum_verify(struct super_block *sb, 141 struct ext4_super_block *es) 142 { 143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 145 return 1; 146 147 return es->s_checksum == ext4_superblock_csum(sb, es); 148 } 149 150 void ext4_superblock_csum_set(struct super_block *sb) 151 { 152 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 153 154 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 155 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 156 return; 157 158 es->s_checksum = ext4_superblock_csum(sb, es); 159 } 160 161 void *ext4_kvmalloc(size_t size, gfp_t flags) 162 { 163 void *ret; 164 165 ret = kmalloc(size, flags | __GFP_NOWARN); 166 if (!ret) 167 ret = __vmalloc(size, flags, PAGE_KERNEL); 168 return ret; 169 } 170 171 void *ext4_kvzalloc(size_t size, gfp_t flags) 172 { 173 void *ret; 174 175 ret = kzalloc(size, flags | __GFP_NOWARN); 176 if (!ret) 177 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 178 return ret; 179 } 180 181 void ext4_kvfree(void *ptr) 182 { 183 if (is_vmalloc_addr(ptr)) 184 vfree(ptr); 185 else 186 kfree(ptr); 187 188 } 189 190 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 191 struct ext4_group_desc *bg) 192 { 193 return le32_to_cpu(bg->bg_block_bitmap_lo) | 194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 195 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 196 } 197 198 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 199 struct ext4_group_desc *bg) 200 { 201 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 202 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 203 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 204 } 205 206 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 207 struct ext4_group_desc *bg) 208 { 209 return le32_to_cpu(bg->bg_inode_table_lo) | 210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 211 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 212 } 213 214 __u32 ext4_free_group_clusters(struct super_block *sb, 215 struct ext4_group_desc *bg) 216 { 217 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 219 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 220 } 221 222 __u32 ext4_free_inodes_count(struct super_block *sb, 223 struct ext4_group_desc *bg) 224 { 225 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 227 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 228 } 229 230 __u32 ext4_used_dirs_count(struct super_block *sb, 231 struct ext4_group_desc *bg) 232 { 233 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 235 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 236 } 237 238 __u32 ext4_itable_unused_count(struct super_block *sb, 239 struct ext4_group_desc *bg) 240 { 241 return le16_to_cpu(bg->bg_itable_unused_lo) | 242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 243 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 244 } 245 246 void ext4_block_bitmap_set(struct super_block *sb, 247 struct ext4_group_desc *bg, ext4_fsblk_t blk) 248 { 249 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 251 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 252 } 253 254 void ext4_inode_bitmap_set(struct super_block *sb, 255 struct ext4_group_desc *bg, ext4_fsblk_t blk) 256 { 257 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 258 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 259 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 260 } 261 262 void ext4_inode_table_set(struct super_block *sb, 263 struct ext4_group_desc *bg, ext4_fsblk_t blk) 264 { 265 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 267 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 268 } 269 270 void ext4_free_group_clusters_set(struct super_block *sb, 271 struct ext4_group_desc *bg, __u32 count) 272 { 273 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 275 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 276 } 277 278 void ext4_free_inodes_set(struct super_block *sb, 279 struct ext4_group_desc *bg, __u32 count) 280 { 281 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 283 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 284 } 285 286 void ext4_used_dirs_set(struct super_block *sb, 287 struct ext4_group_desc *bg, __u32 count) 288 { 289 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 291 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 292 } 293 294 void ext4_itable_unused_set(struct super_block *sb, 295 struct ext4_group_desc *bg, __u32 count) 296 { 297 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 299 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 300 } 301 302 303 static void __save_error_info(struct super_block *sb, const char *func, 304 unsigned int line) 305 { 306 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 307 308 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 309 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 310 es->s_last_error_time = cpu_to_le32(get_seconds()); 311 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 312 es->s_last_error_line = cpu_to_le32(line); 313 if (!es->s_first_error_time) { 314 es->s_first_error_time = es->s_last_error_time; 315 strncpy(es->s_first_error_func, func, 316 sizeof(es->s_first_error_func)); 317 es->s_first_error_line = cpu_to_le32(line); 318 es->s_first_error_ino = es->s_last_error_ino; 319 es->s_first_error_block = es->s_last_error_block; 320 } 321 /* 322 * Start the daily error reporting function if it hasn't been 323 * started already 324 */ 325 if (!es->s_error_count) 326 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 327 le32_add_cpu(&es->s_error_count, 1); 328 } 329 330 static void save_error_info(struct super_block *sb, const char *func, 331 unsigned int line) 332 { 333 __save_error_info(sb, func, line); 334 ext4_commit_super(sb, 1); 335 } 336 337 /* 338 * The del_gendisk() function uninitializes the disk-specific data 339 * structures, including the bdi structure, without telling anyone 340 * else. Once this happens, any attempt to call mark_buffer_dirty() 341 * (for example, by ext4_commit_super), will cause a kernel OOPS. 342 * This is a kludge to prevent these oops until we can put in a proper 343 * hook in del_gendisk() to inform the VFS and file system layers. 344 */ 345 static int block_device_ejected(struct super_block *sb) 346 { 347 struct inode *bd_inode = sb->s_bdev->bd_inode; 348 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info; 349 350 return bdi->dev == NULL; 351 } 352 353 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 354 { 355 struct super_block *sb = journal->j_private; 356 struct ext4_sb_info *sbi = EXT4_SB(sb); 357 int error = is_journal_aborted(journal); 358 struct ext4_journal_cb_entry *jce; 359 360 BUG_ON(txn->t_state == T_FINISHED); 361 spin_lock(&sbi->s_md_lock); 362 while (!list_empty(&txn->t_private_list)) { 363 jce = list_entry(txn->t_private_list.next, 364 struct ext4_journal_cb_entry, jce_list); 365 list_del_init(&jce->jce_list); 366 spin_unlock(&sbi->s_md_lock); 367 jce->jce_func(sb, jce, error); 368 spin_lock(&sbi->s_md_lock); 369 } 370 spin_unlock(&sbi->s_md_lock); 371 } 372 373 /* Deal with the reporting of failure conditions on a filesystem such as 374 * inconsistencies detected or read IO failures. 375 * 376 * On ext2, we can store the error state of the filesystem in the 377 * superblock. That is not possible on ext4, because we may have other 378 * write ordering constraints on the superblock which prevent us from 379 * writing it out straight away; and given that the journal is about to 380 * be aborted, we can't rely on the current, or future, transactions to 381 * write out the superblock safely. 382 * 383 * We'll just use the jbd2_journal_abort() error code to record an error in 384 * the journal instead. On recovery, the journal will complain about 385 * that error until we've noted it down and cleared it. 386 */ 387 388 static void ext4_handle_error(struct super_block *sb) 389 { 390 if (sb->s_flags & MS_RDONLY) 391 return; 392 393 if (!test_opt(sb, ERRORS_CONT)) { 394 journal_t *journal = EXT4_SB(sb)->s_journal; 395 396 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 397 if (journal) 398 jbd2_journal_abort(journal, -EIO); 399 } 400 if (test_opt(sb, ERRORS_RO)) { 401 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 402 /* 403 * Make sure updated value of ->s_mount_flags will be visible 404 * before ->s_flags update 405 */ 406 smp_wmb(); 407 sb->s_flags |= MS_RDONLY; 408 } 409 if (test_opt(sb, ERRORS_PANIC)) 410 panic("EXT4-fs (device %s): panic forced after error\n", 411 sb->s_id); 412 } 413 414 #define ext4_error_ratelimit(sb) \ 415 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 416 "EXT4-fs error") 417 418 void __ext4_error(struct super_block *sb, const char *function, 419 unsigned int line, const char *fmt, ...) 420 { 421 struct va_format vaf; 422 va_list args; 423 424 if (ext4_error_ratelimit(sb)) { 425 va_start(args, fmt); 426 vaf.fmt = fmt; 427 vaf.va = &args; 428 printk(KERN_CRIT 429 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 430 sb->s_id, function, line, current->comm, &vaf); 431 va_end(args); 432 } 433 save_error_info(sb, function, line); 434 ext4_handle_error(sb); 435 } 436 437 void __ext4_error_inode(struct inode *inode, const char *function, 438 unsigned int line, ext4_fsblk_t block, 439 const char *fmt, ...) 440 { 441 va_list args; 442 struct va_format vaf; 443 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 444 445 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 446 es->s_last_error_block = cpu_to_le64(block); 447 if (ext4_error_ratelimit(inode->i_sb)) { 448 va_start(args, fmt); 449 vaf.fmt = fmt; 450 vaf.va = &args; 451 if (block) 452 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 453 "inode #%lu: block %llu: comm %s: %pV\n", 454 inode->i_sb->s_id, function, line, inode->i_ino, 455 block, current->comm, &vaf); 456 else 457 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 458 "inode #%lu: comm %s: %pV\n", 459 inode->i_sb->s_id, function, line, inode->i_ino, 460 current->comm, &vaf); 461 va_end(args); 462 } 463 save_error_info(inode->i_sb, function, line); 464 ext4_handle_error(inode->i_sb); 465 } 466 467 void __ext4_error_file(struct file *file, const char *function, 468 unsigned int line, ext4_fsblk_t block, 469 const char *fmt, ...) 470 { 471 va_list args; 472 struct va_format vaf; 473 struct ext4_super_block *es; 474 struct inode *inode = file_inode(file); 475 char pathname[80], *path; 476 477 es = EXT4_SB(inode->i_sb)->s_es; 478 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 479 if (ext4_error_ratelimit(inode->i_sb)) { 480 path = d_path(&(file->f_path), pathname, sizeof(pathname)); 481 if (IS_ERR(path)) 482 path = "(unknown)"; 483 va_start(args, fmt); 484 vaf.fmt = fmt; 485 vaf.va = &args; 486 if (block) 487 printk(KERN_CRIT 488 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 489 "block %llu: comm %s: path %s: %pV\n", 490 inode->i_sb->s_id, function, line, inode->i_ino, 491 block, current->comm, path, &vaf); 492 else 493 printk(KERN_CRIT 494 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 495 "comm %s: path %s: %pV\n", 496 inode->i_sb->s_id, function, line, inode->i_ino, 497 current->comm, path, &vaf); 498 va_end(args); 499 } 500 save_error_info(inode->i_sb, function, line); 501 ext4_handle_error(inode->i_sb); 502 } 503 504 const char *ext4_decode_error(struct super_block *sb, int errno, 505 char nbuf[16]) 506 { 507 char *errstr = NULL; 508 509 switch (errno) { 510 case -EIO: 511 errstr = "IO failure"; 512 break; 513 case -ENOMEM: 514 errstr = "Out of memory"; 515 break; 516 case -EROFS: 517 if (!sb || (EXT4_SB(sb)->s_journal && 518 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 519 errstr = "Journal has aborted"; 520 else 521 errstr = "Readonly filesystem"; 522 break; 523 default: 524 /* If the caller passed in an extra buffer for unknown 525 * errors, textualise them now. Else we just return 526 * NULL. */ 527 if (nbuf) { 528 /* Check for truncated error codes... */ 529 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 530 errstr = nbuf; 531 } 532 break; 533 } 534 535 return errstr; 536 } 537 538 /* __ext4_std_error decodes expected errors from journaling functions 539 * automatically and invokes the appropriate error response. */ 540 541 void __ext4_std_error(struct super_block *sb, const char *function, 542 unsigned int line, int errno) 543 { 544 char nbuf[16]; 545 const char *errstr; 546 547 /* Special case: if the error is EROFS, and we're not already 548 * inside a transaction, then there's really no point in logging 549 * an error. */ 550 if (errno == -EROFS && journal_current_handle() == NULL && 551 (sb->s_flags & MS_RDONLY)) 552 return; 553 554 if (ext4_error_ratelimit(sb)) { 555 errstr = ext4_decode_error(sb, errno, nbuf); 556 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 557 sb->s_id, function, line, errstr); 558 } 559 560 save_error_info(sb, function, line); 561 ext4_handle_error(sb); 562 } 563 564 /* 565 * ext4_abort is a much stronger failure handler than ext4_error. The 566 * abort function may be used to deal with unrecoverable failures such 567 * as journal IO errors or ENOMEM at a critical moment in log management. 568 * 569 * We unconditionally force the filesystem into an ABORT|READONLY state, 570 * unless the error response on the fs has been set to panic in which 571 * case we take the easy way out and panic immediately. 572 */ 573 574 void __ext4_abort(struct super_block *sb, const char *function, 575 unsigned int line, const char *fmt, ...) 576 { 577 va_list args; 578 579 save_error_info(sb, function, line); 580 va_start(args, fmt); 581 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, 582 function, line); 583 vprintk(fmt, args); 584 printk("\n"); 585 va_end(args); 586 587 if ((sb->s_flags & MS_RDONLY) == 0) { 588 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 589 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 590 /* 591 * Make sure updated value of ->s_mount_flags will be visible 592 * before ->s_flags update 593 */ 594 smp_wmb(); 595 sb->s_flags |= MS_RDONLY; 596 if (EXT4_SB(sb)->s_journal) 597 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 598 save_error_info(sb, function, line); 599 } 600 if (test_opt(sb, ERRORS_PANIC)) 601 panic("EXT4-fs panic from previous error\n"); 602 } 603 604 void __ext4_msg(struct super_block *sb, 605 const char *prefix, const char *fmt, ...) 606 { 607 struct va_format vaf; 608 va_list args; 609 610 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 611 return; 612 613 va_start(args, fmt); 614 vaf.fmt = fmt; 615 vaf.va = &args; 616 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 617 va_end(args); 618 } 619 620 void __ext4_warning(struct super_block *sb, const char *function, 621 unsigned int line, const char *fmt, ...) 622 { 623 struct va_format vaf; 624 va_list args; 625 626 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 627 "EXT4-fs warning")) 628 return; 629 630 va_start(args, fmt); 631 vaf.fmt = fmt; 632 vaf.va = &args; 633 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 634 sb->s_id, function, line, &vaf); 635 va_end(args); 636 } 637 638 void __ext4_grp_locked_error(const char *function, unsigned int line, 639 struct super_block *sb, ext4_group_t grp, 640 unsigned long ino, ext4_fsblk_t block, 641 const char *fmt, ...) 642 __releases(bitlock) 643 __acquires(bitlock) 644 { 645 struct va_format vaf; 646 va_list args; 647 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 648 649 es->s_last_error_ino = cpu_to_le32(ino); 650 es->s_last_error_block = cpu_to_le64(block); 651 __save_error_info(sb, function, line); 652 653 if (ext4_error_ratelimit(sb)) { 654 va_start(args, fmt); 655 vaf.fmt = fmt; 656 vaf.va = &args; 657 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 658 sb->s_id, function, line, grp); 659 if (ino) 660 printk(KERN_CONT "inode %lu: ", ino); 661 if (block) 662 printk(KERN_CONT "block %llu:", 663 (unsigned long long) block); 664 printk(KERN_CONT "%pV\n", &vaf); 665 va_end(args); 666 } 667 668 if (test_opt(sb, ERRORS_CONT)) { 669 ext4_commit_super(sb, 0); 670 return; 671 } 672 673 ext4_unlock_group(sb, grp); 674 ext4_handle_error(sb); 675 /* 676 * We only get here in the ERRORS_RO case; relocking the group 677 * may be dangerous, but nothing bad will happen since the 678 * filesystem will have already been marked read/only and the 679 * journal has been aborted. We return 1 as a hint to callers 680 * who might what to use the return value from 681 * ext4_grp_locked_error() to distinguish between the 682 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 683 * aggressively from the ext4 function in question, with a 684 * more appropriate error code. 685 */ 686 ext4_lock_group(sb, grp); 687 return; 688 } 689 690 void ext4_update_dynamic_rev(struct super_block *sb) 691 { 692 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 693 694 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 695 return; 696 697 ext4_warning(sb, 698 "updating to rev %d because of new feature flag, " 699 "running e2fsck is recommended", 700 EXT4_DYNAMIC_REV); 701 702 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 703 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 704 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 705 /* leave es->s_feature_*compat flags alone */ 706 /* es->s_uuid will be set by e2fsck if empty */ 707 708 /* 709 * The rest of the superblock fields should be zero, and if not it 710 * means they are likely already in use, so leave them alone. We 711 * can leave it up to e2fsck to clean up any inconsistencies there. 712 */ 713 } 714 715 /* 716 * Open the external journal device 717 */ 718 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 719 { 720 struct block_device *bdev; 721 char b[BDEVNAME_SIZE]; 722 723 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 724 if (IS_ERR(bdev)) 725 goto fail; 726 return bdev; 727 728 fail: 729 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 730 __bdevname(dev, b), PTR_ERR(bdev)); 731 return NULL; 732 } 733 734 /* 735 * Release the journal device 736 */ 737 static void ext4_blkdev_put(struct block_device *bdev) 738 { 739 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 740 } 741 742 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 743 { 744 struct block_device *bdev; 745 bdev = sbi->journal_bdev; 746 if (bdev) { 747 ext4_blkdev_put(bdev); 748 sbi->journal_bdev = NULL; 749 } 750 } 751 752 static inline struct inode *orphan_list_entry(struct list_head *l) 753 { 754 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 755 } 756 757 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 758 { 759 struct list_head *l; 760 761 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 762 le32_to_cpu(sbi->s_es->s_last_orphan)); 763 764 printk(KERN_ERR "sb_info orphan list:\n"); 765 list_for_each(l, &sbi->s_orphan) { 766 struct inode *inode = orphan_list_entry(l); 767 printk(KERN_ERR " " 768 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 769 inode->i_sb->s_id, inode->i_ino, inode, 770 inode->i_mode, inode->i_nlink, 771 NEXT_ORPHAN(inode)); 772 } 773 } 774 775 static void ext4_put_super(struct super_block *sb) 776 { 777 struct ext4_sb_info *sbi = EXT4_SB(sb); 778 struct ext4_super_block *es = sbi->s_es; 779 int i, err; 780 781 ext4_unregister_li_request(sb); 782 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 783 784 flush_workqueue(sbi->rsv_conversion_wq); 785 destroy_workqueue(sbi->rsv_conversion_wq); 786 787 if (sbi->s_journal) { 788 err = jbd2_journal_destroy(sbi->s_journal); 789 sbi->s_journal = NULL; 790 if (err < 0) 791 ext4_abort(sb, "Couldn't clean up the journal"); 792 } 793 794 ext4_es_unregister_shrinker(sbi); 795 del_timer(&sbi->s_err_report); 796 ext4_release_system_zone(sb); 797 ext4_mb_release(sb); 798 ext4_ext_release(sb); 799 ext4_xattr_put_super(sb); 800 801 if (!(sb->s_flags & MS_RDONLY)) { 802 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 803 es->s_state = cpu_to_le16(sbi->s_mount_state); 804 } 805 if (!(sb->s_flags & MS_RDONLY)) 806 ext4_commit_super(sb, 1); 807 808 if (sbi->s_proc) { 809 remove_proc_entry("options", sbi->s_proc); 810 remove_proc_entry(sb->s_id, ext4_proc_root); 811 } 812 kobject_del(&sbi->s_kobj); 813 814 for (i = 0; i < sbi->s_gdb_count; i++) 815 brelse(sbi->s_group_desc[i]); 816 ext4_kvfree(sbi->s_group_desc); 817 ext4_kvfree(sbi->s_flex_groups); 818 percpu_counter_destroy(&sbi->s_freeclusters_counter); 819 percpu_counter_destroy(&sbi->s_freeinodes_counter); 820 percpu_counter_destroy(&sbi->s_dirs_counter); 821 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 822 percpu_counter_destroy(&sbi->s_extent_cache_cnt); 823 brelse(sbi->s_sbh); 824 #ifdef CONFIG_QUOTA 825 for (i = 0; i < MAXQUOTAS; i++) 826 kfree(sbi->s_qf_names[i]); 827 #endif 828 829 /* Debugging code just in case the in-memory inode orphan list 830 * isn't empty. The on-disk one can be non-empty if we've 831 * detected an error and taken the fs readonly, but the 832 * in-memory list had better be clean by this point. */ 833 if (!list_empty(&sbi->s_orphan)) 834 dump_orphan_list(sb, sbi); 835 J_ASSERT(list_empty(&sbi->s_orphan)); 836 837 invalidate_bdev(sb->s_bdev); 838 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 839 /* 840 * Invalidate the journal device's buffers. We don't want them 841 * floating about in memory - the physical journal device may 842 * hotswapped, and it breaks the `ro-after' testing code. 843 */ 844 sync_blockdev(sbi->journal_bdev); 845 invalidate_bdev(sbi->journal_bdev); 846 ext4_blkdev_remove(sbi); 847 } 848 if (sbi->s_mmp_tsk) 849 kthread_stop(sbi->s_mmp_tsk); 850 sb->s_fs_info = NULL; 851 /* 852 * Now that we are completely done shutting down the 853 * superblock, we need to actually destroy the kobject. 854 */ 855 kobject_put(&sbi->s_kobj); 856 wait_for_completion(&sbi->s_kobj_unregister); 857 if (sbi->s_chksum_driver) 858 crypto_free_shash(sbi->s_chksum_driver); 859 kfree(sbi->s_blockgroup_lock); 860 kfree(sbi); 861 } 862 863 static struct kmem_cache *ext4_inode_cachep; 864 865 /* 866 * Called inside transaction, so use GFP_NOFS 867 */ 868 static struct inode *ext4_alloc_inode(struct super_block *sb) 869 { 870 struct ext4_inode_info *ei; 871 872 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 873 if (!ei) 874 return NULL; 875 876 ei->vfs_inode.i_version = 1; 877 INIT_LIST_HEAD(&ei->i_prealloc_list); 878 spin_lock_init(&ei->i_prealloc_lock); 879 ext4_es_init_tree(&ei->i_es_tree); 880 rwlock_init(&ei->i_es_lock); 881 INIT_LIST_HEAD(&ei->i_es_lru); 882 ei->i_es_lru_nr = 0; 883 ei->i_touch_when = 0; 884 ei->i_reserved_data_blocks = 0; 885 ei->i_reserved_meta_blocks = 0; 886 ei->i_allocated_meta_blocks = 0; 887 ei->i_da_metadata_calc_len = 0; 888 ei->i_da_metadata_calc_last_lblock = 0; 889 spin_lock_init(&(ei->i_block_reservation_lock)); 890 #ifdef CONFIG_QUOTA 891 ei->i_reserved_quota = 0; 892 #endif 893 ei->jinode = NULL; 894 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 895 spin_lock_init(&ei->i_completed_io_lock); 896 ei->i_sync_tid = 0; 897 ei->i_datasync_tid = 0; 898 atomic_set(&ei->i_ioend_count, 0); 899 atomic_set(&ei->i_unwritten, 0); 900 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 901 902 return &ei->vfs_inode; 903 } 904 905 static int ext4_drop_inode(struct inode *inode) 906 { 907 int drop = generic_drop_inode(inode); 908 909 trace_ext4_drop_inode(inode, drop); 910 return drop; 911 } 912 913 static void ext4_i_callback(struct rcu_head *head) 914 { 915 struct inode *inode = container_of(head, struct inode, i_rcu); 916 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 917 } 918 919 static void ext4_destroy_inode(struct inode *inode) 920 { 921 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 922 ext4_msg(inode->i_sb, KERN_ERR, 923 "Inode %lu (%p): orphan list check failed!", 924 inode->i_ino, EXT4_I(inode)); 925 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 926 EXT4_I(inode), sizeof(struct ext4_inode_info), 927 true); 928 dump_stack(); 929 } 930 call_rcu(&inode->i_rcu, ext4_i_callback); 931 } 932 933 static void init_once(void *foo) 934 { 935 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 936 937 INIT_LIST_HEAD(&ei->i_orphan); 938 init_rwsem(&ei->xattr_sem); 939 init_rwsem(&ei->i_data_sem); 940 inode_init_once(&ei->vfs_inode); 941 } 942 943 static int init_inodecache(void) 944 { 945 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 946 sizeof(struct ext4_inode_info), 947 0, (SLAB_RECLAIM_ACCOUNT| 948 SLAB_MEM_SPREAD), 949 init_once); 950 if (ext4_inode_cachep == NULL) 951 return -ENOMEM; 952 return 0; 953 } 954 955 static void destroy_inodecache(void) 956 { 957 /* 958 * Make sure all delayed rcu free inodes are flushed before we 959 * destroy cache. 960 */ 961 rcu_barrier(); 962 kmem_cache_destroy(ext4_inode_cachep); 963 } 964 965 void ext4_clear_inode(struct inode *inode) 966 { 967 invalidate_inode_buffers(inode); 968 clear_inode(inode); 969 dquot_drop(inode); 970 ext4_discard_preallocations(inode); 971 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 972 ext4_es_lru_del(inode); 973 if (EXT4_I(inode)->jinode) { 974 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 975 EXT4_I(inode)->jinode); 976 jbd2_free_inode(EXT4_I(inode)->jinode); 977 EXT4_I(inode)->jinode = NULL; 978 } 979 } 980 981 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 982 u64 ino, u32 generation) 983 { 984 struct inode *inode; 985 986 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 987 return ERR_PTR(-ESTALE); 988 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 989 return ERR_PTR(-ESTALE); 990 991 /* iget isn't really right if the inode is currently unallocated!! 992 * 993 * ext4_read_inode will return a bad_inode if the inode had been 994 * deleted, so we should be safe. 995 * 996 * Currently we don't know the generation for parent directory, so 997 * a generation of 0 means "accept any" 998 */ 999 inode = ext4_iget(sb, ino); 1000 if (IS_ERR(inode)) 1001 return ERR_CAST(inode); 1002 if (generation && inode->i_generation != generation) { 1003 iput(inode); 1004 return ERR_PTR(-ESTALE); 1005 } 1006 1007 return inode; 1008 } 1009 1010 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1011 int fh_len, int fh_type) 1012 { 1013 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1014 ext4_nfs_get_inode); 1015 } 1016 1017 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1018 int fh_len, int fh_type) 1019 { 1020 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1021 ext4_nfs_get_inode); 1022 } 1023 1024 /* 1025 * Try to release metadata pages (indirect blocks, directories) which are 1026 * mapped via the block device. Since these pages could have journal heads 1027 * which would prevent try_to_free_buffers() from freeing them, we must use 1028 * jbd2 layer's try_to_free_buffers() function to release them. 1029 */ 1030 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1031 gfp_t wait) 1032 { 1033 journal_t *journal = EXT4_SB(sb)->s_journal; 1034 1035 WARN_ON(PageChecked(page)); 1036 if (!page_has_buffers(page)) 1037 return 0; 1038 if (journal) 1039 return jbd2_journal_try_to_free_buffers(journal, page, 1040 wait & ~__GFP_WAIT); 1041 return try_to_free_buffers(page); 1042 } 1043 1044 #ifdef CONFIG_QUOTA 1045 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") 1046 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) 1047 1048 static int ext4_write_dquot(struct dquot *dquot); 1049 static int ext4_acquire_dquot(struct dquot *dquot); 1050 static int ext4_release_dquot(struct dquot *dquot); 1051 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1052 static int ext4_write_info(struct super_block *sb, int type); 1053 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1054 struct path *path); 1055 static int ext4_quota_on_sysfile(struct super_block *sb, int type, 1056 int format_id); 1057 static int ext4_quota_off(struct super_block *sb, int type); 1058 static int ext4_quota_off_sysfile(struct super_block *sb, int type); 1059 static int ext4_quota_on_mount(struct super_block *sb, int type); 1060 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1061 size_t len, loff_t off); 1062 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1063 const char *data, size_t len, loff_t off); 1064 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1065 unsigned int flags); 1066 static int ext4_enable_quotas(struct super_block *sb); 1067 1068 static const struct dquot_operations ext4_quota_operations = { 1069 .get_reserved_space = ext4_get_reserved_space, 1070 .write_dquot = ext4_write_dquot, 1071 .acquire_dquot = ext4_acquire_dquot, 1072 .release_dquot = ext4_release_dquot, 1073 .mark_dirty = ext4_mark_dquot_dirty, 1074 .write_info = ext4_write_info, 1075 .alloc_dquot = dquot_alloc, 1076 .destroy_dquot = dquot_destroy, 1077 }; 1078 1079 static const struct quotactl_ops ext4_qctl_operations = { 1080 .quota_on = ext4_quota_on, 1081 .quota_off = ext4_quota_off, 1082 .quota_sync = dquot_quota_sync, 1083 .get_info = dquot_get_dqinfo, 1084 .set_info = dquot_set_dqinfo, 1085 .get_dqblk = dquot_get_dqblk, 1086 .set_dqblk = dquot_set_dqblk 1087 }; 1088 1089 static const struct quotactl_ops ext4_qctl_sysfile_operations = { 1090 .quota_on_meta = ext4_quota_on_sysfile, 1091 .quota_off = ext4_quota_off_sysfile, 1092 .quota_sync = dquot_quota_sync, 1093 .get_info = dquot_get_dqinfo, 1094 .set_info = dquot_set_dqinfo, 1095 .get_dqblk = dquot_get_dqblk, 1096 .set_dqblk = dquot_set_dqblk 1097 }; 1098 #endif 1099 1100 static const struct super_operations ext4_sops = { 1101 .alloc_inode = ext4_alloc_inode, 1102 .destroy_inode = ext4_destroy_inode, 1103 .write_inode = ext4_write_inode, 1104 .dirty_inode = ext4_dirty_inode, 1105 .drop_inode = ext4_drop_inode, 1106 .evict_inode = ext4_evict_inode, 1107 .put_super = ext4_put_super, 1108 .sync_fs = ext4_sync_fs, 1109 .freeze_fs = ext4_freeze, 1110 .unfreeze_fs = ext4_unfreeze, 1111 .statfs = ext4_statfs, 1112 .remount_fs = ext4_remount, 1113 .show_options = ext4_show_options, 1114 #ifdef CONFIG_QUOTA 1115 .quota_read = ext4_quota_read, 1116 .quota_write = ext4_quota_write, 1117 #endif 1118 .bdev_try_to_free_page = bdev_try_to_free_page, 1119 }; 1120 1121 static const struct super_operations ext4_nojournal_sops = { 1122 .alloc_inode = ext4_alloc_inode, 1123 .destroy_inode = ext4_destroy_inode, 1124 .write_inode = ext4_write_inode, 1125 .dirty_inode = ext4_dirty_inode, 1126 .drop_inode = ext4_drop_inode, 1127 .evict_inode = ext4_evict_inode, 1128 .sync_fs = ext4_sync_fs_nojournal, 1129 .put_super = ext4_put_super, 1130 .statfs = ext4_statfs, 1131 .remount_fs = ext4_remount, 1132 .show_options = ext4_show_options, 1133 #ifdef CONFIG_QUOTA 1134 .quota_read = ext4_quota_read, 1135 .quota_write = ext4_quota_write, 1136 #endif 1137 .bdev_try_to_free_page = bdev_try_to_free_page, 1138 }; 1139 1140 static const struct export_operations ext4_export_ops = { 1141 .fh_to_dentry = ext4_fh_to_dentry, 1142 .fh_to_parent = ext4_fh_to_parent, 1143 .get_parent = ext4_get_parent, 1144 }; 1145 1146 enum { 1147 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1148 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1149 Opt_nouid32, Opt_debug, Opt_removed, 1150 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1151 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1152 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1153 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1154 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1155 Opt_data_err_abort, Opt_data_err_ignore, 1156 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1157 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1158 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1159 Opt_usrquota, Opt_grpquota, Opt_i_version, 1160 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 1161 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1162 Opt_inode_readahead_blks, Opt_journal_ioprio, 1163 Opt_dioread_nolock, Opt_dioread_lock, 1164 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1165 Opt_max_dir_size_kb, 1166 }; 1167 1168 static const match_table_t tokens = { 1169 {Opt_bsd_df, "bsddf"}, 1170 {Opt_minix_df, "minixdf"}, 1171 {Opt_grpid, "grpid"}, 1172 {Opt_grpid, "bsdgroups"}, 1173 {Opt_nogrpid, "nogrpid"}, 1174 {Opt_nogrpid, "sysvgroups"}, 1175 {Opt_resgid, "resgid=%u"}, 1176 {Opt_resuid, "resuid=%u"}, 1177 {Opt_sb, "sb=%u"}, 1178 {Opt_err_cont, "errors=continue"}, 1179 {Opt_err_panic, "errors=panic"}, 1180 {Opt_err_ro, "errors=remount-ro"}, 1181 {Opt_nouid32, "nouid32"}, 1182 {Opt_debug, "debug"}, 1183 {Opt_removed, "oldalloc"}, 1184 {Opt_removed, "orlov"}, 1185 {Opt_user_xattr, "user_xattr"}, 1186 {Opt_nouser_xattr, "nouser_xattr"}, 1187 {Opt_acl, "acl"}, 1188 {Opt_noacl, "noacl"}, 1189 {Opt_noload, "norecovery"}, 1190 {Opt_noload, "noload"}, 1191 {Opt_removed, "nobh"}, 1192 {Opt_removed, "bh"}, 1193 {Opt_commit, "commit=%u"}, 1194 {Opt_min_batch_time, "min_batch_time=%u"}, 1195 {Opt_max_batch_time, "max_batch_time=%u"}, 1196 {Opt_journal_dev, "journal_dev=%u"}, 1197 {Opt_journal_path, "journal_path=%s"}, 1198 {Opt_journal_checksum, "journal_checksum"}, 1199 {Opt_journal_async_commit, "journal_async_commit"}, 1200 {Opt_abort, "abort"}, 1201 {Opt_data_journal, "data=journal"}, 1202 {Opt_data_ordered, "data=ordered"}, 1203 {Opt_data_writeback, "data=writeback"}, 1204 {Opt_data_err_abort, "data_err=abort"}, 1205 {Opt_data_err_ignore, "data_err=ignore"}, 1206 {Opt_offusrjquota, "usrjquota="}, 1207 {Opt_usrjquota, "usrjquota=%s"}, 1208 {Opt_offgrpjquota, "grpjquota="}, 1209 {Opt_grpjquota, "grpjquota=%s"}, 1210 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1211 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1212 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1213 {Opt_grpquota, "grpquota"}, 1214 {Opt_noquota, "noquota"}, 1215 {Opt_quota, "quota"}, 1216 {Opt_usrquota, "usrquota"}, 1217 {Opt_barrier, "barrier=%u"}, 1218 {Opt_barrier, "barrier"}, 1219 {Opt_nobarrier, "nobarrier"}, 1220 {Opt_i_version, "i_version"}, 1221 {Opt_stripe, "stripe=%u"}, 1222 {Opt_delalloc, "delalloc"}, 1223 {Opt_nodelalloc, "nodelalloc"}, 1224 {Opt_removed, "mblk_io_submit"}, 1225 {Opt_removed, "nomblk_io_submit"}, 1226 {Opt_block_validity, "block_validity"}, 1227 {Opt_noblock_validity, "noblock_validity"}, 1228 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1229 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1230 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1231 {Opt_auto_da_alloc, "auto_da_alloc"}, 1232 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1233 {Opt_dioread_nolock, "dioread_nolock"}, 1234 {Opt_dioread_lock, "dioread_lock"}, 1235 {Opt_discard, "discard"}, 1236 {Opt_nodiscard, "nodiscard"}, 1237 {Opt_init_itable, "init_itable=%u"}, 1238 {Opt_init_itable, "init_itable"}, 1239 {Opt_noinit_itable, "noinit_itable"}, 1240 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1241 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1242 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1243 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1244 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1245 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1246 {Opt_err, NULL}, 1247 }; 1248 1249 static ext4_fsblk_t get_sb_block(void **data) 1250 { 1251 ext4_fsblk_t sb_block; 1252 char *options = (char *) *data; 1253 1254 if (!options || strncmp(options, "sb=", 3) != 0) 1255 return 1; /* Default location */ 1256 1257 options += 3; 1258 /* TODO: use simple_strtoll with >32bit ext4 */ 1259 sb_block = simple_strtoul(options, &options, 0); 1260 if (*options && *options != ',') { 1261 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1262 (char *) *data); 1263 return 1; 1264 } 1265 if (*options == ',') 1266 options++; 1267 *data = (void *) options; 1268 1269 return sb_block; 1270 } 1271 1272 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1273 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1274 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1275 1276 #ifdef CONFIG_QUOTA 1277 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1278 { 1279 struct ext4_sb_info *sbi = EXT4_SB(sb); 1280 char *qname; 1281 int ret = -1; 1282 1283 if (sb_any_quota_loaded(sb) && 1284 !sbi->s_qf_names[qtype]) { 1285 ext4_msg(sb, KERN_ERR, 1286 "Cannot change journaled " 1287 "quota options when quota turned on"); 1288 return -1; 1289 } 1290 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) { 1291 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options " 1292 "when QUOTA feature is enabled"); 1293 return -1; 1294 } 1295 qname = match_strdup(args); 1296 if (!qname) { 1297 ext4_msg(sb, KERN_ERR, 1298 "Not enough memory for storing quotafile name"); 1299 return -1; 1300 } 1301 if (sbi->s_qf_names[qtype]) { 1302 if (strcmp(sbi->s_qf_names[qtype], qname) == 0) 1303 ret = 1; 1304 else 1305 ext4_msg(sb, KERN_ERR, 1306 "%s quota file already specified", 1307 QTYPE2NAME(qtype)); 1308 goto errout; 1309 } 1310 if (strchr(qname, '/')) { 1311 ext4_msg(sb, KERN_ERR, 1312 "quotafile must be on filesystem root"); 1313 goto errout; 1314 } 1315 sbi->s_qf_names[qtype] = qname; 1316 set_opt(sb, QUOTA); 1317 return 1; 1318 errout: 1319 kfree(qname); 1320 return ret; 1321 } 1322 1323 static int clear_qf_name(struct super_block *sb, int qtype) 1324 { 1325 1326 struct ext4_sb_info *sbi = EXT4_SB(sb); 1327 1328 if (sb_any_quota_loaded(sb) && 1329 sbi->s_qf_names[qtype]) { 1330 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1331 " when quota turned on"); 1332 return -1; 1333 } 1334 kfree(sbi->s_qf_names[qtype]); 1335 sbi->s_qf_names[qtype] = NULL; 1336 return 1; 1337 } 1338 #endif 1339 1340 #define MOPT_SET 0x0001 1341 #define MOPT_CLEAR 0x0002 1342 #define MOPT_NOSUPPORT 0x0004 1343 #define MOPT_EXPLICIT 0x0008 1344 #define MOPT_CLEAR_ERR 0x0010 1345 #define MOPT_GTE0 0x0020 1346 #ifdef CONFIG_QUOTA 1347 #define MOPT_Q 0 1348 #define MOPT_QFMT 0x0040 1349 #else 1350 #define MOPT_Q MOPT_NOSUPPORT 1351 #define MOPT_QFMT MOPT_NOSUPPORT 1352 #endif 1353 #define MOPT_DATAJ 0x0080 1354 #define MOPT_NO_EXT2 0x0100 1355 #define MOPT_NO_EXT3 0x0200 1356 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1357 #define MOPT_STRING 0x0400 1358 1359 static const struct mount_opts { 1360 int token; 1361 int mount_opt; 1362 int flags; 1363 } ext4_mount_opts[] = { 1364 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1365 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1366 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1367 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1368 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1369 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1370 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1371 MOPT_EXT4_ONLY | MOPT_SET}, 1372 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1373 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1374 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1375 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1376 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1377 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1378 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1379 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1380 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1381 MOPT_EXT4_ONLY | MOPT_SET}, 1382 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1383 EXT4_MOUNT_JOURNAL_CHECKSUM), 1384 MOPT_EXT4_ONLY | MOPT_SET}, 1385 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1386 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1387 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1388 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1389 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1390 MOPT_NO_EXT2 | MOPT_SET}, 1391 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1392 MOPT_NO_EXT2 | MOPT_CLEAR}, 1393 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1394 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1395 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1396 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1397 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1398 {Opt_commit, 0, MOPT_GTE0}, 1399 {Opt_max_batch_time, 0, MOPT_GTE0}, 1400 {Opt_min_batch_time, 0, MOPT_GTE0}, 1401 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1402 {Opt_init_itable, 0, MOPT_GTE0}, 1403 {Opt_stripe, 0, MOPT_GTE0}, 1404 {Opt_resuid, 0, MOPT_GTE0}, 1405 {Opt_resgid, 0, MOPT_GTE0}, 1406 {Opt_journal_dev, 0, MOPT_GTE0}, 1407 {Opt_journal_path, 0, MOPT_STRING}, 1408 {Opt_journal_ioprio, 0, MOPT_GTE0}, 1409 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1410 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1411 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1412 MOPT_NO_EXT2 | MOPT_DATAJ}, 1413 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1414 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1415 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1416 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1417 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1418 #else 1419 {Opt_acl, 0, MOPT_NOSUPPORT}, 1420 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1421 #endif 1422 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1423 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1424 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1425 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1426 MOPT_SET | MOPT_Q}, 1427 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1428 MOPT_SET | MOPT_Q}, 1429 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1430 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q}, 1431 {Opt_usrjquota, 0, MOPT_Q}, 1432 {Opt_grpjquota, 0, MOPT_Q}, 1433 {Opt_offusrjquota, 0, MOPT_Q}, 1434 {Opt_offgrpjquota, 0, MOPT_Q}, 1435 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1436 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1437 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1438 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1439 {Opt_err, 0, 0} 1440 }; 1441 1442 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1443 substring_t *args, unsigned long *journal_devnum, 1444 unsigned int *journal_ioprio, int is_remount) 1445 { 1446 struct ext4_sb_info *sbi = EXT4_SB(sb); 1447 const struct mount_opts *m; 1448 kuid_t uid; 1449 kgid_t gid; 1450 int arg = 0; 1451 1452 #ifdef CONFIG_QUOTA 1453 if (token == Opt_usrjquota) 1454 return set_qf_name(sb, USRQUOTA, &args[0]); 1455 else if (token == Opt_grpjquota) 1456 return set_qf_name(sb, GRPQUOTA, &args[0]); 1457 else if (token == Opt_offusrjquota) 1458 return clear_qf_name(sb, USRQUOTA); 1459 else if (token == Opt_offgrpjquota) 1460 return clear_qf_name(sb, GRPQUOTA); 1461 #endif 1462 switch (token) { 1463 case Opt_noacl: 1464 case Opt_nouser_xattr: 1465 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1466 break; 1467 case Opt_sb: 1468 return 1; /* handled by get_sb_block() */ 1469 case Opt_removed: 1470 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1471 return 1; 1472 case Opt_abort: 1473 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1474 return 1; 1475 case Opt_i_version: 1476 sb->s_flags |= MS_I_VERSION; 1477 return 1; 1478 } 1479 1480 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1481 if (token == m->token) 1482 break; 1483 1484 if (m->token == Opt_err) { 1485 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1486 "or missing value", opt); 1487 return -1; 1488 } 1489 1490 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1491 ext4_msg(sb, KERN_ERR, 1492 "Mount option \"%s\" incompatible with ext2", opt); 1493 return -1; 1494 } 1495 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1496 ext4_msg(sb, KERN_ERR, 1497 "Mount option \"%s\" incompatible with ext3", opt); 1498 return -1; 1499 } 1500 1501 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1502 return -1; 1503 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1504 return -1; 1505 if (m->flags & MOPT_EXPLICIT) 1506 set_opt2(sb, EXPLICIT_DELALLOC); 1507 if (m->flags & MOPT_CLEAR_ERR) 1508 clear_opt(sb, ERRORS_MASK); 1509 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1510 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1511 "options when quota turned on"); 1512 return -1; 1513 } 1514 1515 if (m->flags & MOPT_NOSUPPORT) { 1516 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1517 } else if (token == Opt_commit) { 1518 if (arg == 0) 1519 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1520 sbi->s_commit_interval = HZ * arg; 1521 } else if (token == Opt_max_batch_time) { 1522 if (arg == 0) 1523 arg = EXT4_DEF_MAX_BATCH_TIME; 1524 sbi->s_max_batch_time = arg; 1525 } else if (token == Opt_min_batch_time) { 1526 sbi->s_min_batch_time = arg; 1527 } else if (token == Opt_inode_readahead_blks) { 1528 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1529 ext4_msg(sb, KERN_ERR, 1530 "EXT4-fs: inode_readahead_blks must be " 1531 "0 or a power of 2 smaller than 2^31"); 1532 return -1; 1533 } 1534 sbi->s_inode_readahead_blks = arg; 1535 } else if (token == Opt_init_itable) { 1536 set_opt(sb, INIT_INODE_TABLE); 1537 if (!args->from) 1538 arg = EXT4_DEF_LI_WAIT_MULT; 1539 sbi->s_li_wait_mult = arg; 1540 } else if (token == Opt_max_dir_size_kb) { 1541 sbi->s_max_dir_size_kb = arg; 1542 } else if (token == Opt_stripe) { 1543 sbi->s_stripe = arg; 1544 } else if (token == Opt_resuid) { 1545 uid = make_kuid(current_user_ns(), arg); 1546 if (!uid_valid(uid)) { 1547 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1548 return -1; 1549 } 1550 sbi->s_resuid = uid; 1551 } else if (token == Opt_resgid) { 1552 gid = make_kgid(current_user_ns(), arg); 1553 if (!gid_valid(gid)) { 1554 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1555 return -1; 1556 } 1557 sbi->s_resgid = gid; 1558 } else if (token == Opt_journal_dev) { 1559 if (is_remount) { 1560 ext4_msg(sb, KERN_ERR, 1561 "Cannot specify journal on remount"); 1562 return -1; 1563 } 1564 *journal_devnum = arg; 1565 } else if (token == Opt_journal_path) { 1566 char *journal_path; 1567 struct inode *journal_inode; 1568 struct path path; 1569 int error; 1570 1571 if (is_remount) { 1572 ext4_msg(sb, KERN_ERR, 1573 "Cannot specify journal on remount"); 1574 return -1; 1575 } 1576 journal_path = match_strdup(&args[0]); 1577 if (!journal_path) { 1578 ext4_msg(sb, KERN_ERR, "error: could not dup " 1579 "journal device string"); 1580 return -1; 1581 } 1582 1583 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1584 if (error) { 1585 ext4_msg(sb, KERN_ERR, "error: could not find " 1586 "journal device path: error %d", error); 1587 kfree(journal_path); 1588 return -1; 1589 } 1590 1591 journal_inode = path.dentry->d_inode; 1592 if (!S_ISBLK(journal_inode->i_mode)) { 1593 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1594 "is not a block device", journal_path); 1595 path_put(&path); 1596 kfree(journal_path); 1597 return -1; 1598 } 1599 1600 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1601 path_put(&path); 1602 kfree(journal_path); 1603 } else if (token == Opt_journal_ioprio) { 1604 if (arg > 7) { 1605 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1606 " (must be 0-7)"); 1607 return -1; 1608 } 1609 *journal_ioprio = 1610 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1611 } else if (m->flags & MOPT_DATAJ) { 1612 if (is_remount) { 1613 if (!sbi->s_journal) 1614 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1615 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 1616 ext4_msg(sb, KERN_ERR, 1617 "Cannot change data mode on remount"); 1618 return -1; 1619 } 1620 } else { 1621 clear_opt(sb, DATA_FLAGS); 1622 sbi->s_mount_opt |= m->mount_opt; 1623 } 1624 #ifdef CONFIG_QUOTA 1625 } else if (m->flags & MOPT_QFMT) { 1626 if (sb_any_quota_loaded(sb) && 1627 sbi->s_jquota_fmt != m->mount_opt) { 1628 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 1629 "quota options when quota turned on"); 1630 return -1; 1631 } 1632 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 1633 EXT4_FEATURE_RO_COMPAT_QUOTA)) { 1634 ext4_msg(sb, KERN_ERR, 1635 "Cannot set journaled quota options " 1636 "when QUOTA feature is enabled"); 1637 return -1; 1638 } 1639 sbi->s_jquota_fmt = m->mount_opt; 1640 #endif 1641 } else { 1642 if (!args->from) 1643 arg = 1; 1644 if (m->flags & MOPT_CLEAR) 1645 arg = !arg; 1646 else if (unlikely(!(m->flags & MOPT_SET))) { 1647 ext4_msg(sb, KERN_WARNING, 1648 "buggy handling of option %s", opt); 1649 WARN_ON(1); 1650 return -1; 1651 } 1652 if (arg != 0) 1653 sbi->s_mount_opt |= m->mount_opt; 1654 else 1655 sbi->s_mount_opt &= ~m->mount_opt; 1656 } 1657 return 1; 1658 } 1659 1660 static int parse_options(char *options, struct super_block *sb, 1661 unsigned long *journal_devnum, 1662 unsigned int *journal_ioprio, 1663 int is_remount) 1664 { 1665 struct ext4_sb_info *sbi = EXT4_SB(sb); 1666 char *p; 1667 substring_t args[MAX_OPT_ARGS]; 1668 int token; 1669 1670 if (!options) 1671 return 1; 1672 1673 while ((p = strsep(&options, ",")) != NULL) { 1674 if (!*p) 1675 continue; 1676 /* 1677 * Initialize args struct so we know whether arg was 1678 * found; some options take optional arguments. 1679 */ 1680 args[0].to = args[0].from = NULL; 1681 token = match_token(p, tokens, args); 1682 if (handle_mount_opt(sb, p, token, args, journal_devnum, 1683 journal_ioprio, is_remount) < 0) 1684 return 0; 1685 } 1686 #ifdef CONFIG_QUOTA 1687 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 1688 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) { 1689 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA " 1690 "feature is enabled"); 1691 return 0; 1692 } 1693 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1694 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1695 clear_opt(sb, USRQUOTA); 1696 1697 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1698 clear_opt(sb, GRPQUOTA); 1699 1700 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1701 ext4_msg(sb, KERN_ERR, "old and new quota " 1702 "format mixing"); 1703 return 0; 1704 } 1705 1706 if (!sbi->s_jquota_fmt) { 1707 ext4_msg(sb, KERN_ERR, "journaled quota format " 1708 "not specified"); 1709 return 0; 1710 } 1711 } else { 1712 if (sbi->s_jquota_fmt) { 1713 ext4_msg(sb, KERN_ERR, "journaled quota format " 1714 "specified with no journaling " 1715 "enabled"); 1716 return 0; 1717 } 1718 } 1719 #endif 1720 if (test_opt(sb, DIOREAD_NOLOCK)) { 1721 int blocksize = 1722 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 1723 1724 if (blocksize < PAGE_CACHE_SIZE) { 1725 ext4_msg(sb, KERN_ERR, "can't mount with " 1726 "dioread_nolock if block size != PAGE_SIZE"); 1727 return 0; 1728 } 1729 } 1730 return 1; 1731 } 1732 1733 static inline void ext4_show_quota_options(struct seq_file *seq, 1734 struct super_block *sb) 1735 { 1736 #if defined(CONFIG_QUOTA) 1737 struct ext4_sb_info *sbi = EXT4_SB(sb); 1738 1739 if (sbi->s_jquota_fmt) { 1740 char *fmtname = ""; 1741 1742 switch (sbi->s_jquota_fmt) { 1743 case QFMT_VFS_OLD: 1744 fmtname = "vfsold"; 1745 break; 1746 case QFMT_VFS_V0: 1747 fmtname = "vfsv0"; 1748 break; 1749 case QFMT_VFS_V1: 1750 fmtname = "vfsv1"; 1751 break; 1752 } 1753 seq_printf(seq, ",jqfmt=%s", fmtname); 1754 } 1755 1756 if (sbi->s_qf_names[USRQUOTA]) 1757 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); 1758 1759 if (sbi->s_qf_names[GRPQUOTA]) 1760 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); 1761 #endif 1762 } 1763 1764 static const char *token2str(int token) 1765 { 1766 const struct match_token *t; 1767 1768 for (t = tokens; t->token != Opt_err; t++) 1769 if (t->token == token && !strchr(t->pattern, '=')) 1770 break; 1771 return t->pattern; 1772 } 1773 1774 /* 1775 * Show an option if 1776 * - it's set to a non-default value OR 1777 * - if the per-sb default is different from the global default 1778 */ 1779 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 1780 int nodefs) 1781 { 1782 struct ext4_sb_info *sbi = EXT4_SB(sb); 1783 struct ext4_super_block *es = sbi->s_es; 1784 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; 1785 const struct mount_opts *m; 1786 char sep = nodefs ? '\n' : ','; 1787 1788 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 1789 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 1790 1791 if (sbi->s_sb_block != 1) 1792 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 1793 1794 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 1795 int want_set = m->flags & MOPT_SET; 1796 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 1797 (m->flags & MOPT_CLEAR_ERR)) 1798 continue; 1799 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 1800 continue; /* skip if same as the default */ 1801 if ((want_set && 1802 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 1803 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 1804 continue; /* select Opt_noFoo vs Opt_Foo */ 1805 SEQ_OPTS_PRINT("%s", token2str(m->token)); 1806 } 1807 1808 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 1809 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 1810 SEQ_OPTS_PRINT("resuid=%u", 1811 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 1812 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 1813 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 1814 SEQ_OPTS_PRINT("resgid=%u", 1815 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 1816 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 1817 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 1818 SEQ_OPTS_PUTS("errors=remount-ro"); 1819 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 1820 SEQ_OPTS_PUTS("errors=continue"); 1821 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 1822 SEQ_OPTS_PUTS("errors=panic"); 1823 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 1824 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 1825 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 1826 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 1827 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 1828 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 1829 if (sb->s_flags & MS_I_VERSION) 1830 SEQ_OPTS_PUTS("i_version"); 1831 if (nodefs || sbi->s_stripe) 1832 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 1833 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { 1834 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 1835 SEQ_OPTS_PUTS("data=journal"); 1836 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 1837 SEQ_OPTS_PUTS("data=ordered"); 1838 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 1839 SEQ_OPTS_PUTS("data=writeback"); 1840 } 1841 if (nodefs || 1842 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 1843 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 1844 sbi->s_inode_readahead_blks); 1845 1846 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && 1847 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 1848 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 1849 if (nodefs || sbi->s_max_dir_size_kb) 1850 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 1851 1852 ext4_show_quota_options(seq, sb); 1853 return 0; 1854 } 1855 1856 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 1857 { 1858 return _ext4_show_options(seq, root->d_sb, 0); 1859 } 1860 1861 static int options_seq_show(struct seq_file *seq, void *offset) 1862 { 1863 struct super_block *sb = seq->private; 1864 int rc; 1865 1866 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw"); 1867 rc = _ext4_show_options(seq, sb, 1); 1868 seq_puts(seq, "\n"); 1869 return rc; 1870 } 1871 1872 static int options_open_fs(struct inode *inode, struct file *file) 1873 { 1874 return single_open(file, options_seq_show, PDE_DATA(inode)); 1875 } 1876 1877 static const struct file_operations ext4_seq_options_fops = { 1878 .owner = THIS_MODULE, 1879 .open = options_open_fs, 1880 .read = seq_read, 1881 .llseek = seq_lseek, 1882 .release = single_release, 1883 }; 1884 1885 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 1886 int read_only) 1887 { 1888 struct ext4_sb_info *sbi = EXT4_SB(sb); 1889 int res = 0; 1890 1891 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 1892 ext4_msg(sb, KERN_ERR, "revision level too high, " 1893 "forcing read-only mode"); 1894 res = MS_RDONLY; 1895 } 1896 if (read_only) 1897 goto done; 1898 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1899 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1900 "running e2fsck is recommended"); 1901 else if ((sbi->s_mount_state & EXT4_ERROR_FS)) 1902 ext4_msg(sb, KERN_WARNING, 1903 "warning: mounting fs with errors, " 1904 "running e2fsck is recommended"); 1905 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 1906 le16_to_cpu(es->s_mnt_count) >= 1907 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 1908 ext4_msg(sb, KERN_WARNING, 1909 "warning: maximal mount count reached, " 1910 "running e2fsck is recommended"); 1911 else if (le32_to_cpu(es->s_checkinterval) && 1912 (le32_to_cpu(es->s_lastcheck) + 1913 le32_to_cpu(es->s_checkinterval) <= get_seconds())) 1914 ext4_msg(sb, KERN_WARNING, 1915 "warning: checktime reached, " 1916 "running e2fsck is recommended"); 1917 if (!sbi->s_journal) 1918 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 1919 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1920 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1921 le16_add_cpu(&es->s_mnt_count, 1); 1922 es->s_mtime = cpu_to_le32(get_seconds()); 1923 ext4_update_dynamic_rev(sb); 1924 if (sbi->s_journal) 1925 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1926 1927 ext4_commit_super(sb, 1); 1928 done: 1929 if (test_opt(sb, DEBUG)) 1930 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1931 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1932 sb->s_blocksize, 1933 sbi->s_groups_count, 1934 EXT4_BLOCKS_PER_GROUP(sb), 1935 EXT4_INODES_PER_GROUP(sb), 1936 sbi->s_mount_opt, sbi->s_mount_opt2); 1937 1938 cleancache_init_fs(sb); 1939 return res; 1940 } 1941 1942 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 1943 { 1944 struct ext4_sb_info *sbi = EXT4_SB(sb); 1945 struct flex_groups *new_groups; 1946 int size; 1947 1948 if (!sbi->s_log_groups_per_flex) 1949 return 0; 1950 1951 size = ext4_flex_group(sbi, ngroup - 1) + 1; 1952 if (size <= sbi->s_flex_groups_allocated) 1953 return 0; 1954 1955 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 1956 new_groups = ext4_kvzalloc(size, GFP_KERNEL); 1957 if (!new_groups) { 1958 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 1959 size / (int) sizeof(struct flex_groups)); 1960 return -ENOMEM; 1961 } 1962 1963 if (sbi->s_flex_groups) { 1964 memcpy(new_groups, sbi->s_flex_groups, 1965 (sbi->s_flex_groups_allocated * 1966 sizeof(struct flex_groups))); 1967 ext4_kvfree(sbi->s_flex_groups); 1968 } 1969 sbi->s_flex_groups = new_groups; 1970 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 1971 return 0; 1972 } 1973 1974 static int ext4_fill_flex_info(struct super_block *sb) 1975 { 1976 struct ext4_sb_info *sbi = EXT4_SB(sb); 1977 struct ext4_group_desc *gdp = NULL; 1978 ext4_group_t flex_group; 1979 int i, err; 1980 1981 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 1982 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 1983 sbi->s_log_groups_per_flex = 0; 1984 return 1; 1985 } 1986 1987 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 1988 if (err) 1989 goto failed; 1990 1991 for (i = 0; i < sbi->s_groups_count; i++) { 1992 gdp = ext4_get_group_desc(sb, i, NULL); 1993 1994 flex_group = ext4_flex_group(sbi, i); 1995 atomic_add(ext4_free_inodes_count(sb, gdp), 1996 &sbi->s_flex_groups[flex_group].free_inodes); 1997 atomic64_add(ext4_free_group_clusters(sb, gdp), 1998 &sbi->s_flex_groups[flex_group].free_clusters); 1999 atomic_add(ext4_used_dirs_count(sb, gdp), 2000 &sbi->s_flex_groups[flex_group].used_dirs); 2001 } 2002 2003 return 1; 2004 failed: 2005 return 0; 2006 } 2007 2008 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, 2009 struct ext4_group_desc *gdp) 2010 { 2011 int offset; 2012 __u16 crc = 0; 2013 __le32 le_group = cpu_to_le32(block_group); 2014 2015 if ((sbi->s_es->s_feature_ro_compat & 2016 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) { 2017 /* Use new metadata_csum algorithm */ 2018 __le16 save_csum; 2019 __u32 csum32; 2020 2021 save_csum = gdp->bg_checksum; 2022 gdp->bg_checksum = 0; 2023 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2024 sizeof(le_group)); 2025 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, 2026 sbi->s_desc_size); 2027 gdp->bg_checksum = save_csum; 2028 2029 crc = csum32 & 0xFFFF; 2030 goto out; 2031 } 2032 2033 /* old crc16 code */ 2034 offset = offsetof(struct ext4_group_desc, bg_checksum); 2035 2036 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2037 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2038 crc = crc16(crc, (__u8 *)gdp, offset); 2039 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2040 /* for checksum of struct ext4_group_desc do the rest...*/ 2041 if ((sbi->s_es->s_feature_incompat & 2042 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && 2043 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2044 crc = crc16(crc, (__u8 *)gdp + offset, 2045 le16_to_cpu(sbi->s_es->s_desc_size) - 2046 offset); 2047 2048 out: 2049 return cpu_to_le16(crc); 2050 } 2051 2052 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2053 struct ext4_group_desc *gdp) 2054 { 2055 if (ext4_has_group_desc_csum(sb) && 2056 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb), 2057 block_group, gdp))) 2058 return 0; 2059 2060 return 1; 2061 } 2062 2063 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2064 struct ext4_group_desc *gdp) 2065 { 2066 if (!ext4_has_group_desc_csum(sb)) 2067 return; 2068 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp); 2069 } 2070 2071 /* Called at mount-time, super-block is locked */ 2072 static int ext4_check_descriptors(struct super_block *sb, 2073 ext4_group_t *first_not_zeroed) 2074 { 2075 struct ext4_sb_info *sbi = EXT4_SB(sb); 2076 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2077 ext4_fsblk_t last_block; 2078 ext4_fsblk_t block_bitmap; 2079 ext4_fsblk_t inode_bitmap; 2080 ext4_fsblk_t inode_table; 2081 int flexbg_flag = 0; 2082 ext4_group_t i, grp = sbi->s_groups_count; 2083 2084 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 2085 flexbg_flag = 1; 2086 2087 ext4_debug("Checking group descriptors"); 2088 2089 for (i = 0; i < sbi->s_groups_count; i++) { 2090 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2091 2092 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2093 last_block = ext4_blocks_count(sbi->s_es) - 1; 2094 else 2095 last_block = first_block + 2096 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2097 2098 if ((grp == sbi->s_groups_count) && 2099 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2100 grp = i; 2101 2102 block_bitmap = ext4_block_bitmap(sb, gdp); 2103 if (block_bitmap < first_block || block_bitmap > last_block) { 2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2105 "Block bitmap for group %u not in group " 2106 "(block %llu)!", i, block_bitmap); 2107 return 0; 2108 } 2109 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2110 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2111 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2112 "Inode bitmap for group %u not in group " 2113 "(block %llu)!", i, inode_bitmap); 2114 return 0; 2115 } 2116 inode_table = ext4_inode_table(sb, gdp); 2117 if (inode_table < first_block || 2118 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2119 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2120 "Inode table for group %u not in group " 2121 "(block %llu)!", i, inode_table); 2122 return 0; 2123 } 2124 ext4_lock_group(sb, i); 2125 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2126 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2127 "Checksum for group %u failed (%u!=%u)", 2128 i, le16_to_cpu(ext4_group_desc_csum(sbi, i, 2129 gdp)), le16_to_cpu(gdp->bg_checksum)); 2130 if (!(sb->s_flags & MS_RDONLY)) { 2131 ext4_unlock_group(sb, i); 2132 return 0; 2133 } 2134 } 2135 ext4_unlock_group(sb, i); 2136 if (!flexbg_flag) 2137 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2138 } 2139 if (NULL != first_not_zeroed) 2140 *first_not_zeroed = grp; 2141 2142 ext4_free_blocks_count_set(sbi->s_es, 2143 EXT4_C2B(sbi, ext4_count_free_clusters(sb))); 2144 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2145 return 1; 2146 } 2147 2148 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2149 * the superblock) which were deleted from all directories, but held open by 2150 * a process at the time of a crash. We walk the list and try to delete these 2151 * inodes at recovery time (only with a read-write filesystem). 2152 * 2153 * In order to keep the orphan inode chain consistent during traversal (in 2154 * case of crash during recovery), we link each inode into the superblock 2155 * orphan list_head and handle it the same way as an inode deletion during 2156 * normal operation (which journals the operations for us). 2157 * 2158 * We only do an iget() and an iput() on each inode, which is very safe if we 2159 * accidentally point at an in-use or already deleted inode. The worst that 2160 * can happen in this case is that we get a "bit already cleared" message from 2161 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2162 * e2fsck was run on this filesystem, and it must have already done the orphan 2163 * inode cleanup for us, so we can safely abort without any further action. 2164 */ 2165 static void ext4_orphan_cleanup(struct super_block *sb, 2166 struct ext4_super_block *es) 2167 { 2168 unsigned int s_flags = sb->s_flags; 2169 int nr_orphans = 0, nr_truncates = 0; 2170 #ifdef CONFIG_QUOTA 2171 int i; 2172 #endif 2173 if (!es->s_last_orphan) { 2174 jbd_debug(4, "no orphan inodes to clean up\n"); 2175 return; 2176 } 2177 2178 if (bdev_read_only(sb->s_bdev)) { 2179 ext4_msg(sb, KERN_ERR, "write access " 2180 "unavailable, skipping orphan cleanup"); 2181 return; 2182 } 2183 2184 /* Check if feature set would not allow a r/w mount */ 2185 if (!ext4_feature_set_ok(sb, 0)) { 2186 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2187 "unknown ROCOMPAT features"); 2188 return; 2189 } 2190 2191 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2192 /* don't clear list on RO mount w/ errors */ 2193 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { 2194 jbd_debug(1, "Errors on filesystem, " 2195 "clearing orphan list.\n"); 2196 es->s_last_orphan = 0; 2197 } 2198 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2199 return; 2200 } 2201 2202 if (s_flags & MS_RDONLY) { 2203 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2204 sb->s_flags &= ~MS_RDONLY; 2205 } 2206 #ifdef CONFIG_QUOTA 2207 /* Needed for iput() to work correctly and not trash data */ 2208 sb->s_flags |= MS_ACTIVE; 2209 /* Turn on quotas so that they are updated correctly */ 2210 for (i = 0; i < MAXQUOTAS; i++) { 2211 if (EXT4_SB(sb)->s_qf_names[i]) { 2212 int ret = ext4_quota_on_mount(sb, i); 2213 if (ret < 0) 2214 ext4_msg(sb, KERN_ERR, 2215 "Cannot turn on journaled " 2216 "quota: error %d", ret); 2217 } 2218 } 2219 #endif 2220 2221 while (es->s_last_orphan) { 2222 struct inode *inode; 2223 2224 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2225 if (IS_ERR(inode)) { 2226 es->s_last_orphan = 0; 2227 break; 2228 } 2229 2230 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2231 dquot_initialize(inode); 2232 if (inode->i_nlink) { 2233 if (test_opt(sb, DEBUG)) 2234 ext4_msg(sb, KERN_DEBUG, 2235 "%s: truncating inode %lu to %lld bytes", 2236 __func__, inode->i_ino, inode->i_size); 2237 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2238 inode->i_ino, inode->i_size); 2239 mutex_lock(&inode->i_mutex); 2240 truncate_inode_pages(inode->i_mapping, inode->i_size); 2241 ext4_truncate(inode); 2242 mutex_unlock(&inode->i_mutex); 2243 nr_truncates++; 2244 } else { 2245 if (test_opt(sb, DEBUG)) 2246 ext4_msg(sb, KERN_DEBUG, 2247 "%s: deleting unreferenced inode %lu", 2248 __func__, inode->i_ino); 2249 jbd_debug(2, "deleting unreferenced inode %lu\n", 2250 inode->i_ino); 2251 nr_orphans++; 2252 } 2253 iput(inode); /* The delete magic happens here! */ 2254 } 2255 2256 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2257 2258 if (nr_orphans) 2259 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2260 PLURAL(nr_orphans)); 2261 if (nr_truncates) 2262 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2263 PLURAL(nr_truncates)); 2264 #ifdef CONFIG_QUOTA 2265 /* Turn quotas off */ 2266 for (i = 0; i < MAXQUOTAS; i++) { 2267 if (sb_dqopt(sb)->files[i]) 2268 dquot_quota_off(sb, i); 2269 } 2270 #endif 2271 sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2272 } 2273 2274 /* 2275 * Maximal extent format file size. 2276 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2277 * extent format containers, within a sector_t, and within i_blocks 2278 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2279 * so that won't be a limiting factor. 2280 * 2281 * However there is other limiting factor. We do store extents in the form 2282 * of starting block and length, hence the resulting length of the extent 2283 * covering maximum file size must fit into on-disk format containers as 2284 * well. Given that length is always by 1 unit bigger than max unit (because 2285 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2286 * 2287 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2288 */ 2289 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2290 { 2291 loff_t res; 2292 loff_t upper_limit = MAX_LFS_FILESIZE; 2293 2294 /* small i_blocks in vfs inode? */ 2295 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2296 /* 2297 * CONFIG_LBDAF is not enabled implies the inode 2298 * i_block represent total blocks in 512 bytes 2299 * 32 == size of vfs inode i_blocks * 8 2300 */ 2301 upper_limit = (1LL << 32) - 1; 2302 2303 /* total blocks in file system block size */ 2304 upper_limit >>= (blkbits - 9); 2305 upper_limit <<= blkbits; 2306 } 2307 2308 /* 2309 * 32-bit extent-start container, ee_block. We lower the maxbytes 2310 * by one fs block, so ee_len can cover the extent of maximum file 2311 * size 2312 */ 2313 res = (1LL << 32) - 1; 2314 res <<= blkbits; 2315 2316 /* Sanity check against vm- & vfs- imposed limits */ 2317 if (res > upper_limit) 2318 res = upper_limit; 2319 2320 return res; 2321 } 2322 2323 /* 2324 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2325 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2326 * We need to be 1 filesystem block less than the 2^48 sector limit. 2327 */ 2328 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2329 { 2330 loff_t res = EXT4_NDIR_BLOCKS; 2331 int meta_blocks; 2332 loff_t upper_limit; 2333 /* This is calculated to be the largest file size for a dense, block 2334 * mapped file such that the file's total number of 512-byte sectors, 2335 * including data and all indirect blocks, does not exceed (2^48 - 1). 2336 * 2337 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2338 * number of 512-byte sectors of the file. 2339 */ 2340 2341 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2342 /* 2343 * !has_huge_files or CONFIG_LBDAF not enabled implies that 2344 * the inode i_block field represents total file blocks in 2345 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 2346 */ 2347 upper_limit = (1LL << 32) - 1; 2348 2349 /* total blocks in file system block size */ 2350 upper_limit >>= (bits - 9); 2351 2352 } else { 2353 /* 2354 * We use 48 bit ext4_inode i_blocks 2355 * With EXT4_HUGE_FILE_FL set the i_blocks 2356 * represent total number of blocks in 2357 * file system block size 2358 */ 2359 upper_limit = (1LL << 48) - 1; 2360 2361 } 2362 2363 /* indirect blocks */ 2364 meta_blocks = 1; 2365 /* double indirect blocks */ 2366 meta_blocks += 1 + (1LL << (bits-2)); 2367 /* tripple indirect blocks */ 2368 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2369 2370 upper_limit -= meta_blocks; 2371 upper_limit <<= bits; 2372 2373 res += 1LL << (bits-2); 2374 res += 1LL << (2*(bits-2)); 2375 res += 1LL << (3*(bits-2)); 2376 res <<= bits; 2377 if (res > upper_limit) 2378 res = upper_limit; 2379 2380 if (res > MAX_LFS_FILESIZE) 2381 res = MAX_LFS_FILESIZE; 2382 2383 return res; 2384 } 2385 2386 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2387 ext4_fsblk_t logical_sb_block, int nr) 2388 { 2389 struct ext4_sb_info *sbi = EXT4_SB(sb); 2390 ext4_group_t bg, first_meta_bg; 2391 int has_super = 0; 2392 2393 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2394 2395 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || 2396 nr < first_meta_bg) 2397 return logical_sb_block + nr + 1; 2398 bg = sbi->s_desc_per_block * nr; 2399 if (ext4_bg_has_super(sb, bg)) 2400 has_super = 1; 2401 2402 return (has_super + ext4_group_first_block_no(sb, bg)); 2403 } 2404 2405 /** 2406 * ext4_get_stripe_size: Get the stripe size. 2407 * @sbi: In memory super block info 2408 * 2409 * If we have specified it via mount option, then 2410 * use the mount option value. If the value specified at mount time is 2411 * greater than the blocks per group use the super block value. 2412 * If the super block value is greater than blocks per group return 0. 2413 * Allocator needs it be less than blocks per group. 2414 * 2415 */ 2416 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2417 { 2418 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2419 unsigned long stripe_width = 2420 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2421 int ret; 2422 2423 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2424 ret = sbi->s_stripe; 2425 else if (stripe_width <= sbi->s_blocks_per_group) 2426 ret = stripe_width; 2427 else if (stride <= sbi->s_blocks_per_group) 2428 ret = stride; 2429 else 2430 ret = 0; 2431 2432 /* 2433 * If the stripe width is 1, this makes no sense and 2434 * we set it to 0 to turn off stripe handling code. 2435 */ 2436 if (ret <= 1) 2437 ret = 0; 2438 2439 return ret; 2440 } 2441 2442 /* sysfs supprt */ 2443 2444 struct ext4_attr { 2445 struct attribute attr; 2446 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); 2447 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, 2448 const char *, size_t); 2449 union { 2450 int offset; 2451 int deprecated_val; 2452 } u; 2453 }; 2454 2455 static int parse_strtoull(const char *buf, 2456 unsigned long long max, unsigned long long *value) 2457 { 2458 int ret; 2459 2460 ret = kstrtoull(skip_spaces(buf), 0, value); 2461 if (!ret && *value > max) 2462 ret = -EINVAL; 2463 return ret; 2464 } 2465 2466 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, 2467 struct ext4_sb_info *sbi, 2468 char *buf) 2469 { 2470 return snprintf(buf, PAGE_SIZE, "%llu\n", 2471 (s64) EXT4_C2B(sbi, 2472 percpu_counter_sum(&sbi->s_dirtyclusters_counter))); 2473 } 2474 2475 static ssize_t session_write_kbytes_show(struct ext4_attr *a, 2476 struct ext4_sb_info *sbi, char *buf) 2477 { 2478 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2479 2480 if (!sb->s_bdev->bd_part) 2481 return snprintf(buf, PAGE_SIZE, "0\n"); 2482 return snprintf(buf, PAGE_SIZE, "%lu\n", 2483 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2484 sbi->s_sectors_written_start) >> 1); 2485 } 2486 2487 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, 2488 struct ext4_sb_info *sbi, char *buf) 2489 { 2490 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2491 2492 if (!sb->s_bdev->bd_part) 2493 return snprintf(buf, PAGE_SIZE, "0\n"); 2494 return snprintf(buf, PAGE_SIZE, "%llu\n", 2495 (unsigned long long)(sbi->s_kbytes_written + 2496 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2497 EXT4_SB(sb)->s_sectors_written_start) >> 1))); 2498 } 2499 2500 static ssize_t inode_readahead_blks_store(struct ext4_attr *a, 2501 struct ext4_sb_info *sbi, 2502 const char *buf, size_t count) 2503 { 2504 unsigned long t; 2505 int ret; 2506 2507 ret = kstrtoul(skip_spaces(buf), 0, &t); 2508 if (ret) 2509 return ret; 2510 2511 if (t && (!is_power_of_2(t) || t > 0x40000000)) 2512 return -EINVAL; 2513 2514 sbi->s_inode_readahead_blks = t; 2515 return count; 2516 } 2517 2518 static ssize_t sbi_ui_show(struct ext4_attr *a, 2519 struct ext4_sb_info *sbi, char *buf) 2520 { 2521 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset); 2522 2523 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 2524 } 2525 2526 static ssize_t sbi_ui_store(struct ext4_attr *a, 2527 struct ext4_sb_info *sbi, 2528 const char *buf, size_t count) 2529 { 2530 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset); 2531 unsigned long t; 2532 int ret; 2533 2534 ret = kstrtoul(skip_spaces(buf), 0, &t); 2535 if (ret) 2536 return ret; 2537 *ui = t; 2538 return count; 2539 } 2540 2541 static ssize_t reserved_clusters_show(struct ext4_attr *a, 2542 struct ext4_sb_info *sbi, char *buf) 2543 { 2544 return snprintf(buf, PAGE_SIZE, "%llu\n", 2545 (unsigned long long) atomic64_read(&sbi->s_resv_clusters)); 2546 } 2547 2548 static ssize_t reserved_clusters_store(struct ext4_attr *a, 2549 struct ext4_sb_info *sbi, 2550 const char *buf, size_t count) 2551 { 2552 unsigned long long val; 2553 int ret; 2554 2555 if (parse_strtoull(buf, -1ULL, &val)) 2556 return -EINVAL; 2557 ret = ext4_reserve_clusters(sbi, val); 2558 2559 return ret ? ret : count; 2560 } 2561 2562 static ssize_t trigger_test_error(struct ext4_attr *a, 2563 struct ext4_sb_info *sbi, 2564 const char *buf, size_t count) 2565 { 2566 int len = count; 2567 2568 if (!capable(CAP_SYS_ADMIN)) 2569 return -EPERM; 2570 2571 if (len && buf[len-1] == '\n') 2572 len--; 2573 2574 if (len) 2575 ext4_error(sbi->s_sb, "%.*s", len, buf); 2576 return count; 2577 } 2578 2579 static ssize_t sbi_deprecated_show(struct ext4_attr *a, 2580 struct ext4_sb_info *sbi, char *buf) 2581 { 2582 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val); 2583 } 2584 2585 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ 2586 static struct ext4_attr ext4_attr_##_name = { \ 2587 .attr = {.name = __stringify(_name), .mode = _mode }, \ 2588 .show = _show, \ 2589 .store = _store, \ 2590 .u = { \ 2591 .offset = offsetof(struct ext4_sb_info, _elname),\ 2592 }, \ 2593 } 2594 #define EXT4_ATTR(name, mode, show, store) \ 2595 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) 2596 2597 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) 2598 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) 2599 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) 2600 #define EXT4_RW_ATTR_SBI_UI(name, elname) \ 2601 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) 2602 #define ATTR_LIST(name) &ext4_attr_##name.attr 2603 #define EXT4_DEPRECATED_ATTR(_name, _val) \ 2604 static struct ext4_attr ext4_attr_##_name = { \ 2605 .attr = {.name = __stringify(_name), .mode = 0444 }, \ 2606 .show = sbi_deprecated_show, \ 2607 .u = { \ 2608 .deprecated_val = _val, \ 2609 }, \ 2610 } 2611 2612 EXT4_RO_ATTR(delayed_allocation_blocks); 2613 EXT4_RO_ATTR(session_write_kbytes); 2614 EXT4_RO_ATTR(lifetime_write_kbytes); 2615 EXT4_RW_ATTR(reserved_clusters); 2616 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, 2617 inode_readahead_blks_store, s_inode_readahead_blks); 2618 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); 2619 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); 2620 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); 2621 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); 2622 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); 2623 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); 2624 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); 2625 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128); 2626 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb); 2627 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error); 2628 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval); 2629 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst); 2630 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval); 2631 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst); 2632 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval); 2633 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst); 2634 2635 static struct attribute *ext4_attrs[] = { 2636 ATTR_LIST(delayed_allocation_blocks), 2637 ATTR_LIST(session_write_kbytes), 2638 ATTR_LIST(lifetime_write_kbytes), 2639 ATTR_LIST(reserved_clusters), 2640 ATTR_LIST(inode_readahead_blks), 2641 ATTR_LIST(inode_goal), 2642 ATTR_LIST(mb_stats), 2643 ATTR_LIST(mb_max_to_scan), 2644 ATTR_LIST(mb_min_to_scan), 2645 ATTR_LIST(mb_order2_req), 2646 ATTR_LIST(mb_stream_req), 2647 ATTR_LIST(mb_group_prealloc), 2648 ATTR_LIST(max_writeback_mb_bump), 2649 ATTR_LIST(extent_max_zeroout_kb), 2650 ATTR_LIST(trigger_fs_error), 2651 ATTR_LIST(err_ratelimit_interval_ms), 2652 ATTR_LIST(err_ratelimit_burst), 2653 ATTR_LIST(warning_ratelimit_interval_ms), 2654 ATTR_LIST(warning_ratelimit_burst), 2655 ATTR_LIST(msg_ratelimit_interval_ms), 2656 ATTR_LIST(msg_ratelimit_burst), 2657 NULL, 2658 }; 2659 2660 /* Features this copy of ext4 supports */ 2661 EXT4_INFO_ATTR(lazy_itable_init); 2662 EXT4_INFO_ATTR(batched_discard); 2663 EXT4_INFO_ATTR(meta_bg_resize); 2664 2665 static struct attribute *ext4_feat_attrs[] = { 2666 ATTR_LIST(lazy_itable_init), 2667 ATTR_LIST(batched_discard), 2668 ATTR_LIST(meta_bg_resize), 2669 NULL, 2670 }; 2671 2672 static ssize_t ext4_attr_show(struct kobject *kobj, 2673 struct attribute *attr, char *buf) 2674 { 2675 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2676 s_kobj); 2677 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2678 2679 return a->show ? a->show(a, sbi, buf) : 0; 2680 } 2681 2682 static ssize_t ext4_attr_store(struct kobject *kobj, 2683 struct attribute *attr, 2684 const char *buf, size_t len) 2685 { 2686 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2687 s_kobj); 2688 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2689 2690 return a->store ? a->store(a, sbi, buf, len) : 0; 2691 } 2692 2693 static void ext4_sb_release(struct kobject *kobj) 2694 { 2695 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2696 s_kobj); 2697 complete(&sbi->s_kobj_unregister); 2698 } 2699 2700 static const struct sysfs_ops ext4_attr_ops = { 2701 .show = ext4_attr_show, 2702 .store = ext4_attr_store, 2703 }; 2704 2705 static struct kobj_type ext4_ktype = { 2706 .default_attrs = ext4_attrs, 2707 .sysfs_ops = &ext4_attr_ops, 2708 .release = ext4_sb_release, 2709 }; 2710 2711 static void ext4_feat_release(struct kobject *kobj) 2712 { 2713 complete(&ext4_feat->f_kobj_unregister); 2714 } 2715 2716 static struct kobj_type ext4_feat_ktype = { 2717 .default_attrs = ext4_feat_attrs, 2718 .sysfs_ops = &ext4_attr_ops, 2719 .release = ext4_feat_release, 2720 }; 2721 2722 /* 2723 * Check whether this filesystem can be mounted based on 2724 * the features present and the RDONLY/RDWR mount requested. 2725 * Returns 1 if this filesystem can be mounted as requested, 2726 * 0 if it cannot be. 2727 */ 2728 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2729 { 2730 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { 2731 ext4_msg(sb, KERN_ERR, 2732 "Couldn't mount because of " 2733 "unsupported optional features (%x)", 2734 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2735 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2736 return 0; 2737 } 2738 2739 if (readonly) 2740 return 1; 2741 2742 /* Check that feature set is OK for a read-write mount */ 2743 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { 2744 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2745 "unsupported optional features (%x)", 2746 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2747 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2748 return 0; 2749 } 2750 /* 2751 * Large file size enabled file system can only be mounted 2752 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2753 */ 2754 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { 2755 if (sizeof(blkcnt_t) < sizeof(u64)) { 2756 ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2757 "cannot be mounted RDWR without " 2758 "CONFIG_LBDAF"); 2759 return 0; 2760 } 2761 } 2762 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) && 2763 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { 2764 ext4_msg(sb, KERN_ERR, 2765 "Can't support bigalloc feature without " 2766 "extents feature\n"); 2767 return 0; 2768 } 2769 2770 #ifndef CONFIG_QUOTA 2771 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 2772 !readonly) { 2773 ext4_msg(sb, KERN_ERR, 2774 "Filesystem with quota feature cannot be mounted RDWR " 2775 "without CONFIG_QUOTA"); 2776 return 0; 2777 } 2778 #endif /* CONFIG_QUOTA */ 2779 return 1; 2780 } 2781 2782 /* 2783 * This function is called once a day if we have errors logged 2784 * on the file system 2785 */ 2786 static void print_daily_error_info(unsigned long arg) 2787 { 2788 struct super_block *sb = (struct super_block *) arg; 2789 struct ext4_sb_info *sbi; 2790 struct ext4_super_block *es; 2791 2792 sbi = EXT4_SB(sb); 2793 es = sbi->s_es; 2794 2795 if (es->s_error_count) 2796 ext4_msg(sb, KERN_NOTICE, "error count: %u", 2797 le32_to_cpu(es->s_error_count)); 2798 if (es->s_first_error_time) { 2799 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", 2800 sb->s_id, le32_to_cpu(es->s_first_error_time), 2801 (int) sizeof(es->s_first_error_func), 2802 es->s_first_error_func, 2803 le32_to_cpu(es->s_first_error_line)); 2804 if (es->s_first_error_ino) 2805 printk(": inode %u", 2806 le32_to_cpu(es->s_first_error_ino)); 2807 if (es->s_first_error_block) 2808 printk(": block %llu", (unsigned long long) 2809 le64_to_cpu(es->s_first_error_block)); 2810 printk("\n"); 2811 } 2812 if (es->s_last_error_time) { 2813 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", 2814 sb->s_id, le32_to_cpu(es->s_last_error_time), 2815 (int) sizeof(es->s_last_error_func), 2816 es->s_last_error_func, 2817 le32_to_cpu(es->s_last_error_line)); 2818 if (es->s_last_error_ino) 2819 printk(": inode %u", 2820 le32_to_cpu(es->s_last_error_ino)); 2821 if (es->s_last_error_block) 2822 printk(": block %llu", (unsigned long long) 2823 le64_to_cpu(es->s_last_error_block)); 2824 printk("\n"); 2825 } 2826 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 2827 } 2828 2829 /* Find next suitable group and run ext4_init_inode_table */ 2830 static int ext4_run_li_request(struct ext4_li_request *elr) 2831 { 2832 struct ext4_group_desc *gdp = NULL; 2833 ext4_group_t group, ngroups; 2834 struct super_block *sb; 2835 unsigned long timeout = 0; 2836 int ret = 0; 2837 2838 sb = elr->lr_super; 2839 ngroups = EXT4_SB(sb)->s_groups_count; 2840 2841 sb_start_write(sb); 2842 for (group = elr->lr_next_group; group < ngroups; group++) { 2843 gdp = ext4_get_group_desc(sb, group, NULL); 2844 if (!gdp) { 2845 ret = 1; 2846 break; 2847 } 2848 2849 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2850 break; 2851 } 2852 2853 if (group >= ngroups) 2854 ret = 1; 2855 2856 if (!ret) { 2857 timeout = jiffies; 2858 ret = ext4_init_inode_table(sb, group, 2859 elr->lr_timeout ? 0 : 1); 2860 if (elr->lr_timeout == 0) { 2861 timeout = (jiffies - timeout) * 2862 elr->lr_sbi->s_li_wait_mult; 2863 elr->lr_timeout = timeout; 2864 } 2865 elr->lr_next_sched = jiffies + elr->lr_timeout; 2866 elr->lr_next_group = group + 1; 2867 } 2868 sb_end_write(sb); 2869 2870 return ret; 2871 } 2872 2873 /* 2874 * Remove lr_request from the list_request and free the 2875 * request structure. Should be called with li_list_mtx held 2876 */ 2877 static void ext4_remove_li_request(struct ext4_li_request *elr) 2878 { 2879 struct ext4_sb_info *sbi; 2880 2881 if (!elr) 2882 return; 2883 2884 sbi = elr->lr_sbi; 2885 2886 list_del(&elr->lr_request); 2887 sbi->s_li_request = NULL; 2888 kfree(elr); 2889 } 2890 2891 static void ext4_unregister_li_request(struct super_block *sb) 2892 { 2893 mutex_lock(&ext4_li_mtx); 2894 if (!ext4_li_info) { 2895 mutex_unlock(&ext4_li_mtx); 2896 return; 2897 } 2898 2899 mutex_lock(&ext4_li_info->li_list_mtx); 2900 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2901 mutex_unlock(&ext4_li_info->li_list_mtx); 2902 mutex_unlock(&ext4_li_mtx); 2903 } 2904 2905 static struct task_struct *ext4_lazyinit_task; 2906 2907 /* 2908 * This is the function where ext4lazyinit thread lives. It walks 2909 * through the request list searching for next scheduled filesystem. 2910 * When such a fs is found, run the lazy initialization request 2911 * (ext4_rn_li_request) and keep track of the time spend in this 2912 * function. Based on that time we compute next schedule time of 2913 * the request. When walking through the list is complete, compute 2914 * next waking time and put itself into sleep. 2915 */ 2916 static int ext4_lazyinit_thread(void *arg) 2917 { 2918 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2919 struct list_head *pos, *n; 2920 struct ext4_li_request *elr; 2921 unsigned long next_wakeup, cur; 2922 2923 BUG_ON(NULL == eli); 2924 2925 cont_thread: 2926 while (true) { 2927 next_wakeup = MAX_JIFFY_OFFSET; 2928 2929 mutex_lock(&eli->li_list_mtx); 2930 if (list_empty(&eli->li_request_list)) { 2931 mutex_unlock(&eli->li_list_mtx); 2932 goto exit_thread; 2933 } 2934 2935 list_for_each_safe(pos, n, &eli->li_request_list) { 2936 elr = list_entry(pos, struct ext4_li_request, 2937 lr_request); 2938 2939 if (time_after_eq(jiffies, elr->lr_next_sched)) { 2940 if (ext4_run_li_request(elr) != 0) { 2941 /* error, remove the lazy_init job */ 2942 ext4_remove_li_request(elr); 2943 continue; 2944 } 2945 } 2946 2947 if (time_before(elr->lr_next_sched, next_wakeup)) 2948 next_wakeup = elr->lr_next_sched; 2949 } 2950 mutex_unlock(&eli->li_list_mtx); 2951 2952 try_to_freeze(); 2953 2954 cur = jiffies; 2955 if ((time_after_eq(cur, next_wakeup)) || 2956 (MAX_JIFFY_OFFSET == next_wakeup)) { 2957 cond_resched(); 2958 continue; 2959 } 2960 2961 schedule_timeout_interruptible(next_wakeup - cur); 2962 2963 if (kthread_should_stop()) { 2964 ext4_clear_request_list(); 2965 goto exit_thread; 2966 } 2967 } 2968 2969 exit_thread: 2970 /* 2971 * It looks like the request list is empty, but we need 2972 * to check it under the li_list_mtx lock, to prevent any 2973 * additions into it, and of course we should lock ext4_li_mtx 2974 * to atomically free the list and ext4_li_info, because at 2975 * this point another ext4 filesystem could be registering 2976 * new one. 2977 */ 2978 mutex_lock(&ext4_li_mtx); 2979 mutex_lock(&eli->li_list_mtx); 2980 if (!list_empty(&eli->li_request_list)) { 2981 mutex_unlock(&eli->li_list_mtx); 2982 mutex_unlock(&ext4_li_mtx); 2983 goto cont_thread; 2984 } 2985 mutex_unlock(&eli->li_list_mtx); 2986 kfree(ext4_li_info); 2987 ext4_li_info = NULL; 2988 mutex_unlock(&ext4_li_mtx); 2989 2990 return 0; 2991 } 2992 2993 static void ext4_clear_request_list(void) 2994 { 2995 struct list_head *pos, *n; 2996 struct ext4_li_request *elr; 2997 2998 mutex_lock(&ext4_li_info->li_list_mtx); 2999 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3000 elr = list_entry(pos, struct ext4_li_request, 3001 lr_request); 3002 ext4_remove_li_request(elr); 3003 } 3004 mutex_unlock(&ext4_li_info->li_list_mtx); 3005 } 3006 3007 static int ext4_run_lazyinit_thread(void) 3008 { 3009 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3010 ext4_li_info, "ext4lazyinit"); 3011 if (IS_ERR(ext4_lazyinit_task)) { 3012 int err = PTR_ERR(ext4_lazyinit_task); 3013 ext4_clear_request_list(); 3014 kfree(ext4_li_info); 3015 ext4_li_info = NULL; 3016 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3017 "initialization thread\n", 3018 err); 3019 return err; 3020 } 3021 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3022 return 0; 3023 } 3024 3025 /* 3026 * Check whether it make sense to run itable init. thread or not. 3027 * If there is at least one uninitialized inode table, return 3028 * corresponding group number, else the loop goes through all 3029 * groups and return total number of groups. 3030 */ 3031 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3032 { 3033 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3034 struct ext4_group_desc *gdp = NULL; 3035 3036 for (group = 0; group < ngroups; group++) { 3037 gdp = ext4_get_group_desc(sb, group, NULL); 3038 if (!gdp) 3039 continue; 3040 3041 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3042 break; 3043 } 3044 3045 return group; 3046 } 3047 3048 static int ext4_li_info_new(void) 3049 { 3050 struct ext4_lazy_init *eli = NULL; 3051 3052 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3053 if (!eli) 3054 return -ENOMEM; 3055 3056 INIT_LIST_HEAD(&eli->li_request_list); 3057 mutex_init(&eli->li_list_mtx); 3058 3059 eli->li_state |= EXT4_LAZYINIT_QUIT; 3060 3061 ext4_li_info = eli; 3062 3063 return 0; 3064 } 3065 3066 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3067 ext4_group_t start) 3068 { 3069 struct ext4_sb_info *sbi = EXT4_SB(sb); 3070 struct ext4_li_request *elr; 3071 3072 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3073 if (!elr) 3074 return NULL; 3075 3076 elr->lr_super = sb; 3077 elr->lr_sbi = sbi; 3078 elr->lr_next_group = start; 3079 3080 /* 3081 * Randomize first schedule time of the request to 3082 * spread the inode table initialization requests 3083 * better. 3084 */ 3085 elr->lr_next_sched = jiffies + (prandom_u32() % 3086 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3087 return elr; 3088 } 3089 3090 int ext4_register_li_request(struct super_block *sb, 3091 ext4_group_t first_not_zeroed) 3092 { 3093 struct ext4_sb_info *sbi = EXT4_SB(sb); 3094 struct ext4_li_request *elr = NULL; 3095 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3096 int ret = 0; 3097 3098 mutex_lock(&ext4_li_mtx); 3099 if (sbi->s_li_request != NULL) { 3100 /* 3101 * Reset timeout so it can be computed again, because 3102 * s_li_wait_mult might have changed. 3103 */ 3104 sbi->s_li_request->lr_timeout = 0; 3105 goto out; 3106 } 3107 3108 if (first_not_zeroed == ngroups || 3109 (sb->s_flags & MS_RDONLY) || 3110 !test_opt(sb, INIT_INODE_TABLE)) 3111 goto out; 3112 3113 elr = ext4_li_request_new(sb, first_not_zeroed); 3114 if (!elr) { 3115 ret = -ENOMEM; 3116 goto out; 3117 } 3118 3119 if (NULL == ext4_li_info) { 3120 ret = ext4_li_info_new(); 3121 if (ret) 3122 goto out; 3123 } 3124 3125 mutex_lock(&ext4_li_info->li_list_mtx); 3126 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3127 mutex_unlock(&ext4_li_info->li_list_mtx); 3128 3129 sbi->s_li_request = elr; 3130 /* 3131 * set elr to NULL here since it has been inserted to 3132 * the request_list and the removal and free of it is 3133 * handled by ext4_clear_request_list from now on. 3134 */ 3135 elr = NULL; 3136 3137 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3138 ret = ext4_run_lazyinit_thread(); 3139 if (ret) 3140 goto out; 3141 } 3142 out: 3143 mutex_unlock(&ext4_li_mtx); 3144 if (ret) 3145 kfree(elr); 3146 return ret; 3147 } 3148 3149 /* 3150 * We do not need to lock anything since this is called on 3151 * module unload. 3152 */ 3153 static void ext4_destroy_lazyinit_thread(void) 3154 { 3155 /* 3156 * If thread exited earlier 3157 * there's nothing to be done. 3158 */ 3159 if (!ext4_li_info || !ext4_lazyinit_task) 3160 return; 3161 3162 kthread_stop(ext4_lazyinit_task); 3163 } 3164 3165 static int set_journal_csum_feature_set(struct super_block *sb) 3166 { 3167 int ret = 1; 3168 int compat, incompat; 3169 struct ext4_sb_info *sbi = EXT4_SB(sb); 3170 3171 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3172 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 3173 /* journal checksum v2 */ 3174 compat = 0; 3175 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2; 3176 } else { 3177 /* journal checksum v1 */ 3178 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3179 incompat = 0; 3180 } 3181 3182 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3183 ret = jbd2_journal_set_features(sbi->s_journal, 3184 compat, 0, 3185 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3186 incompat); 3187 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3188 ret = jbd2_journal_set_features(sbi->s_journal, 3189 compat, 0, 3190 incompat); 3191 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3192 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3193 } else { 3194 jbd2_journal_clear_features(sbi->s_journal, 3195 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3196 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3197 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3198 } 3199 3200 return ret; 3201 } 3202 3203 /* 3204 * Note: calculating the overhead so we can be compatible with 3205 * historical BSD practice is quite difficult in the face of 3206 * clusters/bigalloc. This is because multiple metadata blocks from 3207 * different block group can end up in the same allocation cluster. 3208 * Calculating the exact overhead in the face of clustered allocation 3209 * requires either O(all block bitmaps) in memory or O(number of block 3210 * groups**2) in time. We will still calculate the superblock for 3211 * older file systems --- and if we come across with a bigalloc file 3212 * system with zero in s_overhead_clusters the estimate will be close to 3213 * correct especially for very large cluster sizes --- but for newer 3214 * file systems, it's better to calculate this figure once at mkfs 3215 * time, and store it in the superblock. If the superblock value is 3216 * present (even for non-bigalloc file systems), we will use it. 3217 */ 3218 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3219 char *buf) 3220 { 3221 struct ext4_sb_info *sbi = EXT4_SB(sb); 3222 struct ext4_group_desc *gdp; 3223 ext4_fsblk_t first_block, last_block, b; 3224 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3225 int s, j, count = 0; 3226 3227 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC)) 3228 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3229 sbi->s_itb_per_group + 2); 3230 3231 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3232 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3233 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3234 for (i = 0; i < ngroups; i++) { 3235 gdp = ext4_get_group_desc(sb, i, NULL); 3236 b = ext4_block_bitmap(sb, gdp); 3237 if (b >= first_block && b <= last_block) { 3238 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3239 count++; 3240 } 3241 b = ext4_inode_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_table(sb, gdp); 3247 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3248 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3249 int c = EXT4_B2C(sbi, b - first_block); 3250 ext4_set_bit(c, buf); 3251 count++; 3252 } 3253 if (i != grp) 3254 continue; 3255 s = 0; 3256 if (ext4_bg_has_super(sb, grp)) { 3257 ext4_set_bit(s++, buf); 3258 count++; 3259 } 3260 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { 3261 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3262 count++; 3263 } 3264 } 3265 if (!count) 3266 return 0; 3267 return EXT4_CLUSTERS_PER_GROUP(sb) - 3268 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3269 } 3270 3271 /* 3272 * Compute the overhead and stash it in sbi->s_overhead 3273 */ 3274 int ext4_calculate_overhead(struct super_block *sb) 3275 { 3276 struct ext4_sb_info *sbi = EXT4_SB(sb); 3277 struct ext4_super_block *es = sbi->s_es; 3278 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3279 ext4_fsblk_t overhead = 0; 3280 char *buf = (char *) get_zeroed_page(GFP_KERNEL); 3281 3282 if (!buf) 3283 return -ENOMEM; 3284 3285 /* 3286 * Compute the overhead (FS structures). This is constant 3287 * for a given filesystem unless the number of block groups 3288 * changes so we cache the previous value until it does. 3289 */ 3290 3291 /* 3292 * All of the blocks before first_data_block are overhead 3293 */ 3294 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3295 3296 /* 3297 * Add the overhead found in each block group 3298 */ 3299 for (i = 0; i < ngroups; i++) { 3300 int blks; 3301 3302 blks = count_overhead(sb, i, buf); 3303 overhead += blks; 3304 if (blks) 3305 memset(buf, 0, PAGE_SIZE); 3306 cond_resched(); 3307 } 3308 /* Add the journal blocks as well */ 3309 if (sbi->s_journal) 3310 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3311 3312 sbi->s_overhead = overhead; 3313 smp_wmb(); 3314 free_page((unsigned long) buf); 3315 return 0; 3316 } 3317 3318 3319 static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi) 3320 { 3321 ext4_fsblk_t resv_clusters; 3322 3323 /* 3324 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3325 * This should cover the situations where we can not afford to run 3326 * out of space like for example punch hole, or converting 3327 * uninitialized extents in delalloc path. In most cases such 3328 * allocation would require 1, or 2 blocks, higher numbers are 3329 * very rare. 3330 */ 3331 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits; 3332 3333 do_div(resv_clusters, 50); 3334 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3335 3336 return resv_clusters; 3337 } 3338 3339 3340 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count) 3341 { 3342 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >> 3343 sbi->s_cluster_bits; 3344 3345 if (count >= clusters) 3346 return -EINVAL; 3347 3348 atomic64_set(&sbi->s_resv_clusters, count); 3349 return 0; 3350 } 3351 3352 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3353 { 3354 char *orig_data = kstrdup(data, GFP_KERNEL); 3355 struct buffer_head *bh; 3356 struct ext4_super_block *es = NULL; 3357 struct ext4_sb_info *sbi; 3358 ext4_fsblk_t block; 3359 ext4_fsblk_t sb_block = get_sb_block(&data); 3360 ext4_fsblk_t logical_sb_block; 3361 unsigned long offset = 0; 3362 unsigned long journal_devnum = 0; 3363 unsigned long def_mount_opts; 3364 struct inode *root; 3365 char *cp; 3366 const char *descr; 3367 int ret = -ENOMEM; 3368 int blocksize, clustersize; 3369 unsigned int db_count; 3370 unsigned int i; 3371 int needs_recovery, has_huge_files, has_bigalloc; 3372 __u64 blocks_count; 3373 int err = 0; 3374 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3375 ext4_group_t first_not_zeroed; 3376 3377 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3378 if (!sbi) 3379 goto out_free_orig; 3380 3381 sbi->s_blockgroup_lock = 3382 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3383 if (!sbi->s_blockgroup_lock) { 3384 kfree(sbi); 3385 goto out_free_orig; 3386 } 3387 sb->s_fs_info = sbi; 3388 sbi->s_sb = sb; 3389 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3390 sbi->s_sb_block = sb_block; 3391 if (sb->s_bdev->bd_part) 3392 sbi->s_sectors_written_start = 3393 part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3394 3395 /* Cleanup superblock name */ 3396 for (cp = sb->s_id; (cp = strchr(cp, '/'));) 3397 *cp = '!'; 3398 3399 /* -EINVAL is default */ 3400 ret = -EINVAL; 3401 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3402 if (!blocksize) { 3403 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3404 goto out_fail; 3405 } 3406 3407 /* 3408 * The ext4 superblock will not be buffer aligned for other than 1kB 3409 * block sizes. We need to calculate the offset from buffer start. 3410 */ 3411 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3412 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3413 offset = do_div(logical_sb_block, blocksize); 3414 } else { 3415 logical_sb_block = sb_block; 3416 } 3417 3418 if (!(bh = sb_bread(sb, logical_sb_block))) { 3419 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3420 goto out_fail; 3421 } 3422 /* 3423 * Note: s_es must be initialized as soon as possible because 3424 * some ext4 macro-instructions depend on its value 3425 */ 3426 es = (struct ext4_super_block *) (bh->b_data + offset); 3427 sbi->s_es = es; 3428 sb->s_magic = le16_to_cpu(es->s_magic); 3429 if (sb->s_magic != EXT4_SUPER_MAGIC) 3430 goto cantfind_ext4; 3431 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3432 3433 /* Warn if metadata_csum and gdt_csum are both set. */ 3434 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3435 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 3436 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) 3437 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are " 3438 "redundant flags; please run fsck."); 3439 3440 /* Check for a known checksum algorithm */ 3441 if (!ext4_verify_csum_type(sb, es)) { 3442 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3443 "unknown checksum algorithm."); 3444 silent = 1; 3445 goto cantfind_ext4; 3446 } 3447 3448 /* Load the checksum driver */ 3449 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3450 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 3451 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3452 if (IS_ERR(sbi->s_chksum_driver)) { 3453 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3454 ret = PTR_ERR(sbi->s_chksum_driver); 3455 sbi->s_chksum_driver = NULL; 3456 goto failed_mount; 3457 } 3458 } 3459 3460 /* Check superblock checksum */ 3461 if (!ext4_superblock_csum_verify(sb, es)) { 3462 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3463 "invalid superblock checksum. Run e2fsck?"); 3464 silent = 1; 3465 goto cantfind_ext4; 3466 } 3467 3468 /* Precompute checksum seed for all metadata */ 3469 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3470 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 3471 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3472 sizeof(es->s_uuid)); 3473 3474 /* Set defaults before we parse the mount options */ 3475 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3476 set_opt(sb, INIT_INODE_TABLE); 3477 if (def_mount_opts & EXT4_DEFM_DEBUG) 3478 set_opt(sb, DEBUG); 3479 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3480 set_opt(sb, GRPID); 3481 if (def_mount_opts & EXT4_DEFM_UID16) 3482 set_opt(sb, NO_UID32); 3483 /* xattr user namespace & acls are now defaulted on */ 3484 set_opt(sb, XATTR_USER); 3485 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3486 set_opt(sb, POSIX_ACL); 3487 #endif 3488 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3489 set_opt(sb, JOURNAL_DATA); 3490 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3491 set_opt(sb, ORDERED_DATA); 3492 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3493 set_opt(sb, WRITEBACK_DATA); 3494 3495 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3496 set_opt(sb, ERRORS_PANIC); 3497 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3498 set_opt(sb, ERRORS_CONT); 3499 else 3500 set_opt(sb, ERRORS_RO); 3501 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) 3502 set_opt(sb, BLOCK_VALIDITY); 3503 if (def_mount_opts & EXT4_DEFM_DISCARD) 3504 set_opt(sb, DISCARD); 3505 3506 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3507 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3508 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3509 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3510 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3511 3512 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3513 set_opt(sb, BARRIER); 3514 3515 /* 3516 * enable delayed allocation by default 3517 * Use -o nodelalloc to turn it off 3518 */ 3519 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3520 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3521 set_opt(sb, DELALLOC); 3522 3523 /* 3524 * set default s_li_wait_mult for lazyinit, for the case there is 3525 * no mount option specified. 3526 */ 3527 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3528 3529 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3530 &journal_devnum, &journal_ioprio, 0)) { 3531 ext4_msg(sb, KERN_WARNING, 3532 "failed to parse options in superblock: %s", 3533 sbi->s_es->s_mount_opts); 3534 } 3535 sbi->s_def_mount_opt = sbi->s_mount_opt; 3536 if (!parse_options((char *) data, sb, &journal_devnum, 3537 &journal_ioprio, 0)) 3538 goto failed_mount; 3539 3540 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3541 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3542 "with data=journal disables delayed " 3543 "allocation and O_DIRECT support!\n"); 3544 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3545 ext4_msg(sb, KERN_ERR, "can't mount with " 3546 "both data=journal and delalloc"); 3547 goto failed_mount; 3548 } 3549 if (test_opt(sb, DIOREAD_NOLOCK)) { 3550 ext4_msg(sb, KERN_ERR, "can't mount with " 3551 "both data=journal and dioread_nolock"); 3552 goto failed_mount; 3553 } 3554 if (test_opt(sb, DELALLOC)) 3555 clear_opt(sb, DELALLOC); 3556 } 3557 3558 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3559 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3560 3561 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3562 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || 3563 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || 3564 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) 3565 ext4_msg(sb, KERN_WARNING, 3566 "feature flags set on rev 0 fs, " 3567 "running e2fsck is recommended"); 3568 3569 if (IS_EXT2_SB(sb)) { 3570 if (ext2_feature_set_ok(sb)) 3571 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3572 "using the ext4 subsystem"); 3573 else { 3574 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3575 "to feature incompatibilities"); 3576 goto failed_mount; 3577 } 3578 } 3579 3580 if (IS_EXT3_SB(sb)) { 3581 if (ext3_feature_set_ok(sb)) 3582 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3583 "using the ext4 subsystem"); 3584 else { 3585 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3586 "to feature incompatibilities"); 3587 goto failed_mount; 3588 } 3589 } 3590 3591 /* 3592 * Check feature flags regardless of the revision level, since we 3593 * previously didn't change the revision level when setting the flags, 3594 * so there is a chance incompat flags are set on a rev 0 filesystem. 3595 */ 3596 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3597 goto failed_mount; 3598 3599 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3600 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3601 blocksize > EXT4_MAX_BLOCK_SIZE) { 3602 ext4_msg(sb, KERN_ERR, 3603 "Unsupported filesystem blocksize %d", blocksize); 3604 goto failed_mount; 3605 } 3606 3607 if (sb->s_blocksize != blocksize) { 3608 /* Validate the filesystem blocksize */ 3609 if (!sb_set_blocksize(sb, blocksize)) { 3610 ext4_msg(sb, KERN_ERR, "bad block size %d", 3611 blocksize); 3612 goto failed_mount; 3613 } 3614 3615 brelse(bh); 3616 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3617 offset = do_div(logical_sb_block, blocksize); 3618 bh = sb_bread(sb, logical_sb_block); 3619 if (!bh) { 3620 ext4_msg(sb, KERN_ERR, 3621 "Can't read superblock on 2nd try"); 3622 goto failed_mount; 3623 } 3624 es = (struct ext4_super_block *)(bh->b_data + offset); 3625 sbi->s_es = es; 3626 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3627 ext4_msg(sb, KERN_ERR, 3628 "Magic mismatch, very weird!"); 3629 goto failed_mount; 3630 } 3631 } 3632 3633 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3634 EXT4_FEATURE_RO_COMPAT_HUGE_FILE); 3635 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3636 has_huge_files); 3637 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3638 3639 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3640 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3641 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3642 } else { 3643 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3644 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3645 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3646 (!is_power_of_2(sbi->s_inode_size)) || 3647 (sbi->s_inode_size > blocksize)) { 3648 ext4_msg(sb, KERN_ERR, 3649 "unsupported inode size: %d", 3650 sbi->s_inode_size); 3651 goto failed_mount; 3652 } 3653 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3654 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3655 } 3656 3657 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3658 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { 3659 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 3660 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3661 !is_power_of_2(sbi->s_desc_size)) { 3662 ext4_msg(sb, KERN_ERR, 3663 "unsupported descriptor size %lu", 3664 sbi->s_desc_size); 3665 goto failed_mount; 3666 } 3667 } else 3668 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 3669 3670 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3671 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3672 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3673 goto cantfind_ext4; 3674 3675 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3676 if (sbi->s_inodes_per_block == 0) 3677 goto cantfind_ext4; 3678 sbi->s_itb_per_group = sbi->s_inodes_per_group / 3679 sbi->s_inodes_per_block; 3680 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3681 sbi->s_sbh = bh; 3682 sbi->s_mount_state = le16_to_cpu(es->s_state); 3683 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3684 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 3685 3686 for (i = 0; i < 4; i++) 3687 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3688 sbi->s_def_hash_version = es->s_def_hash_version; 3689 i = le32_to_cpu(es->s_flags); 3690 if (i & EXT2_FLAGS_UNSIGNED_HASH) 3691 sbi->s_hash_unsigned = 3; 3692 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3693 #ifdef __CHAR_UNSIGNED__ 3694 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3695 sbi->s_hash_unsigned = 3; 3696 #else 3697 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3698 #endif 3699 } 3700 3701 /* Handle clustersize */ 3702 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3703 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3704 EXT4_FEATURE_RO_COMPAT_BIGALLOC); 3705 if (has_bigalloc) { 3706 if (clustersize < blocksize) { 3707 ext4_msg(sb, KERN_ERR, 3708 "cluster size (%d) smaller than " 3709 "block size (%d)", clustersize, blocksize); 3710 goto failed_mount; 3711 } 3712 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3713 le32_to_cpu(es->s_log_block_size); 3714 sbi->s_clusters_per_group = 3715 le32_to_cpu(es->s_clusters_per_group); 3716 if (sbi->s_clusters_per_group > blocksize * 8) { 3717 ext4_msg(sb, KERN_ERR, 3718 "#clusters per group too big: %lu", 3719 sbi->s_clusters_per_group); 3720 goto failed_mount; 3721 } 3722 if (sbi->s_blocks_per_group != 3723 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3724 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3725 "clusters per group (%lu) inconsistent", 3726 sbi->s_blocks_per_group, 3727 sbi->s_clusters_per_group); 3728 goto failed_mount; 3729 } 3730 } else { 3731 if (clustersize != blocksize) { 3732 ext4_warning(sb, "fragment/cluster size (%d) != " 3733 "block size (%d)", clustersize, 3734 blocksize); 3735 clustersize = blocksize; 3736 } 3737 if (sbi->s_blocks_per_group > blocksize * 8) { 3738 ext4_msg(sb, KERN_ERR, 3739 "#blocks per group too big: %lu", 3740 sbi->s_blocks_per_group); 3741 goto failed_mount; 3742 } 3743 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3744 sbi->s_cluster_bits = 0; 3745 } 3746 sbi->s_cluster_ratio = clustersize / blocksize; 3747 3748 if (sbi->s_inodes_per_group > blocksize * 8) { 3749 ext4_msg(sb, KERN_ERR, 3750 "#inodes per group too big: %lu", 3751 sbi->s_inodes_per_group); 3752 goto failed_mount; 3753 } 3754 3755 /* Do we have standard group size of clustersize * 8 blocks ? */ 3756 if (sbi->s_blocks_per_group == clustersize << 3) 3757 set_opt2(sb, STD_GROUP_SIZE); 3758 3759 /* 3760 * Test whether we have more sectors than will fit in sector_t, 3761 * and whether the max offset is addressable by the page cache. 3762 */ 3763 err = generic_check_addressable(sb->s_blocksize_bits, 3764 ext4_blocks_count(es)); 3765 if (err) { 3766 ext4_msg(sb, KERN_ERR, "filesystem" 3767 " too large to mount safely on this system"); 3768 if (sizeof(sector_t) < 8) 3769 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3770 goto failed_mount; 3771 } 3772 3773 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3774 goto cantfind_ext4; 3775 3776 /* check blocks count against device size */ 3777 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 3778 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3779 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3780 "exceeds size of device (%llu blocks)", 3781 ext4_blocks_count(es), blocks_count); 3782 goto failed_mount; 3783 } 3784 3785 /* 3786 * It makes no sense for the first data block to be beyond the end 3787 * of the filesystem. 3788 */ 3789 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3790 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 3791 "block %u is beyond end of filesystem (%llu)", 3792 le32_to_cpu(es->s_first_data_block), 3793 ext4_blocks_count(es)); 3794 goto failed_mount; 3795 } 3796 blocks_count = (ext4_blocks_count(es) - 3797 le32_to_cpu(es->s_first_data_block) + 3798 EXT4_BLOCKS_PER_GROUP(sb) - 1); 3799 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 3800 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3801 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 3802 "(block count %llu, first data block %u, " 3803 "blocks per group %lu)", sbi->s_groups_count, 3804 ext4_blocks_count(es), 3805 le32_to_cpu(es->s_first_data_block), 3806 EXT4_BLOCKS_PER_GROUP(sb)); 3807 goto failed_mount; 3808 } 3809 sbi->s_groups_count = blocks_count; 3810 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3811 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3812 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3813 EXT4_DESC_PER_BLOCK(sb); 3814 sbi->s_group_desc = ext4_kvmalloc(db_count * 3815 sizeof(struct buffer_head *), 3816 GFP_KERNEL); 3817 if (sbi->s_group_desc == NULL) { 3818 ext4_msg(sb, KERN_ERR, "not enough memory"); 3819 ret = -ENOMEM; 3820 goto failed_mount; 3821 } 3822 3823 if (ext4_proc_root) 3824 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); 3825 3826 if (sbi->s_proc) 3827 proc_create_data("options", S_IRUGO, sbi->s_proc, 3828 &ext4_seq_options_fops, sb); 3829 3830 bgl_lock_init(sbi->s_blockgroup_lock); 3831 3832 for (i = 0; i < db_count; i++) { 3833 block = descriptor_loc(sb, logical_sb_block, i); 3834 sbi->s_group_desc[i] = sb_bread(sb, block); 3835 if (!sbi->s_group_desc[i]) { 3836 ext4_msg(sb, KERN_ERR, 3837 "can't read group descriptor %d", i); 3838 db_count = i; 3839 goto failed_mount2; 3840 } 3841 } 3842 if (!ext4_check_descriptors(sb, &first_not_zeroed)) { 3843 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3844 goto failed_mount2; 3845 } 3846 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 3847 if (!ext4_fill_flex_info(sb)) { 3848 ext4_msg(sb, KERN_ERR, 3849 "unable to initialize " 3850 "flex_bg meta info!"); 3851 goto failed_mount2; 3852 } 3853 3854 sbi->s_gdb_count = db_count; 3855 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3856 spin_lock_init(&sbi->s_next_gen_lock); 3857 3858 init_timer(&sbi->s_err_report); 3859 sbi->s_err_report.function = print_daily_error_info; 3860 sbi->s_err_report.data = (unsigned long) sb; 3861 3862 /* Register extent status tree shrinker */ 3863 ext4_es_register_shrinker(sbi); 3864 3865 err = percpu_counter_init(&sbi->s_freeclusters_counter, 3866 ext4_count_free_clusters(sb)); 3867 if (!err) { 3868 err = percpu_counter_init(&sbi->s_freeinodes_counter, 3869 ext4_count_free_inodes(sb)); 3870 } 3871 if (!err) { 3872 err = percpu_counter_init(&sbi->s_dirs_counter, 3873 ext4_count_dirs(sb)); 3874 } 3875 if (!err) { 3876 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0); 3877 } 3878 if (!err) { 3879 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0); 3880 } 3881 if (err) { 3882 ext4_msg(sb, KERN_ERR, "insufficient memory"); 3883 goto failed_mount3; 3884 } 3885 3886 sbi->s_stripe = ext4_get_stripe_size(sbi); 3887 sbi->s_extent_max_zeroout_kb = 32; 3888 3889 /* 3890 * set up enough so that it can read an inode 3891 */ 3892 if (!test_opt(sb, NOLOAD) && 3893 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 3894 sb->s_op = &ext4_sops; 3895 else 3896 sb->s_op = &ext4_nojournal_sops; 3897 sb->s_export_op = &ext4_export_ops; 3898 sb->s_xattr = ext4_xattr_handlers; 3899 #ifdef CONFIG_QUOTA 3900 sb->dq_op = &ext4_quota_operations; 3901 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 3902 sb->s_qcop = &ext4_qctl_sysfile_operations; 3903 else 3904 sb->s_qcop = &ext4_qctl_operations; 3905 #endif 3906 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3907 3908 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 3909 mutex_init(&sbi->s_orphan_lock); 3910 3911 sb->s_root = NULL; 3912 3913 needs_recovery = (es->s_last_orphan != 0 || 3914 EXT4_HAS_INCOMPAT_FEATURE(sb, 3915 EXT4_FEATURE_INCOMPAT_RECOVER)); 3916 3917 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) && 3918 !(sb->s_flags & MS_RDONLY)) 3919 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 3920 goto failed_mount3; 3921 3922 /* 3923 * The first inode we look at is the journal inode. Don't try 3924 * root first: it may be modified in the journal! 3925 */ 3926 if (!test_opt(sb, NOLOAD) && 3927 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 3928 if (ext4_load_journal(sb, es, journal_devnum)) 3929 goto failed_mount3; 3930 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 3931 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 3932 ext4_msg(sb, KERN_ERR, "required journal recovery " 3933 "suppressed and not mounted read-only"); 3934 goto failed_mount_wq; 3935 } else { 3936 clear_opt(sb, DATA_FLAGS); 3937 sbi->s_journal = NULL; 3938 needs_recovery = 0; 3939 goto no_journal; 3940 } 3941 3942 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) && 3943 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3944 JBD2_FEATURE_INCOMPAT_64BIT)) { 3945 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3946 goto failed_mount_wq; 3947 } 3948 3949 if (!set_journal_csum_feature_set(sb)) { 3950 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 3951 "feature set"); 3952 goto failed_mount_wq; 3953 } 3954 3955 /* We have now updated the journal if required, so we can 3956 * validate the data journaling mode. */ 3957 switch (test_opt(sb, DATA_FLAGS)) { 3958 case 0: 3959 /* No mode set, assume a default based on the journal 3960 * capabilities: ORDERED_DATA if the journal can 3961 * cope, else JOURNAL_DATA 3962 */ 3963 if (jbd2_journal_check_available_features 3964 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 3965 set_opt(sb, ORDERED_DATA); 3966 else 3967 set_opt(sb, JOURNAL_DATA); 3968 break; 3969 3970 case EXT4_MOUNT_ORDERED_DATA: 3971 case EXT4_MOUNT_WRITEBACK_DATA: 3972 if (!jbd2_journal_check_available_features 3973 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 3974 ext4_msg(sb, KERN_ERR, "Journal does not support " 3975 "requested data journaling mode"); 3976 goto failed_mount_wq; 3977 } 3978 default: 3979 break; 3980 } 3981 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 3982 3983 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 3984 3985 /* 3986 * The journal may have updated the bg summary counts, so we 3987 * need to update the global counters. 3988 */ 3989 percpu_counter_set(&sbi->s_freeclusters_counter, 3990 ext4_count_free_clusters(sb)); 3991 percpu_counter_set(&sbi->s_freeinodes_counter, 3992 ext4_count_free_inodes(sb)); 3993 percpu_counter_set(&sbi->s_dirs_counter, 3994 ext4_count_dirs(sb)); 3995 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0); 3996 3997 no_journal: 3998 /* 3999 * Get the # of file system overhead blocks from the 4000 * superblock if present. 4001 */ 4002 if (es->s_overhead_clusters) 4003 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4004 else { 4005 err = ext4_calculate_overhead(sb); 4006 if (err) 4007 goto failed_mount_wq; 4008 } 4009 4010 /* 4011 * The maximum number of concurrent works can be high and 4012 * concurrency isn't really necessary. Limit it to 1. 4013 */ 4014 EXT4_SB(sb)->rsv_conversion_wq = 4015 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4016 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4017 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4018 ret = -ENOMEM; 4019 goto failed_mount4; 4020 } 4021 4022 /* 4023 * The jbd2_journal_load will have done any necessary log recovery, 4024 * so we can safely mount the rest of the filesystem now. 4025 */ 4026 4027 root = ext4_iget(sb, EXT4_ROOT_INO); 4028 if (IS_ERR(root)) { 4029 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4030 ret = PTR_ERR(root); 4031 root = NULL; 4032 goto failed_mount4; 4033 } 4034 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4035 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4036 iput(root); 4037 goto failed_mount4; 4038 } 4039 sb->s_root = d_make_root(root); 4040 if (!sb->s_root) { 4041 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4042 ret = -ENOMEM; 4043 goto failed_mount4; 4044 } 4045 4046 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY)) 4047 sb->s_flags |= MS_RDONLY; 4048 4049 /* determine the minimum size of new large inodes, if present */ 4050 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4051 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4052 EXT4_GOOD_OLD_INODE_SIZE; 4053 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 4054 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { 4055 if (sbi->s_want_extra_isize < 4056 le16_to_cpu(es->s_want_extra_isize)) 4057 sbi->s_want_extra_isize = 4058 le16_to_cpu(es->s_want_extra_isize); 4059 if (sbi->s_want_extra_isize < 4060 le16_to_cpu(es->s_min_extra_isize)) 4061 sbi->s_want_extra_isize = 4062 le16_to_cpu(es->s_min_extra_isize); 4063 } 4064 } 4065 /* Check if enough inode space is available */ 4066 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 4067 sbi->s_inode_size) { 4068 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4069 EXT4_GOOD_OLD_INODE_SIZE; 4070 ext4_msg(sb, KERN_INFO, "required extra inode space not" 4071 "available"); 4072 } 4073 4074 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi)); 4075 if (err) { 4076 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for " 4077 "reserved pool", ext4_calculate_resv_clusters(sbi)); 4078 goto failed_mount4a; 4079 } 4080 4081 err = ext4_setup_system_zone(sb); 4082 if (err) { 4083 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4084 "zone (%d)", err); 4085 goto failed_mount4a; 4086 } 4087 4088 ext4_ext_init(sb); 4089 err = ext4_mb_init(sb); 4090 if (err) { 4091 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4092 err); 4093 goto failed_mount5; 4094 } 4095 4096 err = ext4_register_li_request(sb, first_not_zeroed); 4097 if (err) 4098 goto failed_mount6; 4099 4100 sbi->s_kobj.kset = ext4_kset; 4101 init_completion(&sbi->s_kobj_unregister); 4102 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 4103 "%s", sb->s_id); 4104 if (err) 4105 goto failed_mount7; 4106 4107 #ifdef CONFIG_QUOTA 4108 /* Enable quota usage during mount. */ 4109 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 4110 !(sb->s_flags & MS_RDONLY)) { 4111 err = ext4_enable_quotas(sb); 4112 if (err) 4113 goto failed_mount8; 4114 } 4115 #endif /* CONFIG_QUOTA */ 4116 4117 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4118 ext4_orphan_cleanup(sb, es); 4119 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4120 if (needs_recovery) { 4121 ext4_msg(sb, KERN_INFO, "recovery complete"); 4122 ext4_mark_recovery_complete(sb, es); 4123 } 4124 if (EXT4_SB(sb)->s_journal) { 4125 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4126 descr = " journalled data mode"; 4127 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4128 descr = " ordered data mode"; 4129 else 4130 descr = " writeback data mode"; 4131 } else 4132 descr = "out journal"; 4133 4134 if (test_opt(sb, DISCARD)) { 4135 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4136 if (!blk_queue_discard(q)) 4137 ext4_msg(sb, KERN_WARNING, 4138 "mounting with \"discard\" option, but " 4139 "the device does not support discard"); 4140 } 4141 4142 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4143 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 4144 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4145 4146 if (es->s_error_count) 4147 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4148 4149 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4150 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4151 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4152 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4153 4154 kfree(orig_data); 4155 return 0; 4156 4157 cantfind_ext4: 4158 if (!silent) 4159 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4160 goto failed_mount; 4161 4162 #ifdef CONFIG_QUOTA 4163 failed_mount8: 4164 kobject_del(&sbi->s_kobj); 4165 #endif 4166 failed_mount7: 4167 ext4_unregister_li_request(sb); 4168 failed_mount6: 4169 ext4_mb_release(sb); 4170 failed_mount5: 4171 ext4_ext_release(sb); 4172 ext4_release_system_zone(sb); 4173 failed_mount4a: 4174 dput(sb->s_root); 4175 sb->s_root = NULL; 4176 failed_mount4: 4177 ext4_msg(sb, KERN_ERR, "mount failed"); 4178 if (EXT4_SB(sb)->rsv_conversion_wq) 4179 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4180 failed_mount_wq: 4181 if (sbi->s_journal) { 4182 jbd2_journal_destroy(sbi->s_journal); 4183 sbi->s_journal = NULL; 4184 } 4185 failed_mount3: 4186 ext4_es_unregister_shrinker(sbi); 4187 del_timer(&sbi->s_err_report); 4188 if (sbi->s_flex_groups) 4189 ext4_kvfree(sbi->s_flex_groups); 4190 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4191 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4192 percpu_counter_destroy(&sbi->s_dirs_counter); 4193 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4194 percpu_counter_destroy(&sbi->s_extent_cache_cnt); 4195 if (sbi->s_mmp_tsk) 4196 kthread_stop(sbi->s_mmp_tsk); 4197 failed_mount2: 4198 for (i = 0; i < db_count; i++) 4199 brelse(sbi->s_group_desc[i]); 4200 ext4_kvfree(sbi->s_group_desc); 4201 failed_mount: 4202 if (sbi->s_chksum_driver) 4203 crypto_free_shash(sbi->s_chksum_driver); 4204 if (sbi->s_proc) { 4205 remove_proc_entry("options", sbi->s_proc); 4206 remove_proc_entry(sb->s_id, ext4_proc_root); 4207 } 4208 #ifdef CONFIG_QUOTA 4209 for (i = 0; i < MAXQUOTAS; i++) 4210 kfree(sbi->s_qf_names[i]); 4211 #endif 4212 ext4_blkdev_remove(sbi); 4213 brelse(bh); 4214 out_fail: 4215 sb->s_fs_info = NULL; 4216 kfree(sbi->s_blockgroup_lock); 4217 kfree(sbi); 4218 out_free_orig: 4219 kfree(orig_data); 4220 return err ? err : ret; 4221 } 4222 4223 /* 4224 * Setup any per-fs journal parameters now. We'll do this both on 4225 * initial mount, once the journal has been initialised but before we've 4226 * done any recovery; and again on any subsequent remount. 4227 */ 4228 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4229 { 4230 struct ext4_sb_info *sbi = EXT4_SB(sb); 4231 4232 journal->j_commit_interval = sbi->s_commit_interval; 4233 journal->j_min_batch_time = sbi->s_min_batch_time; 4234 journal->j_max_batch_time = sbi->s_max_batch_time; 4235 4236 write_lock(&journal->j_state_lock); 4237 if (test_opt(sb, BARRIER)) 4238 journal->j_flags |= JBD2_BARRIER; 4239 else 4240 journal->j_flags &= ~JBD2_BARRIER; 4241 if (test_opt(sb, DATA_ERR_ABORT)) 4242 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4243 else 4244 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4245 write_unlock(&journal->j_state_lock); 4246 } 4247 4248 static journal_t *ext4_get_journal(struct super_block *sb, 4249 unsigned int journal_inum) 4250 { 4251 struct inode *journal_inode; 4252 journal_t *journal; 4253 4254 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4255 4256 /* First, test for the existence of a valid inode on disk. Bad 4257 * things happen if we iget() an unused inode, as the subsequent 4258 * iput() will try to delete it. */ 4259 4260 journal_inode = ext4_iget(sb, journal_inum); 4261 if (IS_ERR(journal_inode)) { 4262 ext4_msg(sb, KERN_ERR, "no journal found"); 4263 return NULL; 4264 } 4265 if (!journal_inode->i_nlink) { 4266 make_bad_inode(journal_inode); 4267 iput(journal_inode); 4268 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4269 return NULL; 4270 } 4271 4272 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4273 journal_inode, journal_inode->i_size); 4274 if (!S_ISREG(journal_inode->i_mode)) { 4275 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4276 iput(journal_inode); 4277 return NULL; 4278 } 4279 4280 journal = jbd2_journal_init_inode(journal_inode); 4281 if (!journal) { 4282 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4283 iput(journal_inode); 4284 return NULL; 4285 } 4286 journal->j_private = sb; 4287 ext4_init_journal_params(sb, journal); 4288 return journal; 4289 } 4290 4291 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4292 dev_t j_dev) 4293 { 4294 struct buffer_head *bh; 4295 journal_t *journal; 4296 ext4_fsblk_t start; 4297 ext4_fsblk_t len; 4298 int hblock, blocksize; 4299 ext4_fsblk_t sb_block; 4300 unsigned long offset; 4301 struct ext4_super_block *es; 4302 struct block_device *bdev; 4303 4304 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4305 4306 bdev = ext4_blkdev_get(j_dev, sb); 4307 if (bdev == NULL) 4308 return NULL; 4309 4310 blocksize = sb->s_blocksize; 4311 hblock = bdev_logical_block_size(bdev); 4312 if (blocksize < hblock) { 4313 ext4_msg(sb, KERN_ERR, 4314 "blocksize too small for journal device"); 4315 goto out_bdev; 4316 } 4317 4318 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4319 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4320 set_blocksize(bdev, blocksize); 4321 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4322 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4323 "external journal"); 4324 goto out_bdev; 4325 } 4326 4327 es = (struct ext4_super_block *) (bh->b_data + offset); 4328 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4329 !(le32_to_cpu(es->s_feature_incompat) & 4330 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4331 ext4_msg(sb, KERN_ERR, "external journal has " 4332 "bad superblock"); 4333 brelse(bh); 4334 goto out_bdev; 4335 } 4336 4337 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4338 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4339 brelse(bh); 4340 goto out_bdev; 4341 } 4342 4343 len = ext4_blocks_count(es); 4344 start = sb_block + 1; 4345 brelse(bh); /* we're done with the superblock */ 4346 4347 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4348 start, len, blocksize); 4349 if (!journal) { 4350 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4351 goto out_bdev; 4352 } 4353 journal->j_private = sb; 4354 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4355 wait_on_buffer(journal->j_sb_buffer); 4356 if (!buffer_uptodate(journal->j_sb_buffer)) { 4357 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4358 goto out_journal; 4359 } 4360 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4361 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4362 "user (unsupported) - %d", 4363 be32_to_cpu(journal->j_superblock->s_nr_users)); 4364 goto out_journal; 4365 } 4366 EXT4_SB(sb)->journal_bdev = bdev; 4367 ext4_init_journal_params(sb, journal); 4368 return journal; 4369 4370 out_journal: 4371 jbd2_journal_destroy(journal); 4372 out_bdev: 4373 ext4_blkdev_put(bdev); 4374 return NULL; 4375 } 4376 4377 static int ext4_load_journal(struct super_block *sb, 4378 struct ext4_super_block *es, 4379 unsigned long journal_devnum) 4380 { 4381 journal_t *journal; 4382 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4383 dev_t journal_dev; 4384 int err = 0; 4385 int really_read_only; 4386 4387 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4388 4389 if (journal_devnum && 4390 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4391 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4392 "numbers have changed"); 4393 journal_dev = new_decode_dev(journal_devnum); 4394 } else 4395 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4396 4397 really_read_only = bdev_read_only(sb->s_bdev); 4398 4399 /* 4400 * Are we loading a blank journal or performing recovery after a 4401 * crash? For recovery, we need to check in advance whether we 4402 * can get read-write access to the device. 4403 */ 4404 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 4405 if (sb->s_flags & MS_RDONLY) { 4406 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4407 "required on readonly filesystem"); 4408 if (really_read_only) { 4409 ext4_msg(sb, KERN_ERR, "write access " 4410 "unavailable, cannot proceed"); 4411 return -EROFS; 4412 } 4413 ext4_msg(sb, KERN_INFO, "write access will " 4414 "be enabled during recovery"); 4415 } 4416 } 4417 4418 if (journal_inum && journal_dev) { 4419 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4420 "and inode journals!"); 4421 return -EINVAL; 4422 } 4423 4424 if (journal_inum) { 4425 if (!(journal = ext4_get_journal(sb, journal_inum))) 4426 return -EINVAL; 4427 } else { 4428 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4429 return -EINVAL; 4430 } 4431 4432 if (!(journal->j_flags & JBD2_BARRIER)) 4433 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4434 4435 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) 4436 err = jbd2_journal_wipe(journal, !really_read_only); 4437 if (!err) { 4438 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4439 if (save) 4440 memcpy(save, ((char *) es) + 4441 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4442 err = jbd2_journal_load(journal); 4443 if (save) 4444 memcpy(((char *) es) + EXT4_S_ERR_START, 4445 save, EXT4_S_ERR_LEN); 4446 kfree(save); 4447 } 4448 4449 if (err) { 4450 ext4_msg(sb, KERN_ERR, "error loading journal"); 4451 jbd2_journal_destroy(journal); 4452 return err; 4453 } 4454 4455 EXT4_SB(sb)->s_journal = journal; 4456 ext4_clear_journal_err(sb, es); 4457 4458 if (!really_read_only && journal_devnum && 4459 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4460 es->s_journal_dev = cpu_to_le32(journal_devnum); 4461 4462 /* Make sure we flush the recovery flag to disk. */ 4463 ext4_commit_super(sb, 1); 4464 } 4465 4466 return 0; 4467 } 4468 4469 static int ext4_commit_super(struct super_block *sb, int sync) 4470 { 4471 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4472 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4473 int error = 0; 4474 4475 if (!sbh || block_device_ejected(sb)) 4476 return error; 4477 if (buffer_write_io_error(sbh)) { 4478 /* 4479 * Oh, dear. A previous attempt to write the 4480 * superblock failed. This could happen because the 4481 * USB device was yanked out. Or it could happen to 4482 * be a transient write error and maybe the block will 4483 * be remapped. Nothing we can do but to retry the 4484 * write and hope for the best. 4485 */ 4486 ext4_msg(sb, KERN_ERR, "previous I/O error to " 4487 "superblock detected"); 4488 clear_buffer_write_io_error(sbh); 4489 set_buffer_uptodate(sbh); 4490 } 4491 /* 4492 * If the file system is mounted read-only, don't update the 4493 * superblock write time. This avoids updating the superblock 4494 * write time when we are mounting the root file system 4495 * read/only but we need to replay the journal; at that point, 4496 * for people who are east of GMT and who make their clock 4497 * tick in localtime for Windows bug-for-bug compatibility, 4498 * the clock is set in the future, and this will cause e2fsck 4499 * to complain and force a full file system check. 4500 */ 4501 if (!(sb->s_flags & MS_RDONLY)) 4502 es->s_wtime = cpu_to_le32(get_seconds()); 4503 if (sb->s_bdev->bd_part) 4504 es->s_kbytes_written = 4505 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4506 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4507 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4508 else 4509 es->s_kbytes_written = 4510 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4511 ext4_free_blocks_count_set(es, 4512 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4513 &EXT4_SB(sb)->s_freeclusters_counter))); 4514 es->s_free_inodes_count = 4515 cpu_to_le32(percpu_counter_sum_positive( 4516 &EXT4_SB(sb)->s_freeinodes_counter)); 4517 BUFFER_TRACE(sbh, "marking dirty"); 4518 ext4_superblock_csum_set(sb); 4519 mark_buffer_dirty(sbh); 4520 if (sync) { 4521 error = sync_dirty_buffer(sbh); 4522 if (error) 4523 return error; 4524 4525 error = buffer_write_io_error(sbh); 4526 if (error) { 4527 ext4_msg(sb, KERN_ERR, "I/O error while writing " 4528 "superblock"); 4529 clear_buffer_write_io_error(sbh); 4530 set_buffer_uptodate(sbh); 4531 } 4532 } 4533 return error; 4534 } 4535 4536 /* 4537 * Have we just finished recovery? If so, and if we are mounting (or 4538 * remounting) the filesystem readonly, then we will end up with a 4539 * consistent fs on disk. Record that fact. 4540 */ 4541 static void ext4_mark_recovery_complete(struct super_block *sb, 4542 struct ext4_super_block *es) 4543 { 4544 journal_t *journal = EXT4_SB(sb)->s_journal; 4545 4546 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 4547 BUG_ON(journal != NULL); 4548 return; 4549 } 4550 jbd2_journal_lock_updates(journal); 4551 if (jbd2_journal_flush(journal) < 0) 4552 goto out; 4553 4554 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && 4555 sb->s_flags & MS_RDONLY) { 4556 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4557 ext4_commit_super(sb, 1); 4558 } 4559 4560 out: 4561 jbd2_journal_unlock_updates(journal); 4562 } 4563 4564 /* 4565 * If we are mounting (or read-write remounting) a filesystem whose journal 4566 * has recorded an error from a previous lifetime, move that error to the 4567 * main filesystem now. 4568 */ 4569 static void ext4_clear_journal_err(struct super_block *sb, 4570 struct ext4_super_block *es) 4571 { 4572 journal_t *journal; 4573 int j_errno; 4574 const char *errstr; 4575 4576 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4577 4578 journal = EXT4_SB(sb)->s_journal; 4579 4580 /* 4581 * Now check for any error status which may have been recorded in the 4582 * journal by a prior ext4_error() or ext4_abort() 4583 */ 4584 4585 j_errno = jbd2_journal_errno(journal); 4586 if (j_errno) { 4587 char nbuf[16]; 4588 4589 errstr = ext4_decode_error(sb, j_errno, nbuf); 4590 ext4_warning(sb, "Filesystem error recorded " 4591 "from previous mount: %s", errstr); 4592 ext4_warning(sb, "Marking fs in need of filesystem check."); 4593 4594 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4595 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4596 ext4_commit_super(sb, 1); 4597 4598 jbd2_journal_clear_err(journal); 4599 jbd2_journal_update_sb_errno(journal); 4600 } 4601 } 4602 4603 /* 4604 * Force the running and committing transactions to commit, 4605 * and wait on the commit. 4606 */ 4607 int ext4_force_commit(struct super_block *sb) 4608 { 4609 journal_t *journal; 4610 4611 if (sb->s_flags & MS_RDONLY) 4612 return 0; 4613 4614 journal = EXT4_SB(sb)->s_journal; 4615 return ext4_journal_force_commit(journal); 4616 } 4617 4618 static int ext4_sync_fs(struct super_block *sb, int wait) 4619 { 4620 int ret = 0; 4621 tid_t target; 4622 bool needs_barrier = false; 4623 struct ext4_sb_info *sbi = EXT4_SB(sb); 4624 4625 trace_ext4_sync_fs(sb, wait); 4626 flush_workqueue(sbi->rsv_conversion_wq); 4627 /* 4628 * Writeback quota in non-journalled quota case - journalled quota has 4629 * no dirty dquots 4630 */ 4631 dquot_writeback_dquots(sb, -1); 4632 /* 4633 * Data writeback is possible w/o journal transaction, so barrier must 4634 * being sent at the end of the function. But we can skip it if 4635 * transaction_commit will do it for us. 4636 */ 4637 target = jbd2_get_latest_transaction(sbi->s_journal); 4638 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 4639 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 4640 needs_barrier = true; 4641 4642 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4643 if (wait) 4644 ret = jbd2_log_wait_commit(sbi->s_journal, target); 4645 } 4646 if (needs_barrier) { 4647 int err; 4648 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 4649 if (!ret) 4650 ret = err; 4651 } 4652 4653 return ret; 4654 } 4655 4656 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait) 4657 { 4658 int ret = 0; 4659 4660 trace_ext4_sync_fs(sb, wait); 4661 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4662 dquot_writeback_dquots(sb, -1); 4663 if (wait && test_opt(sb, BARRIER)) 4664 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 4665 4666 return ret; 4667 } 4668 4669 /* 4670 * LVM calls this function before a (read-only) snapshot is created. This 4671 * gives us a chance to flush the journal completely and mark the fs clean. 4672 * 4673 * Note that only this function cannot bring a filesystem to be in a clean 4674 * state independently. It relies on upper layer to stop all data & metadata 4675 * modifications. 4676 */ 4677 static int ext4_freeze(struct super_block *sb) 4678 { 4679 int error = 0; 4680 journal_t *journal; 4681 4682 if (sb->s_flags & MS_RDONLY) 4683 return 0; 4684 4685 journal = EXT4_SB(sb)->s_journal; 4686 4687 /* Now we set up the journal barrier. */ 4688 jbd2_journal_lock_updates(journal); 4689 4690 /* 4691 * Don't clear the needs_recovery flag if we failed to flush 4692 * the journal. 4693 */ 4694 error = jbd2_journal_flush(journal); 4695 if (error < 0) 4696 goto out; 4697 4698 /* Journal blocked and flushed, clear needs_recovery flag. */ 4699 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4700 error = ext4_commit_super(sb, 1); 4701 out: 4702 /* we rely on upper layer to stop further updates */ 4703 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 4704 return error; 4705 } 4706 4707 /* 4708 * Called by LVM after the snapshot is done. We need to reset the RECOVER 4709 * flag here, even though the filesystem is not technically dirty yet. 4710 */ 4711 static int ext4_unfreeze(struct super_block *sb) 4712 { 4713 if (sb->s_flags & MS_RDONLY) 4714 return 0; 4715 4716 /* Reset the needs_recovery flag before the fs is unlocked. */ 4717 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4718 ext4_commit_super(sb, 1); 4719 return 0; 4720 } 4721 4722 /* 4723 * Structure to save mount options for ext4_remount's benefit 4724 */ 4725 struct ext4_mount_options { 4726 unsigned long s_mount_opt; 4727 unsigned long s_mount_opt2; 4728 kuid_t s_resuid; 4729 kgid_t s_resgid; 4730 unsigned long s_commit_interval; 4731 u32 s_min_batch_time, s_max_batch_time; 4732 #ifdef CONFIG_QUOTA 4733 int s_jquota_fmt; 4734 char *s_qf_names[MAXQUOTAS]; 4735 #endif 4736 }; 4737 4738 static int ext4_remount(struct super_block *sb, int *flags, char *data) 4739 { 4740 struct ext4_super_block *es; 4741 struct ext4_sb_info *sbi = EXT4_SB(sb); 4742 unsigned long old_sb_flags; 4743 struct ext4_mount_options old_opts; 4744 int enable_quota = 0; 4745 ext4_group_t g; 4746 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4747 int err = 0; 4748 #ifdef CONFIG_QUOTA 4749 int i, j; 4750 #endif 4751 char *orig_data = kstrdup(data, GFP_KERNEL); 4752 4753 /* Store the original options */ 4754 old_sb_flags = sb->s_flags; 4755 old_opts.s_mount_opt = sbi->s_mount_opt; 4756 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4757 old_opts.s_resuid = sbi->s_resuid; 4758 old_opts.s_resgid = sbi->s_resgid; 4759 old_opts.s_commit_interval = sbi->s_commit_interval; 4760 old_opts.s_min_batch_time = sbi->s_min_batch_time; 4761 old_opts.s_max_batch_time = sbi->s_max_batch_time; 4762 #ifdef CONFIG_QUOTA 4763 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4764 for (i = 0; i < MAXQUOTAS; i++) 4765 if (sbi->s_qf_names[i]) { 4766 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], 4767 GFP_KERNEL); 4768 if (!old_opts.s_qf_names[i]) { 4769 for (j = 0; j < i; j++) 4770 kfree(old_opts.s_qf_names[j]); 4771 kfree(orig_data); 4772 return -ENOMEM; 4773 } 4774 } else 4775 old_opts.s_qf_names[i] = NULL; 4776 #endif 4777 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4778 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4779 4780 /* 4781 * Allow the "check" option to be passed as a remount option. 4782 */ 4783 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 4784 err = -EINVAL; 4785 goto restore_opts; 4786 } 4787 4788 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4789 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4790 ext4_msg(sb, KERN_ERR, "can't mount with " 4791 "both data=journal and delalloc"); 4792 err = -EINVAL; 4793 goto restore_opts; 4794 } 4795 if (test_opt(sb, DIOREAD_NOLOCK)) { 4796 ext4_msg(sb, KERN_ERR, "can't mount with " 4797 "both data=journal and dioread_nolock"); 4798 err = -EINVAL; 4799 goto restore_opts; 4800 } 4801 } 4802 4803 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4804 ext4_abort(sb, "Abort forced by user"); 4805 4806 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4807 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4808 4809 es = sbi->s_es; 4810 4811 if (sbi->s_journal) { 4812 ext4_init_journal_params(sb, sbi->s_journal); 4813 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4814 } 4815 4816 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 4817 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4818 err = -EROFS; 4819 goto restore_opts; 4820 } 4821 4822 if (*flags & MS_RDONLY) { 4823 err = dquot_suspend(sb, -1); 4824 if (err < 0) 4825 goto restore_opts; 4826 4827 /* 4828 * First of all, the unconditional stuff we have to do 4829 * to disable replay of the journal when we next remount 4830 */ 4831 sb->s_flags |= MS_RDONLY; 4832 4833 /* 4834 * OK, test if we are remounting a valid rw partition 4835 * readonly, and if so set the rdonly flag and then 4836 * mark the partition as valid again. 4837 */ 4838 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4839 (sbi->s_mount_state & EXT4_VALID_FS)) 4840 es->s_state = cpu_to_le16(sbi->s_mount_state); 4841 4842 if (sbi->s_journal) 4843 ext4_mark_recovery_complete(sb, es); 4844 } else { 4845 /* Make sure we can mount this feature set readwrite */ 4846 if (!ext4_feature_set_ok(sb, 0)) { 4847 err = -EROFS; 4848 goto restore_opts; 4849 } 4850 /* 4851 * Make sure the group descriptor checksums 4852 * are sane. If they aren't, refuse to remount r/w. 4853 */ 4854 for (g = 0; g < sbi->s_groups_count; g++) { 4855 struct ext4_group_desc *gdp = 4856 ext4_get_group_desc(sb, g, NULL); 4857 4858 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 4859 ext4_msg(sb, KERN_ERR, 4860 "ext4_remount: Checksum for group %u failed (%u!=%u)", 4861 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), 4862 le16_to_cpu(gdp->bg_checksum)); 4863 err = -EINVAL; 4864 goto restore_opts; 4865 } 4866 } 4867 4868 /* 4869 * If we have an unprocessed orphan list hanging 4870 * around from a previously readonly bdev mount, 4871 * require a full umount/remount for now. 4872 */ 4873 if (es->s_last_orphan) { 4874 ext4_msg(sb, KERN_WARNING, "Couldn't " 4875 "remount RDWR because of unprocessed " 4876 "orphan inode list. Please " 4877 "umount/remount instead"); 4878 err = -EINVAL; 4879 goto restore_opts; 4880 } 4881 4882 /* 4883 * Mounting a RDONLY partition read-write, so reread 4884 * and store the current valid flag. (It may have 4885 * been changed by e2fsck since we originally mounted 4886 * the partition.) 4887 */ 4888 if (sbi->s_journal) 4889 ext4_clear_journal_err(sb, es); 4890 sbi->s_mount_state = le16_to_cpu(es->s_state); 4891 if (!ext4_setup_super(sb, es, 0)) 4892 sb->s_flags &= ~MS_RDONLY; 4893 if (EXT4_HAS_INCOMPAT_FEATURE(sb, 4894 EXT4_FEATURE_INCOMPAT_MMP)) 4895 if (ext4_multi_mount_protect(sb, 4896 le64_to_cpu(es->s_mmp_block))) { 4897 err = -EROFS; 4898 goto restore_opts; 4899 } 4900 enable_quota = 1; 4901 } 4902 } 4903 4904 /* 4905 * Reinitialize lazy itable initialization thread based on 4906 * current settings 4907 */ 4908 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4909 ext4_unregister_li_request(sb); 4910 else { 4911 ext4_group_t first_not_zeroed; 4912 first_not_zeroed = ext4_has_uninit_itable(sb); 4913 ext4_register_li_request(sb, first_not_zeroed); 4914 } 4915 4916 ext4_setup_system_zone(sb); 4917 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY)) 4918 ext4_commit_super(sb, 1); 4919 4920 #ifdef CONFIG_QUOTA 4921 /* Release old quota file names */ 4922 for (i = 0; i < MAXQUOTAS; i++) 4923 kfree(old_opts.s_qf_names[i]); 4924 if (enable_quota) { 4925 if (sb_any_quota_suspended(sb)) 4926 dquot_resume(sb, -1); 4927 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 4928 EXT4_FEATURE_RO_COMPAT_QUOTA)) { 4929 err = ext4_enable_quotas(sb); 4930 if (err) 4931 goto restore_opts; 4932 } 4933 } 4934 #endif 4935 4936 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4937 kfree(orig_data); 4938 return 0; 4939 4940 restore_opts: 4941 sb->s_flags = old_sb_flags; 4942 sbi->s_mount_opt = old_opts.s_mount_opt; 4943 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 4944 sbi->s_resuid = old_opts.s_resuid; 4945 sbi->s_resgid = old_opts.s_resgid; 4946 sbi->s_commit_interval = old_opts.s_commit_interval; 4947 sbi->s_min_batch_time = old_opts.s_min_batch_time; 4948 sbi->s_max_batch_time = old_opts.s_max_batch_time; 4949 #ifdef CONFIG_QUOTA 4950 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 4951 for (i = 0; i < MAXQUOTAS; i++) { 4952 kfree(sbi->s_qf_names[i]); 4953 sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 4954 } 4955 #endif 4956 kfree(orig_data); 4957 return err; 4958 } 4959 4960 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 4961 { 4962 struct super_block *sb = dentry->d_sb; 4963 struct ext4_sb_info *sbi = EXT4_SB(sb); 4964 struct ext4_super_block *es = sbi->s_es; 4965 ext4_fsblk_t overhead = 0, resv_blocks; 4966 u64 fsid; 4967 s64 bfree; 4968 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 4969 4970 if (!test_opt(sb, MINIX_DF)) 4971 overhead = sbi->s_overhead; 4972 4973 buf->f_type = EXT4_SUPER_MAGIC; 4974 buf->f_bsize = sb->s_blocksize; 4975 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 4976 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 4977 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 4978 /* prevent underflow in case that few free space is available */ 4979 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 4980 buf->f_bavail = buf->f_bfree - 4981 (ext4_r_blocks_count(es) + resv_blocks); 4982 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 4983 buf->f_bavail = 0; 4984 buf->f_files = le32_to_cpu(es->s_inodes_count); 4985 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 4986 buf->f_namelen = EXT4_NAME_LEN; 4987 fsid = le64_to_cpup((void *)es->s_uuid) ^ 4988 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 4989 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 4990 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 4991 4992 return 0; 4993 } 4994 4995 /* Helper function for writing quotas on sync - we need to start transaction 4996 * before quota file is locked for write. Otherwise the are possible deadlocks: 4997 * Process 1 Process 2 4998 * ext4_create() quota_sync() 4999 * jbd2_journal_start() write_dquot() 5000 * dquot_initialize() down(dqio_mutex) 5001 * down(dqio_mutex) jbd2_journal_start() 5002 * 5003 */ 5004 5005 #ifdef CONFIG_QUOTA 5006 5007 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5008 { 5009 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5010 } 5011 5012 static int ext4_write_dquot(struct dquot *dquot) 5013 { 5014 int ret, err; 5015 handle_t *handle; 5016 struct inode *inode; 5017 5018 inode = dquot_to_inode(dquot); 5019 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5020 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5021 if (IS_ERR(handle)) 5022 return PTR_ERR(handle); 5023 ret = dquot_commit(dquot); 5024 err = ext4_journal_stop(handle); 5025 if (!ret) 5026 ret = err; 5027 return ret; 5028 } 5029 5030 static int ext4_acquire_dquot(struct dquot *dquot) 5031 { 5032 int ret, err; 5033 handle_t *handle; 5034 5035 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5036 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5037 if (IS_ERR(handle)) 5038 return PTR_ERR(handle); 5039 ret = dquot_acquire(dquot); 5040 err = ext4_journal_stop(handle); 5041 if (!ret) 5042 ret = err; 5043 return ret; 5044 } 5045 5046 static int ext4_release_dquot(struct dquot *dquot) 5047 { 5048 int ret, err; 5049 handle_t *handle; 5050 5051 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5052 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5053 if (IS_ERR(handle)) { 5054 /* Release dquot anyway to avoid endless cycle in dqput() */ 5055 dquot_release(dquot); 5056 return PTR_ERR(handle); 5057 } 5058 ret = dquot_release(dquot); 5059 err = ext4_journal_stop(handle); 5060 if (!ret) 5061 ret = err; 5062 return ret; 5063 } 5064 5065 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5066 { 5067 struct super_block *sb = dquot->dq_sb; 5068 struct ext4_sb_info *sbi = EXT4_SB(sb); 5069 5070 /* Are we journaling quotas? */ 5071 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) || 5072 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5073 dquot_mark_dquot_dirty(dquot); 5074 return ext4_write_dquot(dquot); 5075 } else { 5076 return dquot_mark_dquot_dirty(dquot); 5077 } 5078 } 5079 5080 static int ext4_write_info(struct super_block *sb, int type) 5081 { 5082 int ret, err; 5083 handle_t *handle; 5084 5085 /* Data block + inode block */ 5086 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2); 5087 if (IS_ERR(handle)) 5088 return PTR_ERR(handle); 5089 ret = dquot_commit_info(sb, type); 5090 err = ext4_journal_stop(handle); 5091 if (!ret) 5092 ret = err; 5093 return ret; 5094 } 5095 5096 /* 5097 * Turn on quotas during mount time - we need to find 5098 * the quota file and such... 5099 */ 5100 static int ext4_quota_on_mount(struct super_block *sb, int type) 5101 { 5102 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 5103 EXT4_SB(sb)->s_jquota_fmt, type); 5104 } 5105 5106 /* 5107 * Standard function to be called on quota_on 5108 */ 5109 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5110 struct path *path) 5111 { 5112 int err; 5113 5114 if (!test_opt(sb, QUOTA)) 5115 return -EINVAL; 5116 5117 /* Quotafile not on the same filesystem? */ 5118 if (path->dentry->d_sb != sb) 5119 return -EXDEV; 5120 /* Journaling quota? */ 5121 if (EXT4_SB(sb)->s_qf_names[type]) { 5122 /* Quotafile not in fs root? */ 5123 if (path->dentry->d_parent != sb->s_root) 5124 ext4_msg(sb, KERN_WARNING, 5125 "Quota file not on filesystem root. " 5126 "Journaled quota will not work"); 5127 } 5128 5129 /* 5130 * When we journal data on quota file, we have to flush journal to see 5131 * all updates to the file when we bypass pagecache... 5132 */ 5133 if (EXT4_SB(sb)->s_journal && 5134 ext4_should_journal_data(path->dentry->d_inode)) { 5135 /* 5136 * We don't need to lock updates but journal_flush() could 5137 * otherwise be livelocked... 5138 */ 5139 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5140 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5141 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5142 if (err) 5143 return err; 5144 } 5145 5146 return dquot_quota_on(sb, type, format_id, path); 5147 } 5148 5149 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5150 unsigned int flags) 5151 { 5152 int err; 5153 struct inode *qf_inode; 5154 unsigned long qf_inums[MAXQUOTAS] = { 5155 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5156 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 5157 }; 5158 5159 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)); 5160 5161 if (!qf_inums[type]) 5162 return -EPERM; 5163 5164 qf_inode = ext4_iget(sb, qf_inums[type]); 5165 if (IS_ERR(qf_inode)) { 5166 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5167 return PTR_ERR(qf_inode); 5168 } 5169 5170 /* Don't account quota for quota files to avoid recursion */ 5171 qf_inode->i_flags |= S_NOQUOTA; 5172 err = dquot_enable(qf_inode, type, format_id, flags); 5173 iput(qf_inode); 5174 5175 return err; 5176 } 5177 5178 /* Enable usage tracking for all quota types. */ 5179 static int ext4_enable_quotas(struct super_block *sb) 5180 { 5181 int type, err = 0; 5182 unsigned long qf_inums[MAXQUOTAS] = { 5183 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5184 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 5185 }; 5186 5187 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 5188 for (type = 0; type < MAXQUOTAS; type++) { 5189 if (qf_inums[type]) { 5190 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5191 DQUOT_USAGE_ENABLED); 5192 if (err) { 5193 ext4_warning(sb, 5194 "Failed to enable quota tracking " 5195 "(type=%d, err=%d). Please run " 5196 "e2fsck to fix.", type, err); 5197 return err; 5198 } 5199 } 5200 } 5201 return 0; 5202 } 5203 5204 /* 5205 * quota_on function that is used when QUOTA feature is set. 5206 */ 5207 static int ext4_quota_on_sysfile(struct super_block *sb, int type, 5208 int format_id) 5209 { 5210 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 5211 return -EINVAL; 5212 5213 /* 5214 * USAGE was enabled at mount time. Only need to enable LIMITS now. 5215 */ 5216 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED); 5217 } 5218 5219 static int ext4_quota_off(struct super_block *sb, int type) 5220 { 5221 struct inode *inode = sb_dqopt(sb)->files[type]; 5222 handle_t *handle; 5223 5224 /* Force all delayed allocation blocks to be allocated. 5225 * Caller already holds s_umount sem */ 5226 if (test_opt(sb, DELALLOC)) 5227 sync_filesystem(sb); 5228 5229 if (!inode) 5230 goto out; 5231 5232 /* Update modification times of quota files when userspace can 5233 * start looking at them */ 5234 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5235 if (IS_ERR(handle)) 5236 goto out; 5237 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 5238 ext4_mark_inode_dirty(handle, inode); 5239 ext4_journal_stop(handle); 5240 5241 out: 5242 return dquot_quota_off(sb, type); 5243 } 5244 5245 /* 5246 * quota_off function that is used when QUOTA feature is set. 5247 */ 5248 static int ext4_quota_off_sysfile(struct super_block *sb, int type) 5249 { 5250 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 5251 return -EINVAL; 5252 5253 /* Disable only the limits. */ 5254 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED); 5255 } 5256 5257 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5258 * acquiring the locks... As quota files are never truncated and quota code 5259 * itself serializes the operations (and no one else should touch the files) 5260 * we don't have to be afraid of races */ 5261 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5262 size_t len, loff_t off) 5263 { 5264 struct inode *inode = sb_dqopt(sb)->files[type]; 5265 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5266 int err = 0; 5267 int offset = off & (sb->s_blocksize - 1); 5268 int tocopy; 5269 size_t toread; 5270 struct buffer_head *bh; 5271 loff_t i_size = i_size_read(inode); 5272 5273 if (off > i_size) 5274 return 0; 5275 if (off+len > i_size) 5276 len = i_size-off; 5277 toread = len; 5278 while (toread > 0) { 5279 tocopy = sb->s_blocksize - offset < toread ? 5280 sb->s_blocksize - offset : toread; 5281 bh = ext4_bread(NULL, inode, blk, 0, &err); 5282 if (err) 5283 return err; 5284 if (!bh) /* A hole? */ 5285 memset(data, 0, tocopy); 5286 else 5287 memcpy(data, bh->b_data+offset, tocopy); 5288 brelse(bh); 5289 offset = 0; 5290 toread -= tocopy; 5291 data += tocopy; 5292 blk++; 5293 } 5294 return len; 5295 } 5296 5297 /* Write to quotafile (we know the transaction is already started and has 5298 * enough credits) */ 5299 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5300 const char *data, size_t len, loff_t off) 5301 { 5302 struct inode *inode = sb_dqopt(sb)->files[type]; 5303 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5304 int err = 0; 5305 int offset = off & (sb->s_blocksize - 1); 5306 struct buffer_head *bh; 5307 handle_t *handle = journal_current_handle(); 5308 5309 if (EXT4_SB(sb)->s_journal && !handle) { 5310 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5311 " cancelled because transaction is not started", 5312 (unsigned long long)off, (unsigned long long)len); 5313 return -EIO; 5314 } 5315 /* 5316 * Since we account only one data block in transaction credits, 5317 * then it is impossible to cross a block boundary. 5318 */ 5319 if (sb->s_blocksize - offset < len) { 5320 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5321 " cancelled because not block aligned", 5322 (unsigned long long)off, (unsigned long long)len); 5323 return -EIO; 5324 } 5325 5326 bh = ext4_bread(handle, inode, blk, 1, &err); 5327 if (!bh) 5328 goto out; 5329 err = ext4_journal_get_write_access(handle, bh); 5330 if (err) { 5331 brelse(bh); 5332 goto out; 5333 } 5334 lock_buffer(bh); 5335 memcpy(bh->b_data+offset, data, len); 5336 flush_dcache_page(bh->b_page); 5337 unlock_buffer(bh); 5338 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5339 brelse(bh); 5340 out: 5341 if (err) 5342 return err; 5343 if (inode->i_size < off + len) { 5344 i_size_write(inode, off + len); 5345 EXT4_I(inode)->i_disksize = inode->i_size; 5346 ext4_mark_inode_dirty(handle, inode); 5347 } 5348 return len; 5349 } 5350 5351 #endif 5352 5353 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5354 const char *dev_name, void *data) 5355 { 5356 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5357 } 5358 5359 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 5360 static inline void register_as_ext2(void) 5361 { 5362 int err = register_filesystem(&ext2_fs_type); 5363 if (err) 5364 printk(KERN_WARNING 5365 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 5366 } 5367 5368 static inline void unregister_as_ext2(void) 5369 { 5370 unregister_filesystem(&ext2_fs_type); 5371 } 5372 5373 static inline int ext2_feature_set_ok(struct super_block *sb) 5374 { 5375 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) 5376 return 0; 5377 if (sb->s_flags & MS_RDONLY) 5378 return 1; 5379 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) 5380 return 0; 5381 return 1; 5382 } 5383 #else 5384 static inline void register_as_ext2(void) { } 5385 static inline void unregister_as_ext2(void) { } 5386 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 5387 #endif 5388 5389 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 5390 static inline void register_as_ext3(void) 5391 { 5392 int err = register_filesystem(&ext3_fs_type); 5393 if (err) 5394 printk(KERN_WARNING 5395 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 5396 } 5397 5398 static inline void unregister_as_ext3(void) 5399 { 5400 unregister_filesystem(&ext3_fs_type); 5401 } 5402 5403 static inline int ext3_feature_set_ok(struct super_block *sb) 5404 { 5405 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) 5406 return 0; 5407 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 5408 return 0; 5409 if (sb->s_flags & MS_RDONLY) 5410 return 1; 5411 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) 5412 return 0; 5413 return 1; 5414 } 5415 #else 5416 static inline void register_as_ext3(void) { } 5417 static inline void unregister_as_ext3(void) { } 5418 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } 5419 #endif 5420 5421 static struct file_system_type ext4_fs_type = { 5422 .owner = THIS_MODULE, 5423 .name = "ext4", 5424 .mount = ext4_mount, 5425 .kill_sb = kill_block_super, 5426 .fs_flags = FS_REQUIRES_DEV, 5427 }; 5428 MODULE_ALIAS_FS("ext4"); 5429 5430 static int __init ext4_init_feat_adverts(void) 5431 { 5432 struct ext4_features *ef; 5433 int ret = -ENOMEM; 5434 5435 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); 5436 if (!ef) 5437 goto out; 5438 5439 ef->f_kobj.kset = ext4_kset; 5440 init_completion(&ef->f_kobj_unregister); 5441 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, 5442 "features"); 5443 if (ret) { 5444 kfree(ef); 5445 goto out; 5446 } 5447 5448 ext4_feat = ef; 5449 ret = 0; 5450 out: 5451 return ret; 5452 } 5453 5454 static void ext4_exit_feat_adverts(void) 5455 { 5456 kobject_put(&ext4_feat->f_kobj); 5457 wait_for_completion(&ext4_feat->f_kobj_unregister); 5458 kfree(ext4_feat); 5459 } 5460 5461 /* Shared across all ext4 file systems */ 5462 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5463 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 5464 5465 static int __init ext4_init_fs(void) 5466 { 5467 int i, err; 5468 5469 ext4_li_info = NULL; 5470 mutex_init(&ext4_li_mtx); 5471 5472 /* Build-time check for flags consistency */ 5473 ext4_check_flag_values(); 5474 5475 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { 5476 mutex_init(&ext4__aio_mutex[i]); 5477 init_waitqueue_head(&ext4__ioend_wq[i]); 5478 } 5479 5480 err = ext4_init_es(); 5481 if (err) 5482 return err; 5483 5484 err = ext4_init_pageio(); 5485 if (err) 5486 goto out7; 5487 5488 err = ext4_init_system_zone(); 5489 if (err) 5490 goto out6; 5491 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 5492 if (!ext4_kset) { 5493 err = -ENOMEM; 5494 goto out5; 5495 } 5496 ext4_proc_root = proc_mkdir("fs/ext4", NULL); 5497 5498 err = ext4_init_feat_adverts(); 5499 if (err) 5500 goto out4; 5501 5502 err = ext4_init_mballoc(); 5503 if (err) 5504 goto out3; 5505 5506 err = ext4_init_xattr(); 5507 if (err) 5508 goto out2; 5509 err = init_inodecache(); 5510 if (err) 5511 goto out1; 5512 register_as_ext3(); 5513 register_as_ext2(); 5514 err = register_filesystem(&ext4_fs_type); 5515 if (err) 5516 goto out; 5517 5518 return 0; 5519 out: 5520 unregister_as_ext2(); 5521 unregister_as_ext3(); 5522 destroy_inodecache(); 5523 out1: 5524 ext4_exit_xattr(); 5525 out2: 5526 ext4_exit_mballoc(); 5527 out3: 5528 ext4_exit_feat_adverts(); 5529 out4: 5530 if (ext4_proc_root) 5531 remove_proc_entry("fs/ext4", NULL); 5532 kset_unregister(ext4_kset); 5533 out5: 5534 ext4_exit_system_zone(); 5535 out6: 5536 ext4_exit_pageio(); 5537 out7: 5538 ext4_exit_es(); 5539 5540 return err; 5541 } 5542 5543 static void __exit ext4_exit_fs(void) 5544 { 5545 ext4_destroy_lazyinit_thread(); 5546 unregister_as_ext2(); 5547 unregister_as_ext3(); 5548 unregister_filesystem(&ext4_fs_type); 5549 destroy_inodecache(); 5550 ext4_exit_xattr(); 5551 ext4_exit_mballoc(); 5552 ext4_exit_feat_adverts(); 5553 remove_proc_entry("fs/ext4", NULL); 5554 kset_unregister(ext4_kset); 5555 ext4_exit_system_zone(); 5556 ext4_exit_pageio(); 5557 ext4_exit_es(); 5558 } 5559 5560 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 5561 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5562 MODULE_LICENSE("GPL"); 5563 module_init(ext4_init_fs) 5564 module_exit(ext4_exit_fs) 5565