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" 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 const char *ext4_decode_error(struct super_block *sb, int errno, 73 char nbuf[16]); 74 static int ext4_remount(struct super_block *sb, int *flags, char *data); 75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76 static int ext4_unfreeze(struct super_block *sb); 77 static int ext4_freeze(struct super_block *sb); 78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79 const char *dev_name, void *data); 80 static inline int ext2_feature_set_ok(struct super_block *sb); 81 static inline int ext3_feature_set_ok(struct super_block *sb); 82 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83 static void ext4_destroy_lazyinit_thread(void); 84 static void ext4_unregister_li_request(struct super_block *sb); 85 static void ext4_clear_request_list(void); 86 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 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 96 #else 97 #define IS_EXT2_SB(sb) (0) 98 #endif 99 100 101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 102 static struct file_system_type ext3_fs_type = { 103 .owner = THIS_MODULE, 104 .name = "ext3", 105 .mount = ext4_mount, 106 .kill_sb = kill_block_super, 107 .fs_flags = FS_REQUIRES_DEV, 108 }; 109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 110 #else 111 #define IS_EXT3_SB(sb) (0) 112 #endif 113 114 static int ext4_verify_csum_type(struct super_block *sb, 115 struct ext4_super_block *es) 116 { 117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 119 return 1; 120 121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 122 } 123 124 static __le32 ext4_superblock_csum(struct super_block *sb, 125 struct ext4_super_block *es) 126 { 127 struct ext4_sb_info *sbi = EXT4_SB(sb); 128 int offset = offsetof(struct ext4_super_block, s_checksum); 129 __u32 csum; 130 131 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 132 133 return cpu_to_le32(csum); 134 } 135 136 int ext4_superblock_csum_verify(struct super_block *sb, 137 struct ext4_super_block *es) 138 { 139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 141 return 1; 142 143 return es->s_checksum == ext4_superblock_csum(sb, es); 144 } 145 146 void ext4_superblock_csum_set(struct super_block *sb, 147 struct ext4_super_block *es) 148 { 149 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 150 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 151 return; 152 153 es->s_checksum = ext4_superblock_csum(sb, es); 154 } 155 156 void *ext4_kvmalloc(size_t size, gfp_t flags) 157 { 158 void *ret; 159 160 ret = kmalloc(size, flags); 161 if (!ret) 162 ret = __vmalloc(size, flags, PAGE_KERNEL); 163 return ret; 164 } 165 166 void *ext4_kvzalloc(size_t size, gfp_t flags) 167 { 168 void *ret; 169 170 ret = kzalloc(size, flags); 171 if (!ret) 172 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 173 return ret; 174 } 175 176 void ext4_kvfree(void *ptr) 177 { 178 if (is_vmalloc_addr(ptr)) 179 vfree(ptr); 180 else 181 kfree(ptr); 182 183 } 184 185 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 186 struct ext4_group_desc *bg) 187 { 188 return le32_to_cpu(bg->bg_block_bitmap_lo) | 189 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 190 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 191 } 192 193 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 194 struct ext4_group_desc *bg) 195 { 196 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 197 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 198 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 199 } 200 201 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 202 struct ext4_group_desc *bg) 203 { 204 return le32_to_cpu(bg->bg_inode_table_lo) | 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 207 } 208 209 __u32 ext4_free_group_clusters(struct super_block *sb, 210 struct ext4_group_desc *bg) 211 { 212 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 214 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 215 } 216 217 __u32 ext4_free_inodes_count(struct super_block *sb, 218 struct ext4_group_desc *bg) 219 { 220 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 222 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 223 } 224 225 __u32 ext4_used_dirs_count(struct super_block *sb, 226 struct ext4_group_desc *bg) 227 { 228 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 230 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 231 } 232 233 __u32 ext4_itable_unused_count(struct super_block *sb, 234 struct ext4_group_desc *bg) 235 { 236 return le16_to_cpu(bg->bg_itable_unused_lo) | 237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 238 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 239 } 240 241 void ext4_block_bitmap_set(struct super_block *sb, 242 struct ext4_group_desc *bg, ext4_fsblk_t blk) 243 { 244 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 245 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 246 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 247 } 248 249 void ext4_inode_bitmap_set(struct super_block *sb, 250 struct ext4_group_desc *bg, ext4_fsblk_t blk) 251 { 252 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 253 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 254 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 255 } 256 257 void ext4_inode_table_set(struct super_block *sb, 258 struct ext4_group_desc *bg, ext4_fsblk_t blk) 259 { 260 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 262 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 263 } 264 265 void ext4_free_group_clusters_set(struct super_block *sb, 266 struct ext4_group_desc *bg, __u32 count) 267 { 268 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 270 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 271 } 272 273 void ext4_free_inodes_set(struct super_block *sb, 274 struct ext4_group_desc *bg, __u32 count) 275 { 276 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 278 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 279 } 280 281 void ext4_used_dirs_set(struct super_block *sb, 282 struct ext4_group_desc *bg, __u32 count) 283 { 284 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 286 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 287 } 288 289 void ext4_itable_unused_set(struct super_block *sb, 290 struct ext4_group_desc *bg, __u32 count) 291 { 292 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 294 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 295 } 296 297 298 /* Just increment the non-pointer handle value */ 299 static handle_t *ext4_get_nojournal(void) 300 { 301 handle_t *handle = current->journal_info; 302 unsigned long ref_cnt = (unsigned long)handle; 303 304 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT); 305 306 ref_cnt++; 307 handle = (handle_t *)ref_cnt; 308 309 current->journal_info = handle; 310 return handle; 311 } 312 313 314 /* Decrement the non-pointer handle value */ 315 static void ext4_put_nojournal(handle_t *handle) 316 { 317 unsigned long ref_cnt = (unsigned long)handle; 318 319 BUG_ON(ref_cnt == 0); 320 321 ref_cnt--; 322 handle = (handle_t *)ref_cnt; 323 324 current->journal_info = handle; 325 } 326 327 /* 328 * Wrappers for jbd2_journal_start/end. 329 */ 330 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) 331 { 332 journal_t *journal; 333 334 trace_ext4_journal_start(sb, nblocks, _RET_IP_); 335 if (sb->s_flags & MS_RDONLY) 336 return ERR_PTR(-EROFS); 337 338 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE); 339 journal = EXT4_SB(sb)->s_journal; 340 if (!journal) 341 return ext4_get_nojournal(); 342 /* 343 * Special case here: if the journal has aborted behind our 344 * backs (eg. EIO in the commit thread), then we still need to 345 * take the FS itself readonly cleanly. 346 */ 347 if (is_journal_aborted(journal)) { 348 ext4_abort(sb, "Detected aborted journal"); 349 return ERR_PTR(-EROFS); 350 } 351 return jbd2_journal_start(journal, nblocks); 352 } 353 354 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) 355 { 356 struct super_block *sb; 357 int err; 358 int rc; 359 360 if (!ext4_handle_valid(handle)) { 361 ext4_put_nojournal(handle); 362 return 0; 363 } 364 sb = handle->h_transaction->t_journal->j_private; 365 err = handle->h_err; 366 rc = jbd2_journal_stop(handle); 367 368 if (!err) 369 err = rc; 370 if (err) 371 __ext4_std_error(sb, where, line, err); 372 return err; 373 } 374 375 void ext4_journal_abort_handle(const char *caller, unsigned int line, 376 const char *err_fn, struct buffer_head *bh, 377 handle_t *handle, int err) 378 { 379 char nbuf[16]; 380 const char *errstr = ext4_decode_error(NULL, err, nbuf); 381 382 BUG_ON(!ext4_handle_valid(handle)); 383 384 if (bh) 385 BUFFER_TRACE(bh, "abort"); 386 387 if (!handle->h_err) 388 handle->h_err = err; 389 390 if (is_handle_aborted(handle)) 391 return; 392 393 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n", 394 caller, line, errstr, err_fn); 395 396 jbd2_journal_abort_handle(handle); 397 } 398 399 static void __save_error_info(struct super_block *sb, const char *func, 400 unsigned int line) 401 { 402 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 403 404 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 405 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 406 es->s_last_error_time = cpu_to_le32(get_seconds()); 407 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 408 es->s_last_error_line = cpu_to_le32(line); 409 if (!es->s_first_error_time) { 410 es->s_first_error_time = es->s_last_error_time; 411 strncpy(es->s_first_error_func, func, 412 sizeof(es->s_first_error_func)); 413 es->s_first_error_line = cpu_to_le32(line); 414 es->s_first_error_ino = es->s_last_error_ino; 415 es->s_first_error_block = es->s_last_error_block; 416 } 417 /* 418 * Start the daily error reporting function if it hasn't been 419 * started already 420 */ 421 if (!es->s_error_count) 422 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 423 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1); 424 } 425 426 static void save_error_info(struct super_block *sb, const char *func, 427 unsigned int line) 428 { 429 __save_error_info(sb, func, line); 430 ext4_commit_super(sb, 1); 431 } 432 433 /* 434 * The del_gendisk() function uninitializes the disk-specific data 435 * structures, including the bdi structure, without telling anyone 436 * else. Once this happens, any attempt to call mark_buffer_dirty() 437 * (for example, by ext4_commit_super), will cause a kernel OOPS. 438 * This is a kludge to prevent these oops until we can put in a proper 439 * hook in del_gendisk() to inform the VFS and file system layers. 440 */ 441 static int block_device_ejected(struct super_block *sb) 442 { 443 struct inode *bd_inode = sb->s_bdev->bd_inode; 444 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info; 445 446 return bdi->dev == NULL; 447 } 448 449 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 450 { 451 struct super_block *sb = journal->j_private; 452 struct ext4_sb_info *sbi = EXT4_SB(sb); 453 int error = is_journal_aborted(journal); 454 struct ext4_journal_cb_entry *jce, *tmp; 455 456 spin_lock(&sbi->s_md_lock); 457 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) { 458 list_del_init(&jce->jce_list); 459 spin_unlock(&sbi->s_md_lock); 460 jce->jce_func(sb, jce, error); 461 spin_lock(&sbi->s_md_lock); 462 } 463 spin_unlock(&sbi->s_md_lock); 464 } 465 466 /* Deal with the reporting of failure conditions on a filesystem such as 467 * inconsistencies detected or read IO failures. 468 * 469 * On ext2, we can store the error state of the filesystem in the 470 * superblock. That is not possible on ext4, because we may have other 471 * write ordering constraints on the superblock which prevent us from 472 * writing it out straight away; and given that the journal is about to 473 * be aborted, we can't rely on the current, or future, transactions to 474 * write out the superblock safely. 475 * 476 * We'll just use the jbd2_journal_abort() error code to record an error in 477 * the journal instead. On recovery, the journal will complain about 478 * that error until we've noted it down and cleared it. 479 */ 480 481 static void ext4_handle_error(struct super_block *sb) 482 { 483 if (sb->s_flags & MS_RDONLY) 484 return; 485 486 if (!test_opt(sb, ERRORS_CONT)) { 487 journal_t *journal = EXT4_SB(sb)->s_journal; 488 489 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 490 if (journal) 491 jbd2_journal_abort(journal, -EIO); 492 } 493 if (test_opt(sb, ERRORS_RO)) { 494 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 495 sb->s_flags |= MS_RDONLY; 496 } 497 if (test_opt(sb, ERRORS_PANIC)) 498 panic("EXT4-fs (device %s): panic forced after error\n", 499 sb->s_id); 500 } 501 502 void __ext4_error(struct super_block *sb, const char *function, 503 unsigned int line, const char *fmt, ...) 504 { 505 struct va_format vaf; 506 va_list args; 507 508 va_start(args, fmt); 509 vaf.fmt = fmt; 510 vaf.va = &args; 511 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 512 sb->s_id, function, line, current->comm, &vaf); 513 va_end(args); 514 save_error_info(sb, function, line); 515 516 ext4_handle_error(sb); 517 } 518 519 void ext4_error_inode(struct inode *inode, const char *function, 520 unsigned int line, ext4_fsblk_t block, 521 const char *fmt, ...) 522 { 523 va_list args; 524 struct va_format vaf; 525 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 526 527 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 528 es->s_last_error_block = cpu_to_le64(block); 529 save_error_info(inode->i_sb, function, line); 530 va_start(args, fmt); 531 vaf.fmt = fmt; 532 vaf.va = &args; 533 if (block) 534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 535 "inode #%lu: block %llu: comm %s: %pV\n", 536 inode->i_sb->s_id, function, line, inode->i_ino, 537 block, current->comm, &vaf); 538 else 539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 540 "inode #%lu: comm %s: %pV\n", 541 inode->i_sb->s_id, function, line, inode->i_ino, 542 current->comm, &vaf); 543 va_end(args); 544 545 ext4_handle_error(inode->i_sb); 546 } 547 548 void ext4_error_file(struct file *file, const char *function, 549 unsigned int line, ext4_fsblk_t block, 550 const char *fmt, ...) 551 { 552 va_list args; 553 struct va_format vaf; 554 struct ext4_super_block *es; 555 struct inode *inode = file->f_dentry->d_inode; 556 char pathname[80], *path; 557 558 es = EXT4_SB(inode->i_sb)->s_es; 559 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 560 save_error_info(inode->i_sb, function, line); 561 path = d_path(&(file->f_path), pathname, sizeof(pathname)); 562 if (IS_ERR(path)) 563 path = "(unknown)"; 564 va_start(args, fmt); 565 vaf.fmt = fmt; 566 vaf.va = &args; 567 if (block) 568 printk(KERN_CRIT 569 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 570 "block %llu: comm %s: path %s: %pV\n", 571 inode->i_sb->s_id, function, line, inode->i_ino, 572 block, current->comm, path, &vaf); 573 else 574 printk(KERN_CRIT 575 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 576 "comm %s: path %s: %pV\n", 577 inode->i_sb->s_id, function, line, inode->i_ino, 578 current->comm, path, &vaf); 579 va_end(args); 580 581 ext4_handle_error(inode->i_sb); 582 } 583 584 static const char *ext4_decode_error(struct super_block *sb, int errno, 585 char nbuf[16]) 586 { 587 char *errstr = NULL; 588 589 switch (errno) { 590 case -EIO: 591 errstr = "IO failure"; 592 break; 593 case -ENOMEM: 594 errstr = "Out of memory"; 595 break; 596 case -EROFS: 597 if (!sb || (EXT4_SB(sb)->s_journal && 598 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 599 errstr = "Journal has aborted"; 600 else 601 errstr = "Readonly filesystem"; 602 break; 603 default: 604 /* If the caller passed in an extra buffer for unknown 605 * errors, textualise them now. Else we just return 606 * NULL. */ 607 if (nbuf) { 608 /* Check for truncated error codes... */ 609 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 610 errstr = nbuf; 611 } 612 break; 613 } 614 615 return errstr; 616 } 617 618 /* __ext4_std_error decodes expected errors from journaling functions 619 * automatically and invokes the appropriate error response. */ 620 621 void __ext4_std_error(struct super_block *sb, const char *function, 622 unsigned int line, int errno) 623 { 624 char nbuf[16]; 625 const char *errstr; 626 627 /* Special case: if the error is EROFS, and we're not already 628 * inside a transaction, then there's really no point in logging 629 * an error. */ 630 if (errno == -EROFS && journal_current_handle() == NULL && 631 (sb->s_flags & MS_RDONLY)) 632 return; 633 634 errstr = ext4_decode_error(sb, errno, nbuf); 635 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 636 sb->s_id, function, line, errstr); 637 save_error_info(sb, function, line); 638 639 ext4_handle_error(sb); 640 } 641 642 /* 643 * ext4_abort is a much stronger failure handler than ext4_error. The 644 * abort function may be used to deal with unrecoverable failures such 645 * as journal IO errors or ENOMEM at a critical moment in log management. 646 * 647 * We unconditionally force the filesystem into an ABORT|READONLY state, 648 * unless the error response on the fs has been set to panic in which 649 * case we take the easy way out and panic immediately. 650 */ 651 652 void __ext4_abort(struct super_block *sb, const char *function, 653 unsigned int line, const char *fmt, ...) 654 { 655 va_list args; 656 657 save_error_info(sb, function, line); 658 va_start(args, fmt); 659 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, 660 function, line); 661 vprintk(fmt, args); 662 printk("\n"); 663 va_end(args); 664 665 if ((sb->s_flags & MS_RDONLY) == 0) { 666 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 667 sb->s_flags |= MS_RDONLY; 668 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 669 if (EXT4_SB(sb)->s_journal) 670 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 671 save_error_info(sb, function, line); 672 } 673 if (test_opt(sb, ERRORS_PANIC)) 674 panic("EXT4-fs panic from previous error\n"); 675 } 676 677 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...) 678 { 679 struct va_format vaf; 680 va_list args; 681 682 va_start(args, fmt); 683 vaf.fmt = fmt; 684 vaf.va = &args; 685 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 686 va_end(args); 687 } 688 689 void __ext4_warning(struct super_block *sb, const char *function, 690 unsigned int line, const char *fmt, ...) 691 { 692 struct va_format vaf; 693 va_list args; 694 695 va_start(args, fmt); 696 vaf.fmt = fmt; 697 vaf.va = &args; 698 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 699 sb->s_id, function, line, &vaf); 700 va_end(args); 701 } 702 703 void __ext4_grp_locked_error(const char *function, unsigned int line, 704 struct super_block *sb, ext4_group_t grp, 705 unsigned long ino, ext4_fsblk_t block, 706 const char *fmt, ...) 707 __releases(bitlock) 708 __acquires(bitlock) 709 { 710 struct va_format vaf; 711 va_list args; 712 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 713 714 es->s_last_error_ino = cpu_to_le32(ino); 715 es->s_last_error_block = cpu_to_le64(block); 716 __save_error_info(sb, function, line); 717 718 va_start(args, fmt); 719 720 vaf.fmt = fmt; 721 vaf.va = &args; 722 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 723 sb->s_id, function, line, grp); 724 if (ino) 725 printk(KERN_CONT "inode %lu: ", ino); 726 if (block) 727 printk(KERN_CONT "block %llu:", (unsigned long long) block); 728 printk(KERN_CONT "%pV\n", &vaf); 729 va_end(args); 730 731 if (test_opt(sb, ERRORS_CONT)) { 732 ext4_commit_super(sb, 0); 733 return; 734 } 735 736 ext4_unlock_group(sb, grp); 737 ext4_handle_error(sb); 738 /* 739 * We only get here in the ERRORS_RO case; relocking the group 740 * may be dangerous, but nothing bad will happen since the 741 * filesystem will have already been marked read/only and the 742 * journal has been aborted. We return 1 as a hint to callers 743 * who might what to use the return value from 744 * ext4_grp_locked_error() to distinguish between the 745 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 746 * aggressively from the ext4 function in question, with a 747 * more appropriate error code. 748 */ 749 ext4_lock_group(sb, grp); 750 return; 751 } 752 753 void ext4_update_dynamic_rev(struct super_block *sb) 754 { 755 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 756 757 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 758 return; 759 760 ext4_warning(sb, 761 "updating to rev %d because of new feature flag, " 762 "running e2fsck is recommended", 763 EXT4_DYNAMIC_REV); 764 765 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 766 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 767 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 768 /* leave es->s_feature_*compat flags alone */ 769 /* es->s_uuid will be set by e2fsck if empty */ 770 771 /* 772 * The rest of the superblock fields should be zero, and if not it 773 * means they are likely already in use, so leave them alone. We 774 * can leave it up to e2fsck to clean up any inconsistencies there. 775 */ 776 } 777 778 /* 779 * Open the external journal device 780 */ 781 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 782 { 783 struct block_device *bdev; 784 char b[BDEVNAME_SIZE]; 785 786 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 787 if (IS_ERR(bdev)) 788 goto fail; 789 return bdev; 790 791 fail: 792 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 793 __bdevname(dev, b), PTR_ERR(bdev)); 794 return NULL; 795 } 796 797 /* 798 * Release the journal device 799 */ 800 static int ext4_blkdev_put(struct block_device *bdev) 801 { 802 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 803 } 804 805 static int ext4_blkdev_remove(struct ext4_sb_info *sbi) 806 { 807 struct block_device *bdev; 808 int ret = -ENODEV; 809 810 bdev = sbi->journal_bdev; 811 if (bdev) { 812 ret = ext4_blkdev_put(bdev); 813 sbi->journal_bdev = NULL; 814 } 815 return ret; 816 } 817 818 static inline struct inode *orphan_list_entry(struct list_head *l) 819 { 820 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 821 } 822 823 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 824 { 825 struct list_head *l; 826 827 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 828 le32_to_cpu(sbi->s_es->s_last_orphan)); 829 830 printk(KERN_ERR "sb_info orphan list:\n"); 831 list_for_each(l, &sbi->s_orphan) { 832 struct inode *inode = orphan_list_entry(l); 833 printk(KERN_ERR " " 834 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 835 inode->i_sb->s_id, inode->i_ino, inode, 836 inode->i_mode, inode->i_nlink, 837 NEXT_ORPHAN(inode)); 838 } 839 } 840 841 static void ext4_put_super(struct super_block *sb) 842 { 843 struct ext4_sb_info *sbi = EXT4_SB(sb); 844 struct ext4_super_block *es = sbi->s_es; 845 int i, err; 846 847 ext4_unregister_li_request(sb); 848 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 849 850 flush_workqueue(sbi->dio_unwritten_wq); 851 destroy_workqueue(sbi->dio_unwritten_wq); 852 853 lock_super(sb); 854 if (sbi->s_journal) { 855 err = jbd2_journal_destroy(sbi->s_journal); 856 sbi->s_journal = NULL; 857 if (err < 0) 858 ext4_abort(sb, "Couldn't clean up the journal"); 859 } 860 861 del_timer(&sbi->s_err_report); 862 ext4_release_system_zone(sb); 863 ext4_mb_release(sb); 864 ext4_ext_release(sb); 865 ext4_xattr_put_super(sb); 866 867 if (!(sb->s_flags & MS_RDONLY)) { 868 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 869 es->s_state = cpu_to_le16(sbi->s_mount_state); 870 } 871 if (!(sb->s_flags & MS_RDONLY)) 872 ext4_commit_super(sb, 1); 873 874 if (sbi->s_proc) { 875 remove_proc_entry("options", sbi->s_proc); 876 remove_proc_entry(sb->s_id, ext4_proc_root); 877 } 878 kobject_del(&sbi->s_kobj); 879 880 for (i = 0; i < sbi->s_gdb_count; i++) 881 brelse(sbi->s_group_desc[i]); 882 ext4_kvfree(sbi->s_group_desc); 883 ext4_kvfree(sbi->s_flex_groups); 884 percpu_counter_destroy(&sbi->s_freeclusters_counter); 885 percpu_counter_destroy(&sbi->s_freeinodes_counter); 886 percpu_counter_destroy(&sbi->s_dirs_counter); 887 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 888 brelse(sbi->s_sbh); 889 #ifdef CONFIG_QUOTA 890 for (i = 0; i < MAXQUOTAS; i++) 891 kfree(sbi->s_qf_names[i]); 892 #endif 893 894 /* Debugging code just in case the in-memory inode orphan list 895 * isn't empty. The on-disk one can be non-empty if we've 896 * detected an error and taken the fs readonly, but the 897 * in-memory list had better be clean by this point. */ 898 if (!list_empty(&sbi->s_orphan)) 899 dump_orphan_list(sb, sbi); 900 J_ASSERT(list_empty(&sbi->s_orphan)); 901 902 invalidate_bdev(sb->s_bdev); 903 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 904 /* 905 * Invalidate the journal device's buffers. We don't want them 906 * floating about in memory - the physical journal device may 907 * hotswapped, and it breaks the `ro-after' testing code. 908 */ 909 sync_blockdev(sbi->journal_bdev); 910 invalidate_bdev(sbi->journal_bdev); 911 ext4_blkdev_remove(sbi); 912 } 913 if (sbi->s_mmp_tsk) 914 kthread_stop(sbi->s_mmp_tsk); 915 sb->s_fs_info = NULL; 916 /* 917 * Now that we are completely done shutting down the 918 * superblock, we need to actually destroy the kobject. 919 */ 920 unlock_super(sb); 921 kobject_put(&sbi->s_kobj); 922 wait_for_completion(&sbi->s_kobj_unregister); 923 if (sbi->s_chksum_driver) 924 crypto_free_shash(sbi->s_chksum_driver); 925 kfree(sbi->s_blockgroup_lock); 926 kfree(sbi); 927 } 928 929 static struct kmem_cache *ext4_inode_cachep; 930 931 /* 932 * Called inside transaction, so use GFP_NOFS 933 */ 934 static struct inode *ext4_alloc_inode(struct super_block *sb) 935 { 936 struct ext4_inode_info *ei; 937 938 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 939 if (!ei) 940 return NULL; 941 942 ei->vfs_inode.i_version = 1; 943 ei->vfs_inode.i_data.writeback_index = 0; 944 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); 945 INIT_LIST_HEAD(&ei->i_prealloc_list); 946 spin_lock_init(&ei->i_prealloc_lock); 947 ei->i_reserved_data_blocks = 0; 948 ei->i_reserved_meta_blocks = 0; 949 ei->i_allocated_meta_blocks = 0; 950 ei->i_da_metadata_calc_len = 0; 951 spin_lock_init(&(ei->i_block_reservation_lock)); 952 #ifdef CONFIG_QUOTA 953 ei->i_reserved_quota = 0; 954 #endif 955 ei->jinode = NULL; 956 INIT_LIST_HEAD(&ei->i_completed_io_list); 957 spin_lock_init(&ei->i_completed_io_lock); 958 ei->cur_aio_dio = NULL; 959 ei->i_sync_tid = 0; 960 ei->i_datasync_tid = 0; 961 atomic_set(&ei->i_ioend_count, 0); 962 atomic_set(&ei->i_aiodio_unwritten, 0); 963 964 return &ei->vfs_inode; 965 } 966 967 static int ext4_drop_inode(struct inode *inode) 968 { 969 int drop = generic_drop_inode(inode); 970 971 trace_ext4_drop_inode(inode, drop); 972 return drop; 973 } 974 975 static void ext4_i_callback(struct rcu_head *head) 976 { 977 struct inode *inode = container_of(head, struct inode, i_rcu); 978 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 979 } 980 981 static void ext4_destroy_inode(struct inode *inode) 982 { 983 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 984 ext4_msg(inode->i_sb, KERN_ERR, 985 "Inode %lu (%p): orphan list check failed!", 986 inode->i_ino, EXT4_I(inode)); 987 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 988 EXT4_I(inode), sizeof(struct ext4_inode_info), 989 true); 990 dump_stack(); 991 } 992 call_rcu(&inode->i_rcu, ext4_i_callback); 993 } 994 995 static void init_once(void *foo) 996 { 997 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 998 999 INIT_LIST_HEAD(&ei->i_orphan); 1000 #ifdef CONFIG_EXT4_FS_XATTR 1001 init_rwsem(&ei->xattr_sem); 1002 #endif 1003 init_rwsem(&ei->i_data_sem); 1004 inode_init_once(&ei->vfs_inode); 1005 } 1006 1007 static int init_inodecache(void) 1008 { 1009 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 1010 sizeof(struct ext4_inode_info), 1011 0, (SLAB_RECLAIM_ACCOUNT| 1012 SLAB_MEM_SPREAD), 1013 init_once); 1014 if (ext4_inode_cachep == NULL) 1015 return -ENOMEM; 1016 return 0; 1017 } 1018 1019 static void destroy_inodecache(void) 1020 { 1021 kmem_cache_destroy(ext4_inode_cachep); 1022 } 1023 1024 void ext4_clear_inode(struct inode *inode) 1025 { 1026 invalidate_inode_buffers(inode); 1027 clear_inode(inode); 1028 dquot_drop(inode); 1029 ext4_discard_preallocations(inode); 1030 if (EXT4_I(inode)->jinode) { 1031 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1032 EXT4_I(inode)->jinode); 1033 jbd2_free_inode(EXT4_I(inode)->jinode); 1034 EXT4_I(inode)->jinode = NULL; 1035 } 1036 } 1037 1038 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1039 u64 ino, u32 generation) 1040 { 1041 struct inode *inode; 1042 1043 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 1044 return ERR_PTR(-ESTALE); 1045 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 1046 return ERR_PTR(-ESTALE); 1047 1048 /* iget isn't really right if the inode is currently unallocated!! 1049 * 1050 * ext4_read_inode will return a bad_inode if the inode had been 1051 * deleted, so we should be safe. 1052 * 1053 * Currently we don't know the generation for parent directory, so 1054 * a generation of 0 means "accept any" 1055 */ 1056 inode = ext4_iget(sb, ino); 1057 if (IS_ERR(inode)) 1058 return ERR_CAST(inode); 1059 if (generation && inode->i_generation != generation) { 1060 iput(inode); 1061 return ERR_PTR(-ESTALE); 1062 } 1063 1064 return inode; 1065 } 1066 1067 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1068 int fh_len, int fh_type) 1069 { 1070 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1071 ext4_nfs_get_inode); 1072 } 1073 1074 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1075 int fh_len, int fh_type) 1076 { 1077 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1078 ext4_nfs_get_inode); 1079 } 1080 1081 /* 1082 * Try to release metadata pages (indirect blocks, directories) which are 1083 * mapped via the block device. Since these pages could have journal heads 1084 * which would prevent try_to_free_buffers() from freeing them, we must use 1085 * jbd2 layer's try_to_free_buffers() function to release them. 1086 */ 1087 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1088 gfp_t wait) 1089 { 1090 journal_t *journal = EXT4_SB(sb)->s_journal; 1091 1092 WARN_ON(PageChecked(page)); 1093 if (!page_has_buffers(page)) 1094 return 0; 1095 if (journal) 1096 return jbd2_journal_try_to_free_buffers(journal, page, 1097 wait & ~__GFP_WAIT); 1098 return try_to_free_buffers(page); 1099 } 1100 1101 #ifdef CONFIG_QUOTA 1102 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") 1103 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) 1104 1105 static int ext4_write_dquot(struct dquot *dquot); 1106 static int ext4_acquire_dquot(struct dquot *dquot); 1107 static int ext4_release_dquot(struct dquot *dquot); 1108 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1109 static int ext4_write_info(struct super_block *sb, int type); 1110 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1111 struct path *path); 1112 static int ext4_quota_on_sysfile(struct super_block *sb, int type, 1113 int format_id); 1114 static int ext4_quota_off(struct super_block *sb, int type); 1115 static int ext4_quota_off_sysfile(struct super_block *sb, int type); 1116 static int ext4_quota_on_mount(struct super_block *sb, int type); 1117 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1118 size_t len, loff_t off); 1119 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1120 const char *data, size_t len, loff_t off); 1121 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1122 unsigned int flags); 1123 static int ext4_enable_quotas(struct super_block *sb); 1124 1125 static const struct dquot_operations ext4_quota_operations = { 1126 .get_reserved_space = ext4_get_reserved_space, 1127 .write_dquot = ext4_write_dquot, 1128 .acquire_dquot = ext4_acquire_dquot, 1129 .release_dquot = ext4_release_dquot, 1130 .mark_dirty = ext4_mark_dquot_dirty, 1131 .write_info = ext4_write_info, 1132 .alloc_dquot = dquot_alloc, 1133 .destroy_dquot = dquot_destroy, 1134 }; 1135 1136 static const struct quotactl_ops ext4_qctl_operations = { 1137 .quota_on = ext4_quota_on, 1138 .quota_off = ext4_quota_off, 1139 .quota_sync = dquot_quota_sync, 1140 .get_info = dquot_get_dqinfo, 1141 .set_info = dquot_set_dqinfo, 1142 .get_dqblk = dquot_get_dqblk, 1143 .set_dqblk = dquot_set_dqblk 1144 }; 1145 1146 static const struct quotactl_ops ext4_qctl_sysfile_operations = { 1147 .quota_on_meta = ext4_quota_on_sysfile, 1148 .quota_off = ext4_quota_off_sysfile, 1149 .quota_sync = dquot_quota_sync, 1150 .get_info = dquot_get_dqinfo, 1151 .set_info = dquot_set_dqinfo, 1152 .get_dqblk = dquot_get_dqblk, 1153 .set_dqblk = dquot_set_dqblk 1154 }; 1155 #endif 1156 1157 static const struct super_operations ext4_sops = { 1158 .alloc_inode = ext4_alloc_inode, 1159 .destroy_inode = ext4_destroy_inode, 1160 .write_inode = ext4_write_inode, 1161 .dirty_inode = ext4_dirty_inode, 1162 .drop_inode = ext4_drop_inode, 1163 .evict_inode = ext4_evict_inode, 1164 .put_super = ext4_put_super, 1165 .sync_fs = ext4_sync_fs, 1166 .freeze_fs = ext4_freeze, 1167 .unfreeze_fs = ext4_unfreeze, 1168 .statfs = ext4_statfs, 1169 .remount_fs = ext4_remount, 1170 .show_options = ext4_show_options, 1171 #ifdef CONFIG_QUOTA 1172 .quota_read = ext4_quota_read, 1173 .quota_write = ext4_quota_write, 1174 #endif 1175 .bdev_try_to_free_page = bdev_try_to_free_page, 1176 }; 1177 1178 static const struct super_operations ext4_nojournal_sops = { 1179 .alloc_inode = ext4_alloc_inode, 1180 .destroy_inode = ext4_destroy_inode, 1181 .write_inode = ext4_write_inode, 1182 .dirty_inode = ext4_dirty_inode, 1183 .drop_inode = ext4_drop_inode, 1184 .evict_inode = ext4_evict_inode, 1185 .put_super = ext4_put_super, 1186 .statfs = ext4_statfs, 1187 .remount_fs = ext4_remount, 1188 .show_options = ext4_show_options, 1189 #ifdef CONFIG_QUOTA 1190 .quota_read = ext4_quota_read, 1191 .quota_write = ext4_quota_write, 1192 #endif 1193 .bdev_try_to_free_page = bdev_try_to_free_page, 1194 }; 1195 1196 static const struct export_operations ext4_export_ops = { 1197 .fh_to_dentry = ext4_fh_to_dentry, 1198 .fh_to_parent = ext4_fh_to_parent, 1199 .get_parent = ext4_get_parent, 1200 }; 1201 1202 enum { 1203 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1204 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1205 Opt_nouid32, Opt_debug, Opt_removed, 1206 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1207 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1208 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, 1209 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit, 1210 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1211 Opt_data_err_abort, Opt_data_err_ignore, 1212 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1213 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1214 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1215 Opt_usrquota, Opt_grpquota, Opt_i_version, 1216 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 1217 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1218 Opt_inode_readahead_blks, Opt_journal_ioprio, 1219 Opt_dioread_nolock, Opt_dioread_lock, 1220 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1221 }; 1222 1223 static const match_table_t tokens = { 1224 {Opt_bsd_df, "bsddf"}, 1225 {Opt_minix_df, "minixdf"}, 1226 {Opt_grpid, "grpid"}, 1227 {Opt_grpid, "bsdgroups"}, 1228 {Opt_nogrpid, "nogrpid"}, 1229 {Opt_nogrpid, "sysvgroups"}, 1230 {Opt_resgid, "resgid=%u"}, 1231 {Opt_resuid, "resuid=%u"}, 1232 {Opt_sb, "sb=%u"}, 1233 {Opt_err_cont, "errors=continue"}, 1234 {Opt_err_panic, "errors=panic"}, 1235 {Opt_err_ro, "errors=remount-ro"}, 1236 {Opt_nouid32, "nouid32"}, 1237 {Opt_debug, "debug"}, 1238 {Opt_removed, "oldalloc"}, 1239 {Opt_removed, "orlov"}, 1240 {Opt_user_xattr, "user_xattr"}, 1241 {Opt_nouser_xattr, "nouser_xattr"}, 1242 {Opt_acl, "acl"}, 1243 {Opt_noacl, "noacl"}, 1244 {Opt_noload, "norecovery"}, 1245 {Opt_noload, "noload"}, 1246 {Opt_removed, "nobh"}, 1247 {Opt_removed, "bh"}, 1248 {Opt_commit, "commit=%u"}, 1249 {Opt_min_batch_time, "min_batch_time=%u"}, 1250 {Opt_max_batch_time, "max_batch_time=%u"}, 1251 {Opt_journal_dev, "journal_dev=%u"}, 1252 {Opt_journal_checksum, "journal_checksum"}, 1253 {Opt_journal_async_commit, "journal_async_commit"}, 1254 {Opt_abort, "abort"}, 1255 {Opt_data_journal, "data=journal"}, 1256 {Opt_data_ordered, "data=ordered"}, 1257 {Opt_data_writeback, "data=writeback"}, 1258 {Opt_data_err_abort, "data_err=abort"}, 1259 {Opt_data_err_ignore, "data_err=ignore"}, 1260 {Opt_offusrjquota, "usrjquota="}, 1261 {Opt_usrjquota, "usrjquota=%s"}, 1262 {Opt_offgrpjquota, "grpjquota="}, 1263 {Opt_grpjquota, "grpjquota=%s"}, 1264 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1265 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1266 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1267 {Opt_grpquota, "grpquota"}, 1268 {Opt_noquota, "noquota"}, 1269 {Opt_quota, "quota"}, 1270 {Opt_usrquota, "usrquota"}, 1271 {Opt_barrier, "barrier=%u"}, 1272 {Opt_barrier, "barrier"}, 1273 {Opt_nobarrier, "nobarrier"}, 1274 {Opt_i_version, "i_version"}, 1275 {Opt_stripe, "stripe=%u"}, 1276 {Opt_delalloc, "delalloc"}, 1277 {Opt_nodelalloc, "nodelalloc"}, 1278 {Opt_mblk_io_submit, "mblk_io_submit"}, 1279 {Opt_nomblk_io_submit, "nomblk_io_submit"}, 1280 {Opt_block_validity, "block_validity"}, 1281 {Opt_noblock_validity, "noblock_validity"}, 1282 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1283 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1284 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1285 {Opt_auto_da_alloc, "auto_da_alloc"}, 1286 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1287 {Opt_dioread_nolock, "dioread_nolock"}, 1288 {Opt_dioread_lock, "dioread_lock"}, 1289 {Opt_discard, "discard"}, 1290 {Opt_nodiscard, "nodiscard"}, 1291 {Opt_init_itable, "init_itable=%u"}, 1292 {Opt_init_itable, "init_itable"}, 1293 {Opt_noinit_itable, "noinit_itable"}, 1294 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1295 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1296 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1297 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1298 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1299 {Opt_err, NULL}, 1300 }; 1301 1302 static ext4_fsblk_t get_sb_block(void **data) 1303 { 1304 ext4_fsblk_t sb_block; 1305 char *options = (char *) *data; 1306 1307 if (!options || strncmp(options, "sb=", 3) != 0) 1308 return 1; /* Default location */ 1309 1310 options += 3; 1311 /* TODO: use simple_strtoll with >32bit ext4 */ 1312 sb_block = simple_strtoul(options, &options, 0); 1313 if (*options && *options != ',') { 1314 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1315 (char *) *data); 1316 return 1; 1317 } 1318 if (*options == ',') 1319 options++; 1320 *data = (void *) options; 1321 1322 return sb_block; 1323 } 1324 1325 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1326 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1327 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1328 1329 #ifdef CONFIG_QUOTA 1330 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1331 { 1332 struct ext4_sb_info *sbi = EXT4_SB(sb); 1333 char *qname; 1334 1335 if (sb_any_quota_loaded(sb) && 1336 !sbi->s_qf_names[qtype]) { 1337 ext4_msg(sb, KERN_ERR, 1338 "Cannot change journaled " 1339 "quota options when quota turned on"); 1340 return -1; 1341 } 1342 qname = match_strdup(args); 1343 if (!qname) { 1344 ext4_msg(sb, KERN_ERR, 1345 "Not enough memory for storing quotafile name"); 1346 return -1; 1347 } 1348 if (sbi->s_qf_names[qtype] && 1349 strcmp(sbi->s_qf_names[qtype], qname)) { 1350 ext4_msg(sb, KERN_ERR, 1351 "%s quota file already specified", QTYPE2NAME(qtype)); 1352 kfree(qname); 1353 return -1; 1354 } 1355 sbi->s_qf_names[qtype] = qname; 1356 if (strchr(sbi->s_qf_names[qtype], '/')) { 1357 ext4_msg(sb, KERN_ERR, 1358 "quotafile must be on filesystem root"); 1359 kfree(sbi->s_qf_names[qtype]); 1360 sbi->s_qf_names[qtype] = NULL; 1361 return -1; 1362 } 1363 set_opt(sb, QUOTA); 1364 return 1; 1365 } 1366 1367 static int clear_qf_name(struct super_block *sb, int qtype) 1368 { 1369 1370 struct ext4_sb_info *sbi = EXT4_SB(sb); 1371 1372 if (sb_any_quota_loaded(sb) && 1373 sbi->s_qf_names[qtype]) { 1374 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1375 " when quota turned on"); 1376 return -1; 1377 } 1378 /* 1379 * The space will be released later when all options are confirmed 1380 * to be correct 1381 */ 1382 sbi->s_qf_names[qtype] = NULL; 1383 return 1; 1384 } 1385 #endif 1386 1387 #define MOPT_SET 0x0001 1388 #define MOPT_CLEAR 0x0002 1389 #define MOPT_NOSUPPORT 0x0004 1390 #define MOPT_EXPLICIT 0x0008 1391 #define MOPT_CLEAR_ERR 0x0010 1392 #define MOPT_GTE0 0x0020 1393 #ifdef CONFIG_QUOTA 1394 #define MOPT_Q 0 1395 #define MOPT_QFMT 0x0040 1396 #else 1397 #define MOPT_Q MOPT_NOSUPPORT 1398 #define MOPT_QFMT MOPT_NOSUPPORT 1399 #endif 1400 #define MOPT_DATAJ 0x0080 1401 1402 static const struct mount_opts { 1403 int token; 1404 int mount_opt; 1405 int flags; 1406 } ext4_mount_opts[] = { 1407 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1408 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1409 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1410 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1411 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET}, 1412 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR}, 1413 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1414 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1415 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET}, 1416 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR}, 1417 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1418 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1419 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT}, 1420 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT}, 1421 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET}, 1422 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1423 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET}, 1424 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET}, 1425 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1426 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1427 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1428 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET}, 1429 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR}, 1430 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1431 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1432 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1433 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1434 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1435 {Opt_commit, 0, MOPT_GTE0}, 1436 {Opt_max_batch_time, 0, MOPT_GTE0}, 1437 {Opt_min_batch_time, 0, MOPT_GTE0}, 1438 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1439 {Opt_init_itable, 0, MOPT_GTE0}, 1440 {Opt_stripe, 0, MOPT_GTE0}, 1441 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ}, 1442 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ}, 1443 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ}, 1444 #ifdef CONFIG_EXT4_FS_XATTR 1445 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1446 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1447 #else 1448 {Opt_user_xattr, 0, MOPT_NOSUPPORT}, 1449 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT}, 1450 #endif 1451 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1452 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1453 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1454 #else 1455 {Opt_acl, 0, MOPT_NOSUPPORT}, 1456 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1457 #endif 1458 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1459 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1460 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1461 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1462 MOPT_SET | MOPT_Q}, 1463 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1464 MOPT_SET | MOPT_Q}, 1465 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1466 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q}, 1467 {Opt_usrjquota, 0, MOPT_Q}, 1468 {Opt_grpjquota, 0, MOPT_Q}, 1469 {Opt_offusrjquota, 0, MOPT_Q}, 1470 {Opt_offgrpjquota, 0, MOPT_Q}, 1471 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1472 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1473 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1474 {Opt_err, 0, 0} 1475 }; 1476 1477 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1478 substring_t *args, unsigned long *journal_devnum, 1479 unsigned int *journal_ioprio, int is_remount) 1480 { 1481 struct ext4_sb_info *sbi = EXT4_SB(sb); 1482 const struct mount_opts *m; 1483 kuid_t uid; 1484 kgid_t gid; 1485 int arg = 0; 1486 1487 #ifdef CONFIG_QUOTA 1488 if (token == Opt_usrjquota) 1489 return set_qf_name(sb, USRQUOTA, &args[0]); 1490 else if (token == Opt_grpjquota) 1491 return set_qf_name(sb, GRPQUOTA, &args[0]); 1492 else if (token == Opt_offusrjquota) 1493 return clear_qf_name(sb, USRQUOTA); 1494 else if (token == Opt_offgrpjquota) 1495 return clear_qf_name(sb, GRPQUOTA); 1496 #endif 1497 if (args->from && match_int(args, &arg)) 1498 return -1; 1499 switch (token) { 1500 case Opt_noacl: 1501 case Opt_nouser_xattr: 1502 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1503 break; 1504 case Opt_sb: 1505 return 1; /* handled by get_sb_block() */ 1506 case Opt_removed: 1507 ext4_msg(sb, KERN_WARNING, 1508 "Ignoring removed %s option", opt); 1509 return 1; 1510 case Opt_resuid: 1511 uid = make_kuid(current_user_ns(), arg); 1512 if (!uid_valid(uid)) { 1513 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1514 return -1; 1515 } 1516 sbi->s_resuid = uid; 1517 return 1; 1518 case Opt_resgid: 1519 gid = make_kgid(current_user_ns(), arg); 1520 if (!gid_valid(gid)) { 1521 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1522 return -1; 1523 } 1524 sbi->s_resgid = gid; 1525 return 1; 1526 case Opt_abort: 1527 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1528 return 1; 1529 case Opt_i_version: 1530 sb->s_flags |= MS_I_VERSION; 1531 return 1; 1532 case Opt_journal_dev: 1533 if (is_remount) { 1534 ext4_msg(sb, KERN_ERR, 1535 "Cannot specify journal on remount"); 1536 return -1; 1537 } 1538 *journal_devnum = arg; 1539 return 1; 1540 case Opt_journal_ioprio: 1541 if (arg < 0 || arg > 7) 1542 return -1; 1543 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1544 return 1; 1545 } 1546 1547 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 1548 if (token != m->token) 1549 continue; 1550 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1551 return -1; 1552 if (m->flags & MOPT_EXPLICIT) 1553 set_opt2(sb, EXPLICIT_DELALLOC); 1554 if (m->flags & MOPT_CLEAR_ERR) 1555 clear_opt(sb, ERRORS_MASK); 1556 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1557 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1558 "options when quota turned on"); 1559 return -1; 1560 } 1561 1562 if (m->flags & MOPT_NOSUPPORT) { 1563 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1564 } else if (token == Opt_commit) { 1565 if (arg == 0) 1566 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1567 sbi->s_commit_interval = HZ * arg; 1568 } else if (token == Opt_max_batch_time) { 1569 if (arg == 0) 1570 arg = EXT4_DEF_MAX_BATCH_TIME; 1571 sbi->s_max_batch_time = arg; 1572 } else if (token == Opt_min_batch_time) { 1573 sbi->s_min_batch_time = arg; 1574 } else if (token == Opt_inode_readahead_blks) { 1575 if (arg > (1 << 30)) 1576 return -1; 1577 if (arg && !is_power_of_2(arg)) { 1578 ext4_msg(sb, KERN_ERR, 1579 "EXT4-fs: inode_readahead_blks" 1580 " must be a power of 2"); 1581 return -1; 1582 } 1583 sbi->s_inode_readahead_blks = arg; 1584 } else if (token == Opt_init_itable) { 1585 set_opt(sb, INIT_INODE_TABLE); 1586 if (!args->from) 1587 arg = EXT4_DEF_LI_WAIT_MULT; 1588 sbi->s_li_wait_mult = arg; 1589 } else if (token == Opt_stripe) { 1590 sbi->s_stripe = arg; 1591 } else if (m->flags & MOPT_DATAJ) { 1592 if (is_remount) { 1593 if (!sbi->s_journal) 1594 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1595 else if (test_opt(sb, DATA_FLAGS) != 1596 m->mount_opt) { 1597 ext4_msg(sb, KERN_ERR, 1598 "Cannot change data mode on remount"); 1599 return -1; 1600 } 1601 } else { 1602 clear_opt(sb, DATA_FLAGS); 1603 sbi->s_mount_opt |= m->mount_opt; 1604 } 1605 #ifdef CONFIG_QUOTA 1606 } else if (m->flags & MOPT_QFMT) { 1607 if (sb_any_quota_loaded(sb) && 1608 sbi->s_jquota_fmt != m->mount_opt) { 1609 ext4_msg(sb, KERN_ERR, "Cannot " 1610 "change journaled quota options " 1611 "when quota turned on"); 1612 return -1; 1613 } 1614 sbi->s_jquota_fmt = m->mount_opt; 1615 #endif 1616 } else { 1617 if (!args->from) 1618 arg = 1; 1619 if (m->flags & MOPT_CLEAR) 1620 arg = !arg; 1621 else if (unlikely(!(m->flags & MOPT_SET))) { 1622 ext4_msg(sb, KERN_WARNING, 1623 "buggy handling of option %s", opt); 1624 WARN_ON(1); 1625 return -1; 1626 } 1627 if (arg != 0) 1628 sbi->s_mount_opt |= m->mount_opt; 1629 else 1630 sbi->s_mount_opt &= ~m->mount_opt; 1631 } 1632 return 1; 1633 } 1634 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1635 "or missing value", opt); 1636 return -1; 1637 } 1638 1639 static int parse_options(char *options, struct super_block *sb, 1640 unsigned long *journal_devnum, 1641 unsigned int *journal_ioprio, 1642 int is_remount) 1643 { 1644 #ifdef CONFIG_QUOTA 1645 struct ext4_sb_info *sbi = EXT4_SB(sb); 1646 #endif 1647 char *p; 1648 substring_t args[MAX_OPT_ARGS]; 1649 int token; 1650 1651 if (!options) 1652 return 1; 1653 1654 while ((p = strsep(&options, ",")) != NULL) { 1655 if (!*p) 1656 continue; 1657 /* 1658 * Initialize args struct so we know whether arg was 1659 * found; some options take optional arguments. 1660 */ 1661 args[0].to = args[0].from = 0; 1662 token = match_token(p, tokens, args); 1663 if (handle_mount_opt(sb, p, token, args, journal_devnum, 1664 journal_ioprio, is_remount) < 0) 1665 return 0; 1666 } 1667 #ifdef CONFIG_QUOTA 1668 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1669 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1670 clear_opt(sb, USRQUOTA); 1671 1672 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1673 clear_opt(sb, GRPQUOTA); 1674 1675 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1676 ext4_msg(sb, KERN_ERR, "old and new quota " 1677 "format mixing"); 1678 return 0; 1679 } 1680 1681 if (!sbi->s_jquota_fmt) { 1682 ext4_msg(sb, KERN_ERR, "journaled quota format " 1683 "not specified"); 1684 return 0; 1685 } 1686 } else { 1687 if (sbi->s_jquota_fmt) { 1688 ext4_msg(sb, KERN_ERR, "journaled quota format " 1689 "specified with no journaling " 1690 "enabled"); 1691 return 0; 1692 } 1693 } 1694 #endif 1695 return 1; 1696 } 1697 1698 static inline void ext4_show_quota_options(struct seq_file *seq, 1699 struct super_block *sb) 1700 { 1701 #if defined(CONFIG_QUOTA) 1702 struct ext4_sb_info *sbi = EXT4_SB(sb); 1703 1704 if (sbi->s_jquota_fmt) { 1705 char *fmtname = ""; 1706 1707 switch (sbi->s_jquota_fmt) { 1708 case QFMT_VFS_OLD: 1709 fmtname = "vfsold"; 1710 break; 1711 case QFMT_VFS_V0: 1712 fmtname = "vfsv0"; 1713 break; 1714 case QFMT_VFS_V1: 1715 fmtname = "vfsv1"; 1716 break; 1717 } 1718 seq_printf(seq, ",jqfmt=%s", fmtname); 1719 } 1720 1721 if (sbi->s_qf_names[USRQUOTA]) 1722 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); 1723 1724 if (sbi->s_qf_names[GRPQUOTA]) 1725 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); 1726 1727 if (test_opt(sb, USRQUOTA)) 1728 seq_puts(seq, ",usrquota"); 1729 1730 if (test_opt(sb, GRPQUOTA)) 1731 seq_puts(seq, ",grpquota"); 1732 #endif 1733 } 1734 1735 static const char *token2str(int token) 1736 { 1737 static const struct match_token *t; 1738 1739 for (t = tokens; t->token != Opt_err; t++) 1740 if (t->token == token && !strchr(t->pattern, '=')) 1741 break; 1742 return t->pattern; 1743 } 1744 1745 /* 1746 * Show an option if 1747 * - it's set to a non-default value OR 1748 * - if the per-sb default is different from the global default 1749 */ 1750 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 1751 int nodefs) 1752 { 1753 struct ext4_sb_info *sbi = EXT4_SB(sb); 1754 struct ext4_super_block *es = sbi->s_es; 1755 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; 1756 const struct mount_opts *m; 1757 char sep = nodefs ? '\n' : ','; 1758 1759 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 1760 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 1761 1762 if (sbi->s_sb_block != 1) 1763 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 1764 1765 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 1766 int want_set = m->flags & MOPT_SET; 1767 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 1768 (m->flags & MOPT_CLEAR_ERR)) 1769 continue; 1770 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 1771 continue; /* skip if same as the default */ 1772 if ((want_set && 1773 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 1774 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 1775 continue; /* select Opt_noFoo vs Opt_Foo */ 1776 SEQ_OPTS_PRINT("%s", token2str(m->token)); 1777 } 1778 1779 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 1780 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 1781 SEQ_OPTS_PRINT("resuid=%u", 1782 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 1783 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 1784 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 1785 SEQ_OPTS_PRINT("resgid=%u", 1786 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 1787 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 1788 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 1789 SEQ_OPTS_PUTS("errors=remount-ro"); 1790 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 1791 SEQ_OPTS_PUTS("errors=continue"); 1792 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 1793 SEQ_OPTS_PUTS("errors=panic"); 1794 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 1795 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 1796 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 1797 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 1798 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 1799 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 1800 if (sb->s_flags & MS_I_VERSION) 1801 SEQ_OPTS_PUTS("i_version"); 1802 if (nodefs || sbi->s_stripe) 1803 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 1804 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { 1805 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 1806 SEQ_OPTS_PUTS("data=journal"); 1807 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 1808 SEQ_OPTS_PUTS("data=ordered"); 1809 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 1810 SEQ_OPTS_PUTS("data=writeback"); 1811 } 1812 if (nodefs || 1813 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 1814 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 1815 sbi->s_inode_readahead_blks); 1816 1817 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && 1818 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 1819 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 1820 1821 ext4_show_quota_options(seq, sb); 1822 return 0; 1823 } 1824 1825 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 1826 { 1827 return _ext4_show_options(seq, root->d_sb, 0); 1828 } 1829 1830 static int options_seq_show(struct seq_file *seq, void *offset) 1831 { 1832 struct super_block *sb = seq->private; 1833 int rc; 1834 1835 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw"); 1836 rc = _ext4_show_options(seq, sb, 1); 1837 seq_puts(seq, "\n"); 1838 return rc; 1839 } 1840 1841 static int options_open_fs(struct inode *inode, struct file *file) 1842 { 1843 return single_open(file, options_seq_show, PDE(inode)->data); 1844 } 1845 1846 static const struct file_operations ext4_seq_options_fops = { 1847 .owner = THIS_MODULE, 1848 .open = options_open_fs, 1849 .read = seq_read, 1850 .llseek = seq_lseek, 1851 .release = single_release, 1852 }; 1853 1854 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 1855 int read_only) 1856 { 1857 struct ext4_sb_info *sbi = EXT4_SB(sb); 1858 int res = 0; 1859 1860 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 1861 ext4_msg(sb, KERN_ERR, "revision level too high, " 1862 "forcing read-only mode"); 1863 res = MS_RDONLY; 1864 } 1865 if (read_only) 1866 goto done; 1867 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1868 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1869 "running e2fsck is recommended"); 1870 else if ((sbi->s_mount_state & EXT4_ERROR_FS)) 1871 ext4_msg(sb, KERN_WARNING, 1872 "warning: mounting fs with errors, " 1873 "running e2fsck is recommended"); 1874 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 1875 le16_to_cpu(es->s_mnt_count) >= 1876 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 1877 ext4_msg(sb, KERN_WARNING, 1878 "warning: maximal mount count reached, " 1879 "running e2fsck is recommended"); 1880 else if (le32_to_cpu(es->s_checkinterval) && 1881 (le32_to_cpu(es->s_lastcheck) + 1882 le32_to_cpu(es->s_checkinterval) <= get_seconds())) 1883 ext4_msg(sb, KERN_WARNING, 1884 "warning: checktime reached, " 1885 "running e2fsck is recommended"); 1886 if (!sbi->s_journal) 1887 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 1888 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1889 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1890 le16_add_cpu(&es->s_mnt_count, 1); 1891 es->s_mtime = cpu_to_le32(get_seconds()); 1892 ext4_update_dynamic_rev(sb); 1893 if (sbi->s_journal) 1894 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1895 1896 ext4_commit_super(sb, 1); 1897 done: 1898 if (test_opt(sb, DEBUG)) 1899 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1900 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1901 sb->s_blocksize, 1902 sbi->s_groups_count, 1903 EXT4_BLOCKS_PER_GROUP(sb), 1904 EXT4_INODES_PER_GROUP(sb), 1905 sbi->s_mount_opt, sbi->s_mount_opt2); 1906 1907 cleancache_init_fs(sb); 1908 return res; 1909 } 1910 1911 static int ext4_fill_flex_info(struct super_block *sb) 1912 { 1913 struct ext4_sb_info *sbi = EXT4_SB(sb); 1914 struct ext4_group_desc *gdp = NULL; 1915 ext4_group_t flex_group_count; 1916 ext4_group_t flex_group; 1917 unsigned int groups_per_flex = 0; 1918 size_t size; 1919 int i; 1920 1921 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 1922 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 1923 sbi->s_log_groups_per_flex = 0; 1924 return 1; 1925 } 1926 groups_per_flex = 1 << sbi->s_log_groups_per_flex; 1927 1928 /* We allocate both existing and potentially added groups */ 1929 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + 1930 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << 1931 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; 1932 size = flex_group_count * sizeof(struct flex_groups); 1933 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL); 1934 if (sbi->s_flex_groups == NULL) { 1935 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups", 1936 flex_group_count); 1937 goto failed; 1938 } 1939 1940 for (i = 0; i < sbi->s_groups_count; i++) { 1941 gdp = ext4_get_group_desc(sb, i, NULL); 1942 1943 flex_group = ext4_flex_group(sbi, i); 1944 atomic_add(ext4_free_inodes_count(sb, gdp), 1945 &sbi->s_flex_groups[flex_group].free_inodes); 1946 atomic_add(ext4_free_group_clusters(sb, gdp), 1947 &sbi->s_flex_groups[flex_group].free_clusters); 1948 atomic_add(ext4_used_dirs_count(sb, gdp), 1949 &sbi->s_flex_groups[flex_group].used_dirs); 1950 } 1951 1952 return 1; 1953 failed: 1954 return 0; 1955 } 1956 1957 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, 1958 struct ext4_group_desc *gdp) 1959 { 1960 int offset; 1961 __u16 crc = 0; 1962 __le32 le_group = cpu_to_le32(block_group); 1963 1964 if ((sbi->s_es->s_feature_ro_compat & 1965 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) { 1966 /* Use new metadata_csum algorithm */ 1967 __u16 old_csum; 1968 __u32 csum32; 1969 1970 old_csum = gdp->bg_checksum; 1971 gdp->bg_checksum = 0; 1972 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 1973 sizeof(le_group)); 1974 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, 1975 sbi->s_desc_size); 1976 gdp->bg_checksum = old_csum; 1977 1978 crc = csum32 & 0xFFFF; 1979 goto out; 1980 } 1981 1982 /* old crc16 code */ 1983 offset = offsetof(struct ext4_group_desc, bg_checksum); 1984 1985 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 1986 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 1987 crc = crc16(crc, (__u8 *)gdp, offset); 1988 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 1989 /* for checksum of struct ext4_group_desc do the rest...*/ 1990 if ((sbi->s_es->s_feature_incompat & 1991 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && 1992 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 1993 crc = crc16(crc, (__u8 *)gdp + offset, 1994 le16_to_cpu(sbi->s_es->s_desc_size) - 1995 offset); 1996 1997 out: 1998 return cpu_to_le16(crc); 1999 } 2000 2001 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2002 struct ext4_group_desc *gdp) 2003 { 2004 if (ext4_has_group_desc_csum(sb) && 2005 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb), 2006 block_group, gdp))) 2007 return 0; 2008 2009 return 1; 2010 } 2011 2012 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2013 struct ext4_group_desc *gdp) 2014 { 2015 if (!ext4_has_group_desc_csum(sb)) 2016 return; 2017 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp); 2018 } 2019 2020 /* Called at mount-time, super-block is locked */ 2021 static int ext4_check_descriptors(struct super_block *sb, 2022 ext4_group_t *first_not_zeroed) 2023 { 2024 struct ext4_sb_info *sbi = EXT4_SB(sb); 2025 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2026 ext4_fsblk_t last_block; 2027 ext4_fsblk_t block_bitmap; 2028 ext4_fsblk_t inode_bitmap; 2029 ext4_fsblk_t inode_table; 2030 int flexbg_flag = 0; 2031 ext4_group_t i, grp = sbi->s_groups_count; 2032 2033 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 2034 flexbg_flag = 1; 2035 2036 ext4_debug("Checking group descriptors"); 2037 2038 for (i = 0; i < sbi->s_groups_count; i++) { 2039 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2040 2041 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2042 last_block = ext4_blocks_count(sbi->s_es) - 1; 2043 else 2044 last_block = first_block + 2045 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2046 2047 if ((grp == sbi->s_groups_count) && 2048 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2049 grp = i; 2050 2051 block_bitmap = ext4_block_bitmap(sb, gdp); 2052 if (block_bitmap < first_block || block_bitmap > last_block) { 2053 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2054 "Block bitmap for group %u not in group " 2055 "(block %llu)!", i, block_bitmap); 2056 return 0; 2057 } 2058 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2059 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2060 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2061 "Inode bitmap for group %u not in group " 2062 "(block %llu)!", i, inode_bitmap); 2063 return 0; 2064 } 2065 inode_table = ext4_inode_table(sb, gdp); 2066 if (inode_table < first_block || 2067 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2068 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2069 "Inode table for group %u not in group " 2070 "(block %llu)!", i, inode_table); 2071 return 0; 2072 } 2073 ext4_lock_group(sb, i); 2074 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2075 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2076 "Checksum for group %u failed (%u!=%u)", 2077 i, le16_to_cpu(ext4_group_desc_csum(sbi, i, 2078 gdp)), le16_to_cpu(gdp->bg_checksum)); 2079 if (!(sb->s_flags & MS_RDONLY)) { 2080 ext4_unlock_group(sb, i); 2081 return 0; 2082 } 2083 } 2084 ext4_unlock_group(sb, i); 2085 if (!flexbg_flag) 2086 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2087 } 2088 if (NULL != first_not_zeroed) 2089 *first_not_zeroed = grp; 2090 2091 ext4_free_blocks_count_set(sbi->s_es, 2092 EXT4_C2B(sbi, ext4_count_free_clusters(sb))); 2093 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2094 return 1; 2095 } 2096 2097 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2098 * the superblock) which were deleted from all directories, but held open by 2099 * a process at the time of a crash. We walk the list and try to delete these 2100 * inodes at recovery time (only with a read-write filesystem). 2101 * 2102 * In order to keep the orphan inode chain consistent during traversal (in 2103 * case of crash during recovery), we link each inode into the superblock 2104 * orphan list_head and handle it the same way as an inode deletion during 2105 * normal operation (which journals the operations for us). 2106 * 2107 * We only do an iget() and an iput() on each inode, which is very safe if we 2108 * accidentally point at an in-use or already deleted inode. The worst that 2109 * can happen in this case is that we get a "bit already cleared" message from 2110 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2111 * e2fsck was run on this filesystem, and it must have already done the orphan 2112 * inode cleanup for us, so we can safely abort without any further action. 2113 */ 2114 static void ext4_orphan_cleanup(struct super_block *sb, 2115 struct ext4_super_block *es) 2116 { 2117 unsigned int s_flags = sb->s_flags; 2118 int nr_orphans = 0, nr_truncates = 0; 2119 #ifdef CONFIG_QUOTA 2120 int i; 2121 #endif 2122 if (!es->s_last_orphan) { 2123 jbd_debug(4, "no orphan inodes to clean up\n"); 2124 return; 2125 } 2126 2127 if (bdev_read_only(sb->s_bdev)) { 2128 ext4_msg(sb, KERN_ERR, "write access " 2129 "unavailable, skipping orphan cleanup"); 2130 return; 2131 } 2132 2133 /* Check if feature set would not allow a r/w mount */ 2134 if (!ext4_feature_set_ok(sb, 0)) { 2135 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2136 "unknown ROCOMPAT features"); 2137 return; 2138 } 2139 2140 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2141 if (es->s_last_orphan) 2142 jbd_debug(1, "Errors on filesystem, " 2143 "clearing orphan list.\n"); 2144 es->s_last_orphan = 0; 2145 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2146 return; 2147 } 2148 2149 if (s_flags & MS_RDONLY) { 2150 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2151 sb->s_flags &= ~MS_RDONLY; 2152 } 2153 #ifdef CONFIG_QUOTA 2154 /* Needed for iput() to work correctly and not trash data */ 2155 sb->s_flags |= MS_ACTIVE; 2156 /* Turn on quotas so that they are updated correctly */ 2157 for (i = 0; i < MAXQUOTAS; i++) { 2158 if (EXT4_SB(sb)->s_qf_names[i]) { 2159 int ret = ext4_quota_on_mount(sb, i); 2160 if (ret < 0) 2161 ext4_msg(sb, KERN_ERR, 2162 "Cannot turn on journaled " 2163 "quota: error %d", ret); 2164 } 2165 } 2166 #endif 2167 2168 while (es->s_last_orphan) { 2169 struct inode *inode; 2170 2171 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2172 if (IS_ERR(inode)) { 2173 es->s_last_orphan = 0; 2174 break; 2175 } 2176 2177 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2178 dquot_initialize(inode); 2179 if (inode->i_nlink) { 2180 ext4_msg(sb, KERN_DEBUG, 2181 "%s: truncating inode %lu to %lld bytes", 2182 __func__, inode->i_ino, inode->i_size); 2183 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2184 inode->i_ino, inode->i_size); 2185 ext4_truncate(inode); 2186 nr_truncates++; 2187 } else { 2188 ext4_msg(sb, KERN_DEBUG, 2189 "%s: deleting unreferenced inode %lu", 2190 __func__, inode->i_ino); 2191 jbd_debug(2, "deleting unreferenced inode %lu\n", 2192 inode->i_ino); 2193 nr_orphans++; 2194 } 2195 iput(inode); /* The delete magic happens here! */ 2196 } 2197 2198 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2199 2200 if (nr_orphans) 2201 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2202 PLURAL(nr_orphans)); 2203 if (nr_truncates) 2204 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2205 PLURAL(nr_truncates)); 2206 #ifdef CONFIG_QUOTA 2207 /* Turn quotas off */ 2208 for (i = 0; i < MAXQUOTAS; i++) { 2209 if (sb_dqopt(sb)->files[i]) 2210 dquot_quota_off(sb, i); 2211 } 2212 #endif 2213 sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2214 } 2215 2216 /* 2217 * Maximal extent format file size. 2218 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2219 * extent format containers, within a sector_t, and within i_blocks 2220 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2221 * so that won't be a limiting factor. 2222 * 2223 * However there is other limiting factor. We do store extents in the form 2224 * of starting block and length, hence the resulting length of the extent 2225 * covering maximum file size must fit into on-disk format containers as 2226 * well. Given that length is always by 1 unit bigger than max unit (because 2227 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2228 * 2229 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2230 */ 2231 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2232 { 2233 loff_t res; 2234 loff_t upper_limit = MAX_LFS_FILESIZE; 2235 2236 /* small i_blocks in vfs inode? */ 2237 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2238 /* 2239 * CONFIG_LBDAF is not enabled implies the inode 2240 * i_block represent total blocks in 512 bytes 2241 * 32 == size of vfs inode i_blocks * 8 2242 */ 2243 upper_limit = (1LL << 32) - 1; 2244 2245 /* total blocks in file system block size */ 2246 upper_limit >>= (blkbits - 9); 2247 upper_limit <<= blkbits; 2248 } 2249 2250 /* 2251 * 32-bit extent-start container, ee_block. We lower the maxbytes 2252 * by one fs block, so ee_len can cover the extent of maximum file 2253 * size 2254 */ 2255 res = (1LL << 32) - 1; 2256 res <<= blkbits; 2257 2258 /* Sanity check against vm- & vfs- imposed limits */ 2259 if (res > upper_limit) 2260 res = upper_limit; 2261 2262 return res; 2263 } 2264 2265 /* 2266 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2267 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2268 * We need to be 1 filesystem block less than the 2^48 sector limit. 2269 */ 2270 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2271 { 2272 loff_t res = EXT4_NDIR_BLOCKS; 2273 int meta_blocks; 2274 loff_t upper_limit; 2275 /* This is calculated to be the largest file size for a dense, block 2276 * mapped file such that the file's total number of 512-byte sectors, 2277 * including data and all indirect blocks, does not exceed (2^48 - 1). 2278 * 2279 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2280 * number of 512-byte sectors of the file. 2281 */ 2282 2283 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2284 /* 2285 * !has_huge_files or CONFIG_LBDAF not enabled implies that 2286 * the inode i_block field represents total file blocks in 2287 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 2288 */ 2289 upper_limit = (1LL << 32) - 1; 2290 2291 /* total blocks in file system block size */ 2292 upper_limit >>= (bits - 9); 2293 2294 } else { 2295 /* 2296 * We use 48 bit ext4_inode i_blocks 2297 * With EXT4_HUGE_FILE_FL set the i_blocks 2298 * represent total number of blocks in 2299 * file system block size 2300 */ 2301 upper_limit = (1LL << 48) - 1; 2302 2303 } 2304 2305 /* indirect blocks */ 2306 meta_blocks = 1; 2307 /* double indirect blocks */ 2308 meta_blocks += 1 + (1LL << (bits-2)); 2309 /* tripple indirect blocks */ 2310 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2311 2312 upper_limit -= meta_blocks; 2313 upper_limit <<= bits; 2314 2315 res += 1LL << (bits-2); 2316 res += 1LL << (2*(bits-2)); 2317 res += 1LL << (3*(bits-2)); 2318 res <<= bits; 2319 if (res > upper_limit) 2320 res = upper_limit; 2321 2322 if (res > MAX_LFS_FILESIZE) 2323 res = MAX_LFS_FILESIZE; 2324 2325 return res; 2326 } 2327 2328 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2329 ext4_fsblk_t logical_sb_block, int nr) 2330 { 2331 struct ext4_sb_info *sbi = EXT4_SB(sb); 2332 ext4_group_t bg, first_meta_bg; 2333 int has_super = 0; 2334 2335 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2336 2337 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || 2338 nr < first_meta_bg) 2339 return logical_sb_block + nr + 1; 2340 bg = sbi->s_desc_per_block * nr; 2341 if (ext4_bg_has_super(sb, bg)) 2342 has_super = 1; 2343 2344 return (has_super + ext4_group_first_block_no(sb, bg)); 2345 } 2346 2347 /** 2348 * ext4_get_stripe_size: Get the stripe size. 2349 * @sbi: In memory super block info 2350 * 2351 * If we have specified it via mount option, then 2352 * use the mount option value. If the value specified at mount time is 2353 * greater than the blocks per group use the super block value. 2354 * If the super block value is greater than blocks per group return 0. 2355 * Allocator needs it be less than blocks per group. 2356 * 2357 */ 2358 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2359 { 2360 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2361 unsigned long stripe_width = 2362 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2363 int ret; 2364 2365 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2366 ret = sbi->s_stripe; 2367 else if (stripe_width <= sbi->s_blocks_per_group) 2368 ret = stripe_width; 2369 else if (stride <= sbi->s_blocks_per_group) 2370 ret = stride; 2371 else 2372 ret = 0; 2373 2374 /* 2375 * If the stripe width is 1, this makes no sense and 2376 * we set it to 0 to turn off stripe handling code. 2377 */ 2378 if (ret <= 1) 2379 ret = 0; 2380 2381 return ret; 2382 } 2383 2384 /* sysfs supprt */ 2385 2386 struct ext4_attr { 2387 struct attribute attr; 2388 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); 2389 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, 2390 const char *, size_t); 2391 int offset; 2392 }; 2393 2394 static int parse_strtoul(const char *buf, 2395 unsigned long max, unsigned long *value) 2396 { 2397 char *endp; 2398 2399 *value = simple_strtoul(skip_spaces(buf), &endp, 0); 2400 endp = skip_spaces(endp); 2401 if (*endp || *value > max) 2402 return -EINVAL; 2403 2404 return 0; 2405 } 2406 2407 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, 2408 struct ext4_sb_info *sbi, 2409 char *buf) 2410 { 2411 return snprintf(buf, PAGE_SIZE, "%llu\n", 2412 (s64) EXT4_C2B(sbi, 2413 percpu_counter_sum(&sbi->s_dirtyclusters_counter))); 2414 } 2415 2416 static ssize_t session_write_kbytes_show(struct ext4_attr *a, 2417 struct ext4_sb_info *sbi, char *buf) 2418 { 2419 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2420 2421 if (!sb->s_bdev->bd_part) 2422 return snprintf(buf, PAGE_SIZE, "0\n"); 2423 return snprintf(buf, PAGE_SIZE, "%lu\n", 2424 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2425 sbi->s_sectors_written_start) >> 1); 2426 } 2427 2428 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, 2429 struct ext4_sb_info *sbi, char *buf) 2430 { 2431 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2432 2433 if (!sb->s_bdev->bd_part) 2434 return snprintf(buf, PAGE_SIZE, "0\n"); 2435 return snprintf(buf, PAGE_SIZE, "%llu\n", 2436 (unsigned long long)(sbi->s_kbytes_written + 2437 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2438 EXT4_SB(sb)->s_sectors_written_start) >> 1))); 2439 } 2440 2441 static ssize_t inode_readahead_blks_store(struct ext4_attr *a, 2442 struct ext4_sb_info *sbi, 2443 const char *buf, size_t count) 2444 { 2445 unsigned long t; 2446 2447 if (parse_strtoul(buf, 0x40000000, &t)) 2448 return -EINVAL; 2449 2450 if (t && !is_power_of_2(t)) 2451 return -EINVAL; 2452 2453 sbi->s_inode_readahead_blks = t; 2454 return count; 2455 } 2456 2457 static ssize_t sbi_ui_show(struct ext4_attr *a, 2458 struct ext4_sb_info *sbi, char *buf) 2459 { 2460 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); 2461 2462 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 2463 } 2464 2465 static ssize_t sbi_ui_store(struct ext4_attr *a, 2466 struct ext4_sb_info *sbi, 2467 const char *buf, size_t count) 2468 { 2469 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); 2470 unsigned long t; 2471 2472 if (parse_strtoul(buf, 0xffffffff, &t)) 2473 return -EINVAL; 2474 *ui = t; 2475 return count; 2476 } 2477 2478 static ssize_t trigger_test_error(struct ext4_attr *a, 2479 struct ext4_sb_info *sbi, 2480 const char *buf, size_t count) 2481 { 2482 int len = count; 2483 2484 if (!capable(CAP_SYS_ADMIN)) 2485 return -EPERM; 2486 2487 if (len && buf[len-1] == '\n') 2488 len--; 2489 2490 if (len) 2491 ext4_error(sbi->s_sb, "%.*s", len, buf); 2492 return count; 2493 } 2494 2495 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ 2496 static struct ext4_attr ext4_attr_##_name = { \ 2497 .attr = {.name = __stringify(_name), .mode = _mode }, \ 2498 .show = _show, \ 2499 .store = _store, \ 2500 .offset = offsetof(struct ext4_sb_info, _elname), \ 2501 } 2502 #define EXT4_ATTR(name, mode, show, store) \ 2503 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) 2504 2505 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) 2506 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) 2507 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) 2508 #define EXT4_RW_ATTR_SBI_UI(name, elname) \ 2509 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) 2510 #define ATTR_LIST(name) &ext4_attr_##name.attr 2511 2512 EXT4_RO_ATTR(delayed_allocation_blocks); 2513 EXT4_RO_ATTR(session_write_kbytes); 2514 EXT4_RO_ATTR(lifetime_write_kbytes); 2515 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, 2516 inode_readahead_blks_store, s_inode_readahead_blks); 2517 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); 2518 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); 2519 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); 2520 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); 2521 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); 2522 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); 2523 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); 2524 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); 2525 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error); 2526 2527 static struct attribute *ext4_attrs[] = { 2528 ATTR_LIST(delayed_allocation_blocks), 2529 ATTR_LIST(session_write_kbytes), 2530 ATTR_LIST(lifetime_write_kbytes), 2531 ATTR_LIST(inode_readahead_blks), 2532 ATTR_LIST(inode_goal), 2533 ATTR_LIST(mb_stats), 2534 ATTR_LIST(mb_max_to_scan), 2535 ATTR_LIST(mb_min_to_scan), 2536 ATTR_LIST(mb_order2_req), 2537 ATTR_LIST(mb_stream_req), 2538 ATTR_LIST(mb_group_prealloc), 2539 ATTR_LIST(max_writeback_mb_bump), 2540 ATTR_LIST(trigger_fs_error), 2541 NULL, 2542 }; 2543 2544 /* Features this copy of ext4 supports */ 2545 EXT4_INFO_ATTR(lazy_itable_init); 2546 EXT4_INFO_ATTR(batched_discard); 2547 2548 static struct attribute *ext4_feat_attrs[] = { 2549 ATTR_LIST(lazy_itable_init), 2550 ATTR_LIST(batched_discard), 2551 NULL, 2552 }; 2553 2554 static ssize_t ext4_attr_show(struct kobject *kobj, 2555 struct attribute *attr, char *buf) 2556 { 2557 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2558 s_kobj); 2559 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2560 2561 return a->show ? a->show(a, sbi, buf) : 0; 2562 } 2563 2564 static ssize_t ext4_attr_store(struct kobject *kobj, 2565 struct attribute *attr, 2566 const char *buf, size_t len) 2567 { 2568 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2569 s_kobj); 2570 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2571 2572 return a->store ? a->store(a, sbi, buf, len) : 0; 2573 } 2574 2575 static void ext4_sb_release(struct kobject *kobj) 2576 { 2577 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2578 s_kobj); 2579 complete(&sbi->s_kobj_unregister); 2580 } 2581 2582 static const struct sysfs_ops ext4_attr_ops = { 2583 .show = ext4_attr_show, 2584 .store = ext4_attr_store, 2585 }; 2586 2587 static struct kobj_type ext4_ktype = { 2588 .default_attrs = ext4_attrs, 2589 .sysfs_ops = &ext4_attr_ops, 2590 .release = ext4_sb_release, 2591 }; 2592 2593 static void ext4_feat_release(struct kobject *kobj) 2594 { 2595 complete(&ext4_feat->f_kobj_unregister); 2596 } 2597 2598 static struct kobj_type ext4_feat_ktype = { 2599 .default_attrs = ext4_feat_attrs, 2600 .sysfs_ops = &ext4_attr_ops, 2601 .release = ext4_feat_release, 2602 }; 2603 2604 /* 2605 * Check whether this filesystem can be mounted based on 2606 * the features present and the RDONLY/RDWR mount requested. 2607 * Returns 1 if this filesystem can be mounted as requested, 2608 * 0 if it cannot be. 2609 */ 2610 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2611 { 2612 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { 2613 ext4_msg(sb, KERN_ERR, 2614 "Couldn't mount because of " 2615 "unsupported optional features (%x)", 2616 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2617 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2618 return 0; 2619 } 2620 2621 if (readonly) 2622 return 1; 2623 2624 /* Check that feature set is OK for a read-write mount */ 2625 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { 2626 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2627 "unsupported optional features (%x)", 2628 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2629 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2630 return 0; 2631 } 2632 /* 2633 * Large file size enabled file system can only be mounted 2634 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2635 */ 2636 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { 2637 if (sizeof(blkcnt_t) < sizeof(u64)) { 2638 ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2639 "cannot be mounted RDWR without " 2640 "CONFIG_LBDAF"); 2641 return 0; 2642 } 2643 } 2644 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) && 2645 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { 2646 ext4_msg(sb, KERN_ERR, 2647 "Can't support bigalloc feature without " 2648 "extents feature\n"); 2649 return 0; 2650 } 2651 2652 #ifndef CONFIG_QUOTA 2653 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 2654 !readonly) { 2655 ext4_msg(sb, KERN_ERR, 2656 "Filesystem with quota feature cannot be mounted RDWR " 2657 "without CONFIG_QUOTA"); 2658 return 0; 2659 } 2660 #endif /* CONFIG_QUOTA */ 2661 return 1; 2662 } 2663 2664 /* 2665 * This function is called once a day if we have errors logged 2666 * on the file system 2667 */ 2668 static void print_daily_error_info(unsigned long arg) 2669 { 2670 struct super_block *sb = (struct super_block *) arg; 2671 struct ext4_sb_info *sbi; 2672 struct ext4_super_block *es; 2673 2674 sbi = EXT4_SB(sb); 2675 es = sbi->s_es; 2676 2677 if (es->s_error_count) 2678 ext4_msg(sb, KERN_NOTICE, "error count: %u", 2679 le32_to_cpu(es->s_error_count)); 2680 if (es->s_first_error_time) { 2681 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", 2682 sb->s_id, le32_to_cpu(es->s_first_error_time), 2683 (int) sizeof(es->s_first_error_func), 2684 es->s_first_error_func, 2685 le32_to_cpu(es->s_first_error_line)); 2686 if (es->s_first_error_ino) 2687 printk(": inode %u", 2688 le32_to_cpu(es->s_first_error_ino)); 2689 if (es->s_first_error_block) 2690 printk(": block %llu", (unsigned long long) 2691 le64_to_cpu(es->s_first_error_block)); 2692 printk("\n"); 2693 } 2694 if (es->s_last_error_time) { 2695 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", 2696 sb->s_id, le32_to_cpu(es->s_last_error_time), 2697 (int) sizeof(es->s_last_error_func), 2698 es->s_last_error_func, 2699 le32_to_cpu(es->s_last_error_line)); 2700 if (es->s_last_error_ino) 2701 printk(": inode %u", 2702 le32_to_cpu(es->s_last_error_ino)); 2703 if (es->s_last_error_block) 2704 printk(": block %llu", (unsigned long long) 2705 le64_to_cpu(es->s_last_error_block)); 2706 printk("\n"); 2707 } 2708 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 2709 } 2710 2711 /* Find next suitable group and run ext4_init_inode_table */ 2712 static int ext4_run_li_request(struct ext4_li_request *elr) 2713 { 2714 struct ext4_group_desc *gdp = NULL; 2715 ext4_group_t group, ngroups; 2716 struct super_block *sb; 2717 unsigned long timeout = 0; 2718 int ret = 0; 2719 2720 sb = elr->lr_super; 2721 ngroups = EXT4_SB(sb)->s_groups_count; 2722 2723 sb_start_write(sb); 2724 for (group = elr->lr_next_group; group < ngroups; group++) { 2725 gdp = ext4_get_group_desc(sb, group, NULL); 2726 if (!gdp) { 2727 ret = 1; 2728 break; 2729 } 2730 2731 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2732 break; 2733 } 2734 2735 if (group == ngroups) 2736 ret = 1; 2737 2738 if (!ret) { 2739 timeout = jiffies; 2740 ret = ext4_init_inode_table(sb, group, 2741 elr->lr_timeout ? 0 : 1); 2742 if (elr->lr_timeout == 0) { 2743 timeout = (jiffies - timeout) * 2744 elr->lr_sbi->s_li_wait_mult; 2745 elr->lr_timeout = timeout; 2746 } 2747 elr->lr_next_sched = jiffies + elr->lr_timeout; 2748 elr->lr_next_group = group + 1; 2749 } 2750 sb_end_write(sb); 2751 2752 return ret; 2753 } 2754 2755 /* 2756 * Remove lr_request from the list_request and free the 2757 * request structure. Should be called with li_list_mtx held 2758 */ 2759 static void ext4_remove_li_request(struct ext4_li_request *elr) 2760 { 2761 struct ext4_sb_info *sbi; 2762 2763 if (!elr) 2764 return; 2765 2766 sbi = elr->lr_sbi; 2767 2768 list_del(&elr->lr_request); 2769 sbi->s_li_request = NULL; 2770 kfree(elr); 2771 } 2772 2773 static void ext4_unregister_li_request(struct super_block *sb) 2774 { 2775 mutex_lock(&ext4_li_mtx); 2776 if (!ext4_li_info) { 2777 mutex_unlock(&ext4_li_mtx); 2778 return; 2779 } 2780 2781 mutex_lock(&ext4_li_info->li_list_mtx); 2782 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2783 mutex_unlock(&ext4_li_info->li_list_mtx); 2784 mutex_unlock(&ext4_li_mtx); 2785 } 2786 2787 static struct task_struct *ext4_lazyinit_task; 2788 2789 /* 2790 * This is the function where ext4lazyinit thread lives. It walks 2791 * through the request list searching for next scheduled filesystem. 2792 * When such a fs is found, run the lazy initialization request 2793 * (ext4_rn_li_request) and keep track of the time spend in this 2794 * function. Based on that time we compute next schedule time of 2795 * the request. When walking through the list is complete, compute 2796 * next waking time and put itself into sleep. 2797 */ 2798 static int ext4_lazyinit_thread(void *arg) 2799 { 2800 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2801 struct list_head *pos, *n; 2802 struct ext4_li_request *elr; 2803 unsigned long next_wakeup, cur; 2804 2805 BUG_ON(NULL == eli); 2806 2807 cont_thread: 2808 while (true) { 2809 next_wakeup = MAX_JIFFY_OFFSET; 2810 2811 mutex_lock(&eli->li_list_mtx); 2812 if (list_empty(&eli->li_request_list)) { 2813 mutex_unlock(&eli->li_list_mtx); 2814 goto exit_thread; 2815 } 2816 2817 list_for_each_safe(pos, n, &eli->li_request_list) { 2818 elr = list_entry(pos, struct ext4_li_request, 2819 lr_request); 2820 2821 if (time_after_eq(jiffies, elr->lr_next_sched)) { 2822 if (ext4_run_li_request(elr) != 0) { 2823 /* error, remove the lazy_init job */ 2824 ext4_remove_li_request(elr); 2825 continue; 2826 } 2827 } 2828 2829 if (time_before(elr->lr_next_sched, next_wakeup)) 2830 next_wakeup = elr->lr_next_sched; 2831 } 2832 mutex_unlock(&eli->li_list_mtx); 2833 2834 try_to_freeze(); 2835 2836 cur = jiffies; 2837 if ((time_after_eq(cur, next_wakeup)) || 2838 (MAX_JIFFY_OFFSET == next_wakeup)) { 2839 cond_resched(); 2840 continue; 2841 } 2842 2843 schedule_timeout_interruptible(next_wakeup - cur); 2844 2845 if (kthread_should_stop()) { 2846 ext4_clear_request_list(); 2847 goto exit_thread; 2848 } 2849 } 2850 2851 exit_thread: 2852 /* 2853 * It looks like the request list is empty, but we need 2854 * to check it under the li_list_mtx lock, to prevent any 2855 * additions into it, and of course we should lock ext4_li_mtx 2856 * to atomically free the list and ext4_li_info, because at 2857 * this point another ext4 filesystem could be registering 2858 * new one. 2859 */ 2860 mutex_lock(&ext4_li_mtx); 2861 mutex_lock(&eli->li_list_mtx); 2862 if (!list_empty(&eli->li_request_list)) { 2863 mutex_unlock(&eli->li_list_mtx); 2864 mutex_unlock(&ext4_li_mtx); 2865 goto cont_thread; 2866 } 2867 mutex_unlock(&eli->li_list_mtx); 2868 kfree(ext4_li_info); 2869 ext4_li_info = NULL; 2870 mutex_unlock(&ext4_li_mtx); 2871 2872 return 0; 2873 } 2874 2875 static void ext4_clear_request_list(void) 2876 { 2877 struct list_head *pos, *n; 2878 struct ext4_li_request *elr; 2879 2880 mutex_lock(&ext4_li_info->li_list_mtx); 2881 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 2882 elr = list_entry(pos, struct ext4_li_request, 2883 lr_request); 2884 ext4_remove_li_request(elr); 2885 } 2886 mutex_unlock(&ext4_li_info->li_list_mtx); 2887 } 2888 2889 static int ext4_run_lazyinit_thread(void) 2890 { 2891 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 2892 ext4_li_info, "ext4lazyinit"); 2893 if (IS_ERR(ext4_lazyinit_task)) { 2894 int err = PTR_ERR(ext4_lazyinit_task); 2895 ext4_clear_request_list(); 2896 kfree(ext4_li_info); 2897 ext4_li_info = NULL; 2898 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 2899 "initialization thread\n", 2900 err); 2901 return err; 2902 } 2903 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 2904 return 0; 2905 } 2906 2907 /* 2908 * Check whether it make sense to run itable init. thread or not. 2909 * If there is at least one uninitialized inode table, return 2910 * corresponding group number, else the loop goes through all 2911 * groups and return total number of groups. 2912 */ 2913 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 2914 { 2915 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 2916 struct ext4_group_desc *gdp = NULL; 2917 2918 for (group = 0; group < ngroups; group++) { 2919 gdp = ext4_get_group_desc(sb, group, NULL); 2920 if (!gdp) 2921 continue; 2922 2923 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2924 break; 2925 } 2926 2927 return group; 2928 } 2929 2930 static int ext4_li_info_new(void) 2931 { 2932 struct ext4_lazy_init *eli = NULL; 2933 2934 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 2935 if (!eli) 2936 return -ENOMEM; 2937 2938 INIT_LIST_HEAD(&eli->li_request_list); 2939 mutex_init(&eli->li_list_mtx); 2940 2941 eli->li_state |= EXT4_LAZYINIT_QUIT; 2942 2943 ext4_li_info = eli; 2944 2945 return 0; 2946 } 2947 2948 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 2949 ext4_group_t start) 2950 { 2951 struct ext4_sb_info *sbi = EXT4_SB(sb); 2952 struct ext4_li_request *elr; 2953 unsigned long rnd; 2954 2955 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 2956 if (!elr) 2957 return NULL; 2958 2959 elr->lr_super = sb; 2960 elr->lr_sbi = sbi; 2961 elr->lr_next_group = start; 2962 2963 /* 2964 * Randomize first schedule time of the request to 2965 * spread the inode table initialization requests 2966 * better. 2967 */ 2968 get_random_bytes(&rnd, sizeof(rnd)); 2969 elr->lr_next_sched = jiffies + (unsigned long)rnd % 2970 (EXT4_DEF_LI_MAX_START_DELAY * HZ); 2971 2972 return elr; 2973 } 2974 2975 static int ext4_register_li_request(struct super_block *sb, 2976 ext4_group_t first_not_zeroed) 2977 { 2978 struct ext4_sb_info *sbi = EXT4_SB(sb); 2979 struct ext4_li_request *elr; 2980 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 2981 int ret = 0; 2982 2983 if (sbi->s_li_request != NULL) { 2984 /* 2985 * Reset timeout so it can be computed again, because 2986 * s_li_wait_mult might have changed. 2987 */ 2988 sbi->s_li_request->lr_timeout = 0; 2989 return 0; 2990 } 2991 2992 if (first_not_zeroed == ngroups || 2993 (sb->s_flags & MS_RDONLY) || 2994 !test_opt(sb, INIT_INODE_TABLE)) 2995 return 0; 2996 2997 elr = ext4_li_request_new(sb, first_not_zeroed); 2998 if (!elr) 2999 return -ENOMEM; 3000 3001 mutex_lock(&ext4_li_mtx); 3002 3003 if (NULL == ext4_li_info) { 3004 ret = ext4_li_info_new(); 3005 if (ret) 3006 goto out; 3007 } 3008 3009 mutex_lock(&ext4_li_info->li_list_mtx); 3010 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3011 mutex_unlock(&ext4_li_info->li_list_mtx); 3012 3013 sbi->s_li_request = elr; 3014 /* 3015 * set elr to NULL here since it has been inserted to 3016 * the request_list and the removal and free of it is 3017 * handled by ext4_clear_request_list from now on. 3018 */ 3019 elr = NULL; 3020 3021 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3022 ret = ext4_run_lazyinit_thread(); 3023 if (ret) 3024 goto out; 3025 } 3026 out: 3027 mutex_unlock(&ext4_li_mtx); 3028 if (ret) 3029 kfree(elr); 3030 return ret; 3031 } 3032 3033 /* 3034 * We do not need to lock anything since this is called on 3035 * module unload. 3036 */ 3037 static void ext4_destroy_lazyinit_thread(void) 3038 { 3039 /* 3040 * If thread exited earlier 3041 * there's nothing to be done. 3042 */ 3043 if (!ext4_li_info || !ext4_lazyinit_task) 3044 return; 3045 3046 kthread_stop(ext4_lazyinit_task); 3047 } 3048 3049 static int set_journal_csum_feature_set(struct super_block *sb) 3050 { 3051 int ret = 1; 3052 int compat, incompat; 3053 struct ext4_sb_info *sbi = EXT4_SB(sb); 3054 3055 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3056 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 3057 /* journal checksum v2 */ 3058 compat = 0; 3059 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2; 3060 } else { 3061 /* journal checksum v1 */ 3062 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3063 incompat = 0; 3064 } 3065 3066 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3067 ret = jbd2_journal_set_features(sbi->s_journal, 3068 compat, 0, 3069 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3070 incompat); 3071 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3072 ret = jbd2_journal_set_features(sbi->s_journal, 3073 compat, 0, 3074 incompat); 3075 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3076 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3077 } else { 3078 jbd2_journal_clear_features(sbi->s_journal, 3079 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3080 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3081 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3082 } 3083 3084 return ret; 3085 } 3086 3087 /* 3088 * Note: calculating the overhead so we can be compatible with 3089 * historical BSD practice is quite difficult in the face of 3090 * clusters/bigalloc. This is because multiple metadata blocks from 3091 * different block group can end up in the same allocation cluster. 3092 * Calculating the exact overhead in the face of clustered allocation 3093 * requires either O(all block bitmaps) in memory or O(number of block 3094 * groups**2) in time. We will still calculate the superblock for 3095 * older file systems --- and if we come across with a bigalloc file 3096 * system with zero in s_overhead_clusters the estimate will be close to 3097 * correct especially for very large cluster sizes --- but for newer 3098 * file systems, it's better to calculate this figure once at mkfs 3099 * time, and store it in the superblock. If the superblock value is 3100 * present (even for non-bigalloc file systems), we will use it. 3101 */ 3102 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3103 char *buf) 3104 { 3105 struct ext4_sb_info *sbi = EXT4_SB(sb); 3106 struct ext4_group_desc *gdp; 3107 ext4_fsblk_t first_block, last_block, b; 3108 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3109 int s, j, count = 0; 3110 3111 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3112 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3113 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3114 for (i = 0; i < ngroups; i++) { 3115 gdp = ext4_get_group_desc(sb, i, NULL); 3116 b = ext4_block_bitmap(sb, gdp); 3117 if (b >= first_block && b <= last_block) { 3118 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3119 count++; 3120 } 3121 b = ext4_inode_bitmap(sb, gdp); 3122 if (b >= first_block && b <= last_block) { 3123 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3124 count++; 3125 } 3126 b = ext4_inode_table(sb, gdp); 3127 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3128 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3129 int c = EXT4_B2C(sbi, b - first_block); 3130 ext4_set_bit(c, buf); 3131 count++; 3132 } 3133 if (i != grp) 3134 continue; 3135 s = 0; 3136 if (ext4_bg_has_super(sb, grp)) { 3137 ext4_set_bit(s++, buf); 3138 count++; 3139 } 3140 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { 3141 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3142 count++; 3143 } 3144 } 3145 if (!count) 3146 return 0; 3147 return EXT4_CLUSTERS_PER_GROUP(sb) - 3148 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3149 } 3150 3151 /* 3152 * Compute the overhead and stash it in sbi->s_overhead 3153 */ 3154 int ext4_calculate_overhead(struct super_block *sb) 3155 { 3156 struct ext4_sb_info *sbi = EXT4_SB(sb); 3157 struct ext4_super_block *es = sbi->s_es; 3158 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3159 ext4_fsblk_t overhead = 0; 3160 char *buf = (char *) get_zeroed_page(GFP_KERNEL); 3161 3162 memset(buf, 0, PAGE_SIZE); 3163 if (!buf) 3164 return -ENOMEM; 3165 3166 /* 3167 * Compute the overhead (FS structures). This is constant 3168 * for a given filesystem unless the number of block groups 3169 * changes so we cache the previous value until it does. 3170 */ 3171 3172 /* 3173 * All of the blocks before first_data_block are overhead 3174 */ 3175 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3176 3177 /* 3178 * Add the overhead found in each block group 3179 */ 3180 for (i = 0; i < ngroups; i++) { 3181 int blks; 3182 3183 blks = count_overhead(sb, i, buf); 3184 overhead += blks; 3185 if (blks) 3186 memset(buf, 0, PAGE_SIZE); 3187 cond_resched(); 3188 } 3189 sbi->s_overhead = overhead; 3190 smp_wmb(); 3191 free_page((unsigned long) buf); 3192 return 0; 3193 } 3194 3195 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3196 { 3197 char *orig_data = kstrdup(data, GFP_KERNEL); 3198 struct buffer_head *bh; 3199 struct ext4_super_block *es = NULL; 3200 struct ext4_sb_info *sbi; 3201 ext4_fsblk_t block; 3202 ext4_fsblk_t sb_block = get_sb_block(&data); 3203 ext4_fsblk_t logical_sb_block; 3204 unsigned long offset = 0; 3205 unsigned long journal_devnum = 0; 3206 unsigned long def_mount_opts; 3207 struct inode *root; 3208 char *cp; 3209 const char *descr; 3210 int ret = -ENOMEM; 3211 int blocksize, clustersize; 3212 unsigned int db_count; 3213 unsigned int i; 3214 int needs_recovery, has_huge_files, has_bigalloc; 3215 __u64 blocks_count; 3216 int err; 3217 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3218 ext4_group_t first_not_zeroed; 3219 3220 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3221 if (!sbi) 3222 goto out_free_orig; 3223 3224 sbi->s_blockgroup_lock = 3225 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3226 if (!sbi->s_blockgroup_lock) { 3227 kfree(sbi); 3228 goto out_free_orig; 3229 } 3230 sb->s_fs_info = sbi; 3231 sbi->s_sb = sb; 3232 sbi->s_mount_opt = 0; 3233 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID); 3234 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID); 3235 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3236 sbi->s_sb_block = sb_block; 3237 if (sb->s_bdev->bd_part) 3238 sbi->s_sectors_written_start = 3239 part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3240 3241 /* Cleanup superblock name */ 3242 for (cp = sb->s_id; (cp = strchr(cp, '/'));) 3243 *cp = '!'; 3244 3245 ret = -EINVAL; 3246 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3247 if (!blocksize) { 3248 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3249 goto out_fail; 3250 } 3251 3252 /* 3253 * The ext4 superblock will not be buffer aligned for other than 1kB 3254 * block sizes. We need to calculate the offset from buffer start. 3255 */ 3256 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3257 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3258 offset = do_div(logical_sb_block, blocksize); 3259 } else { 3260 logical_sb_block = sb_block; 3261 } 3262 3263 if (!(bh = sb_bread(sb, logical_sb_block))) { 3264 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3265 goto out_fail; 3266 } 3267 /* 3268 * Note: s_es must be initialized as soon as possible because 3269 * some ext4 macro-instructions depend on its value 3270 */ 3271 es = (struct ext4_super_block *) (bh->b_data + offset); 3272 sbi->s_es = es; 3273 sb->s_magic = le16_to_cpu(es->s_magic); 3274 if (sb->s_magic != EXT4_SUPER_MAGIC) 3275 goto cantfind_ext4; 3276 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3277 3278 /* Warn if metadata_csum and gdt_csum are both set. */ 3279 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3280 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 3281 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) 3282 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are " 3283 "redundant flags; please run fsck."); 3284 3285 /* Check for a known checksum algorithm */ 3286 if (!ext4_verify_csum_type(sb, es)) { 3287 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3288 "unknown checksum algorithm."); 3289 silent = 1; 3290 goto cantfind_ext4; 3291 } 3292 3293 /* Load the checksum driver */ 3294 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3295 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 3296 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3297 if (IS_ERR(sbi->s_chksum_driver)) { 3298 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3299 ret = PTR_ERR(sbi->s_chksum_driver); 3300 sbi->s_chksum_driver = NULL; 3301 goto failed_mount; 3302 } 3303 } 3304 3305 /* Check superblock checksum */ 3306 if (!ext4_superblock_csum_verify(sb, es)) { 3307 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3308 "invalid superblock checksum. Run e2fsck?"); 3309 silent = 1; 3310 goto cantfind_ext4; 3311 } 3312 3313 /* Precompute checksum seed for all metadata */ 3314 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3315 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 3316 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3317 sizeof(es->s_uuid)); 3318 3319 /* Set defaults before we parse the mount options */ 3320 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3321 set_opt(sb, INIT_INODE_TABLE); 3322 if (def_mount_opts & EXT4_DEFM_DEBUG) 3323 set_opt(sb, DEBUG); 3324 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3325 set_opt(sb, GRPID); 3326 if (def_mount_opts & EXT4_DEFM_UID16) 3327 set_opt(sb, NO_UID32); 3328 /* xattr user namespace & acls are now defaulted on */ 3329 #ifdef CONFIG_EXT4_FS_XATTR 3330 set_opt(sb, XATTR_USER); 3331 #endif 3332 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3333 set_opt(sb, POSIX_ACL); 3334 #endif 3335 set_opt(sb, MBLK_IO_SUBMIT); 3336 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3337 set_opt(sb, JOURNAL_DATA); 3338 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3339 set_opt(sb, ORDERED_DATA); 3340 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3341 set_opt(sb, WRITEBACK_DATA); 3342 3343 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3344 set_opt(sb, ERRORS_PANIC); 3345 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3346 set_opt(sb, ERRORS_CONT); 3347 else 3348 set_opt(sb, ERRORS_RO); 3349 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) 3350 set_opt(sb, BLOCK_VALIDITY); 3351 if (def_mount_opts & EXT4_DEFM_DISCARD) 3352 set_opt(sb, DISCARD); 3353 3354 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3355 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3356 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3357 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3358 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3359 3360 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3361 set_opt(sb, BARRIER); 3362 3363 /* 3364 * enable delayed allocation by default 3365 * Use -o nodelalloc to turn it off 3366 */ 3367 if (!IS_EXT3_SB(sb) && 3368 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3369 set_opt(sb, DELALLOC); 3370 3371 /* 3372 * set default s_li_wait_mult for lazyinit, for the case there is 3373 * no mount option specified. 3374 */ 3375 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3376 3377 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3378 &journal_devnum, &journal_ioprio, 0)) { 3379 ext4_msg(sb, KERN_WARNING, 3380 "failed to parse options in superblock: %s", 3381 sbi->s_es->s_mount_opts); 3382 } 3383 sbi->s_def_mount_opt = sbi->s_mount_opt; 3384 if (!parse_options((char *) data, sb, &journal_devnum, 3385 &journal_ioprio, 0)) 3386 goto failed_mount; 3387 3388 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3389 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3390 "with data=journal disables delayed " 3391 "allocation and O_DIRECT support!\n"); 3392 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3393 ext4_msg(sb, KERN_ERR, "can't mount with " 3394 "both data=journal and delalloc"); 3395 goto failed_mount; 3396 } 3397 if (test_opt(sb, DIOREAD_NOLOCK)) { 3398 ext4_msg(sb, KERN_ERR, "can't mount with " 3399 "both data=journal and delalloc"); 3400 goto failed_mount; 3401 } 3402 if (test_opt(sb, DELALLOC)) 3403 clear_opt(sb, DELALLOC); 3404 } 3405 3406 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3407 if (test_opt(sb, DIOREAD_NOLOCK)) { 3408 if (blocksize < PAGE_SIZE) { 3409 ext4_msg(sb, KERN_ERR, "can't mount with " 3410 "dioread_nolock if block size != PAGE_SIZE"); 3411 goto failed_mount; 3412 } 3413 } 3414 3415 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3416 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3417 3418 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3419 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || 3420 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || 3421 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) 3422 ext4_msg(sb, KERN_WARNING, 3423 "feature flags set on rev 0 fs, " 3424 "running e2fsck is recommended"); 3425 3426 if (IS_EXT2_SB(sb)) { 3427 if (ext2_feature_set_ok(sb)) 3428 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3429 "using the ext4 subsystem"); 3430 else { 3431 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3432 "to feature incompatibilities"); 3433 goto failed_mount; 3434 } 3435 } 3436 3437 if (IS_EXT3_SB(sb)) { 3438 if (ext3_feature_set_ok(sb)) 3439 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3440 "using the ext4 subsystem"); 3441 else { 3442 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3443 "to feature incompatibilities"); 3444 goto failed_mount; 3445 } 3446 } 3447 3448 /* 3449 * Check feature flags regardless of the revision level, since we 3450 * previously didn't change the revision level when setting the flags, 3451 * so there is a chance incompat flags are set on a rev 0 filesystem. 3452 */ 3453 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3454 goto failed_mount; 3455 3456 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3457 blocksize > EXT4_MAX_BLOCK_SIZE) { 3458 ext4_msg(sb, KERN_ERR, 3459 "Unsupported filesystem blocksize %d", blocksize); 3460 goto failed_mount; 3461 } 3462 3463 if (sb->s_blocksize != blocksize) { 3464 /* Validate the filesystem blocksize */ 3465 if (!sb_set_blocksize(sb, blocksize)) { 3466 ext4_msg(sb, KERN_ERR, "bad block size %d", 3467 blocksize); 3468 goto failed_mount; 3469 } 3470 3471 brelse(bh); 3472 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3473 offset = do_div(logical_sb_block, blocksize); 3474 bh = sb_bread(sb, logical_sb_block); 3475 if (!bh) { 3476 ext4_msg(sb, KERN_ERR, 3477 "Can't read superblock on 2nd try"); 3478 goto failed_mount; 3479 } 3480 es = (struct ext4_super_block *)(bh->b_data + offset); 3481 sbi->s_es = es; 3482 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3483 ext4_msg(sb, KERN_ERR, 3484 "Magic mismatch, very weird!"); 3485 goto failed_mount; 3486 } 3487 } 3488 3489 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3490 EXT4_FEATURE_RO_COMPAT_HUGE_FILE); 3491 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3492 has_huge_files); 3493 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3494 3495 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3496 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3497 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3498 } else { 3499 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3500 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3501 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3502 (!is_power_of_2(sbi->s_inode_size)) || 3503 (sbi->s_inode_size > blocksize)) { 3504 ext4_msg(sb, KERN_ERR, 3505 "unsupported inode size: %d", 3506 sbi->s_inode_size); 3507 goto failed_mount; 3508 } 3509 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3510 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3511 } 3512 3513 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3514 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { 3515 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 3516 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3517 !is_power_of_2(sbi->s_desc_size)) { 3518 ext4_msg(sb, KERN_ERR, 3519 "unsupported descriptor size %lu", 3520 sbi->s_desc_size); 3521 goto failed_mount; 3522 } 3523 } else 3524 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 3525 3526 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3527 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3528 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3529 goto cantfind_ext4; 3530 3531 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3532 if (sbi->s_inodes_per_block == 0) 3533 goto cantfind_ext4; 3534 sbi->s_itb_per_group = sbi->s_inodes_per_group / 3535 sbi->s_inodes_per_block; 3536 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3537 sbi->s_sbh = bh; 3538 sbi->s_mount_state = le16_to_cpu(es->s_state); 3539 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3540 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 3541 3542 for (i = 0; i < 4; i++) 3543 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3544 sbi->s_def_hash_version = es->s_def_hash_version; 3545 i = le32_to_cpu(es->s_flags); 3546 if (i & EXT2_FLAGS_UNSIGNED_HASH) 3547 sbi->s_hash_unsigned = 3; 3548 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3549 #ifdef __CHAR_UNSIGNED__ 3550 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3551 sbi->s_hash_unsigned = 3; 3552 #else 3553 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3554 #endif 3555 } 3556 3557 /* Handle clustersize */ 3558 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3559 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3560 EXT4_FEATURE_RO_COMPAT_BIGALLOC); 3561 if (has_bigalloc) { 3562 if (clustersize < blocksize) { 3563 ext4_msg(sb, KERN_ERR, 3564 "cluster size (%d) smaller than " 3565 "block size (%d)", clustersize, blocksize); 3566 goto failed_mount; 3567 } 3568 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3569 le32_to_cpu(es->s_log_block_size); 3570 sbi->s_clusters_per_group = 3571 le32_to_cpu(es->s_clusters_per_group); 3572 if (sbi->s_clusters_per_group > blocksize * 8) { 3573 ext4_msg(sb, KERN_ERR, 3574 "#clusters per group too big: %lu", 3575 sbi->s_clusters_per_group); 3576 goto failed_mount; 3577 } 3578 if (sbi->s_blocks_per_group != 3579 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3580 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3581 "clusters per group (%lu) inconsistent", 3582 sbi->s_blocks_per_group, 3583 sbi->s_clusters_per_group); 3584 goto failed_mount; 3585 } 3586 } else { 3587 if (clustersize != blocksize) { 3588 ext4_warning(sb, "fragment/cluster size (%d) != " 3589 "block size (%d)", clustersize, 3590 blocksize); 3591 clustersize = blocksize; 3592 } 3593 if (sbi->s_blocks_per_group > blocksize * 8) { 3594 ext4_msg(sb, KERN_ERR, 3595 "#blocks per group too big: %lu", 3596 sbi->s_blocks_per_group); 3597 goto failed_mount; 3598 } 3599 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3600 sbi->s_cluster_bits = 0; 3601 } 3602 sbi->s_cluster_ratio = clustersize / blocksize; 3603 3604 if (sbi->s_inodes_per_group > blocksize * 8) { 3605 ext4_msg(sb, KERN_ERR, 3606 "#inodes per group too big: %lu", 3607 sbi->s_inodes_per_group); 3608 goto failed_mount; 3609 } 3610 3611 /* 3612 * Test whether we have more sectors than will fit in sector_t, 3613 * and whether the max offset is addressable by the page cache. 3614 */ 3615 err = generic_check_addressable(sb->s_blocksize_bits, 3616 ext4_blocks_count(es)); 3617 if (err) { 3618 ext4_msg(sb, KERN_ERR, "filesystem" 3619 " too large to mount safely on this system"); 3620 if (sizeof(sector_t) < 8) 3621 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3622 ret = err; 3623 goto failed_mount; 3624 } 3625 3626 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3627 goto cantfind_ext4; 3628 3629 /* check blocks count against device size */ 3630 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 3631 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3632 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3633 "exceeds size of device (%llu blocks)", 3634 ext4_blocks_count(es), blocks_count); 3635 goto failed_mount; 3636 } 3637 3638 /* 3639 * It makes no sense for the first data block to be beyond the end 3640 * of the filesystem. 3641 */ 3642 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3643 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 3644 "block %u is beyond end of filesystem (%llu)", 3645 le32_to_cpu(es->s_first_data_block), 3646 ext4_blocks_count(es)); 3647 goto failed_mount; 3648 } 3649 blocks_count = (ext4_blocks_count(es) - 3650 le32_to_cpu(es->s_first_data_block) + 3651 EXT4_BLOCKS_PER_GROUP(sb) - 1); 3652 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 3653 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3654 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 3655 "(block count %llu, first data block %u, " 3656 "blocks per group %lu)", sbi->s_groups_count, 3657 ext4_blocks_count(es), 3658 le32_to_cpu(es->s_first_data_block), 3659 EXT4_BLOCKS_PER_GROUP(sb)); 3660 goto failed_mount; 3661 } 3662 sbi->s_groups_count = blocks_count; 3663 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3664 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3665 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3666 EXT4_DESC_PER_BLOCK(sb); 3667 sbi->s_group_desc = ext4_kvmalloc(db_count * 3668 sizeof(struct buffer_head *), 3669 GFP_KERNEL); 3670 if (sbi->s_group_desc == NULL) { 3671 ext4_msg(sb, KERN_ERR, "not enough memory"); 3672 ret = -ENOMEM; 3673 goto failed_mount; 3674 } 3675 3676 if (ext4_proc_root) 3677 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); 3678 3679 if (sbi->s_proc) 3680 proc_create_data("options", S_IRUGO, sbi->s_proc, 3681 &ext4_seq_options_fops, sb); 3682 3683 bgl_lock_init(sbi->s_blockgroup_lock); 3684 3685 for (i = 0; i < db_count; i++) { 3686 block = descriptor_loc(sb, logical_sb_block, i); 3687 sbi->s_group_desc[i] = sb_bread(sb, block); 3688 if (!sbi->s_group_desc[i]) { 3689 ext4_msg(sb, KERN_ERR, 3690 "can't read group descriptor %d", i); 3691 db_count = i; 3692 goto failed_mount2; 3693 } 3694 } 3695 if (!ext4_check_descriptors(sb, &first_not_zeroed)) { 3696 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3697 goto failed_mount2; 3698 } 3699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 3700 if (!ext4_fill_flex_info(sb)) { 3701 ext4_msg(sb, KERN_ERR, 3702 "unable to initialize " 3703 "flex_bg meta info!"); 3704 goto failed_mount2; 3705 } 3706 3707 sbi->s_gdb_count = db_count; 3708 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3709 spin_lock_init(&sbi->s_next_gen_lock); 3710 3711 init_timer(&sbi->s_err_report); 3712 sbi->s_err_report.function = print_daily_error_info; 3713 sbi->s_err_report.data = (unsigned long) sb; 3714 3715 err = percpu_counter_init(&sbi->s_freeclusters_counter, 3716 ext4_count_free_clusters(sb)); 3717 if (!err) { 3718 err = percpu_counter_init(&sbi->s_freeinodes_counter, 3719 ext4_count_free_inodes(sb)); 3720 } 3721 if (!err) { 3722 err = percpu_counter_init(&sbi->s_dirs_counter, 3723 ext4_count_dirs(sb)); 3724 } 3725 if (!err) { 3726 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0); 3727 } 3728 if (err) { 3729 ext4_msg(sb, KERN_ERR, "insufficient memory"); 3730 ret = err; 3731 goto failed_mount3; 3732 } 3733 3734 sbi->s_stripe = ext4_get_stripe_size(sbi); 3735 sbi->s_max_writeback_mb_bump = 128; 3736 3737 /* 3738 * set up enough so that it can read an inode 3739 */ 3740 if (!test_opt(sb, NOLOAD) && 3741 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 3742 sb->s_op = &ext4_sops; 3743 else 3744 sb->s_op = &ext4_nojournal_sops; 3745 sb->s_export_op = &ext4_export_ops; 3746 sb->s_xattr = ext4_xattr_handlers; 3747 #ifdef CONFIG_QUOTA 3748 sb->s_qcop = &ext4_qctl_operations; 3749 sb->dq_op = &ext4_quota_operations; 3750 3751 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) { 3752 /* Use qctl operations for hidden quota files. */ 3753 sb->s_qcop = &ext4_qctl_sysfile_operations; 3754 } 3755 #endif 3756 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3757 3758 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 3759 mutex_init(&sbi->s_orphan_lock); 3760 sbi->s_resize_flags = 0; 3761 3762 sb->s_root = NULL; 3763 3764 needs_recovery = (es->s_last_orphan != 0 || 3765 EXT4_HAS_INCOMPAT_FEATURE(sb, 3766 EXT4_FEATURE_INCOMPAT_RECOVER)); 3767 3768 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) && 3769 !(sb->s_flags & MS_RDONLY)) 3770 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 3771 goto failed_mount3; 3772 3773 /* 3774 * The first inode we look at is the journal inode. Don't try 3775 * root first: it may be modified in the journal! 3776 */ 3777 if (!test_opt(sb, NOLOAD) && 3778 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 3779 if (ext4_load_journal(sb, es, journal_devnum)) 3780 goto failed_mount3; 3781 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 3782 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 3783 ext4_msg(sb, KERN_ERR, "required journal recovery " 3784 "suppressed and not mounted read-only"); 3785 goto failed_mount_wq; 3786 } else { 3787 clear_opt(sb, DATA_FLAGS); 3788 sbi->s_journal = NULL; 3789 needs_recovery = 0; 3790 goto no_journal; 3791 } 3792 3793 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) && 3794 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3795 JBD2_FEATURE_INCOMPAT_64BIT)) { 3796 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3797 goto failed_mount_wq; 3798 } 3799 3800 if (!set_journal_csum_feature_set(sb)) { 3801 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 3802 "feature set"); 3803 goto failed_mount_wq; 3804 } 3805 3806 /* We have now updated the journal if required, so we can 3807 * validate the data journaling mode. */ 3808 switch (test_opt(sb, DATA_FLAGS)) { 3809 case 0: 3810 /* No mode set, assume a default based on the journal 3811 * capabilities: ORDERED_DATA if the journal can 3812 * cope, else JOURNAL_DATA 3813 */ 3814 if (jbd2_journal_check_available_features 3815 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 3816 set_opt(sb, ORDERED_DATA); 3817 else 3818 set_opt(sb, JOURNAL_DATA); 3819 break; 3820 3821 case EXT4_MOUNT_ORDERED_DATA: 3822 case EXT4_MOUNT_WRITEBACK_DATA: 3823 if (!jbd2_journal_check_available_features 3824 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 3825 ext4_msg(sb, KERN_ERR, "Journal does not support " 3826 "requested data journaling mode"); 3827 goto failed_mount_wq; 3828 } 3829 default: 3830 break; 3831 } 3832 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 3833 3834 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 3835 3836 /* 3837 * The journal may have updated the bg summary counts, so we 3838 * need to update the global counters. 3839 */ 3840 percpu_counter_set(&sbi->s_freeclusters_counter, 3841 ext4_count_free_clusters(sb)); 3842 percpu_counter_set(&sbi->s_freeinodes_counter, 3843 ext4_count_free_inodes(sb)); 3844 percpu_counter_set(&sbi->s_dirs_counter, 3845 ext4_count_dirs(sb)); 3846 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0); 3847 3848 no_journal: 3849 /* 3850 * Get the # of file system overhead blocks from the 3851 * superblock if present. 3852 */ 3853 if (es->s_overhead_clusters) 3854 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 3855 else { 3856 ret = ext4_calculate_overhead(sb); 3857 if (ret) 3858 goto failed_mount_wq; 3859 } 3860 3861 /* 3862 * The maximum number of concurrent works can be high and 3863 * concurrency isn't really necessary. Limit it to 1. 3864 */ 3865 EXT4_SB(sb)->dio_unwritten_wq = 3866 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 3867 if (!EXT4_SB(sb)->dio_unwritten_wq) { 3868 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); 3869 goto failed_mount_wq; 3870 } 3871 3872 /* 3873 * The jbd2_journal_load will have done any necessary log recovery, 3874 * so we can safely mount the rest of the filesystem now. 3875 */ 3876 3877 root = ext4_iget(sb, EXT4_ROOT_INO); 3878 if (IS_ERR(root)) { 3879 ext4_msg(sb, KERN_ERR, "get root inode failed"); 3880 ret = PTR_ERR(root); 3881 root = NULL; 3882 goto failed_mount4; 3883 } 3884 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 3885 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 3886 iput(root); 3887 goto failed_mount4; 3888 } 3889 sb->s_root = d_make_root(root); 3890 if (!sb->s_root) { 3891 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 3892 ret = -ENOMEM; 3893 goto failed_mount4; 3894 } 3895 3896 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY)) 3897 sb->s_flags |= MS_RDONLY; 3898 3899 /* determine the minimum size of new large inodes, if present */ 3900 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 3901 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3902 EXT4_GOOD_OLD_INODE_SIZE; 3903 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3904 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { 3905 if (sbi->s_want_extra_isize < 3906 le16_to_cpu(es->s_want_extra_isize)) 3907 sbi->s_want_extra_isize = 3908 le16_to_cpu(es->s_want_extra_isize); 3909 if (sbi->s_want_extra_isize < 3910 le16_to_cpu(es->s_min_extra_isize)) 3911 sbi->s_want_extra_isize = 3912 le16_to_cpu(es->s_min_extra_isize); 3913 } 3914 } 3915 /* Check if enough inode space is available */ 3916 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 3917 sbi->s_inode_size) { 3918 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3919 EXT4_GOOD_OLD_INODE_SIZE; 3920 ext4_msg(sb, KERN_INFO, "required extra inode space not" 3921 "available"); 3922 } 3923 3924 err = ext4_setup_system_zone(sb); 3925 if (err) { 3926 ext4_msg(sb, KERN_ERR, "failed to initialize system " 3927 "zone (%d)", err); 3928 goto failed_mount4a; 3929 } 3930 3931 ext4_ext_init(sb); 3932 err = ext4_mb_init(sb); 3933 if (err) { 3934 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 3935 err); 3936 goto failed_mount5; 3937 } 3938 3939 err = ext4_register_li_request(sb, first_not_zeroed); 3940 if (err) 3941 goto failed_mount6; 3942 3943 sbi->s_kobj.kset = ext4_kset; 3944 init_completion(&sbi->s_kobj_unregister); 3945 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 3946 "%s", sb->s_id); 3947 if (err) 3948 goto failed_mount7; 3949 3950 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 3951 ext4_orphan_cleanup(sb, es); 3952 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 3953 if (needs_recovery) { 3954 ext4_msg(sb, KERN_INFO, "recovery complete"); 3955 ext4_mark_recovery_complete(sb, es); 3956 } 3957 if (EXT4_SB(sb)->s_journal) { 3958 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 3959 descr = " journalled data mode"; 3960 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 3961 descr = " ordered data mode"; 3962 else 3963 descr = " writeback data mode"; 3964 } else 3965 descr = "out journal"; 3966 3967 #ifdef CONFIG_QUOTA 3968 /* Enable quota usage during mount. */ 3969 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 3970 !(sb->s_flags & MS_RDONLY)) { 3971 ret = ext4_enable_quotas(sb); 3972 if (ret) 3973 goto failed_mount7; 3974 } 3975 #endif /* CONFIG_QUOTA */ 3976 3977 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 3978 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 3979 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 3980 3981 if (es->s_error_count) 3982 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 3983 3984 kfree(orig_data); 3985 return 0; 3986 3987 cantfind_ext4: 3988 if (!silent) 3989 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 3990 goto failed_mount; 3991 3992 failed_mount7: 3993 ext4_unregister_li_request(sb); 3994 failed_mount6: 3995 ext4_mb_release(sb); 3996 failed_mount5: 3997 ext4_ext_release(sb); 3998 ext4_release_system_zone(sb); 3999 failed_mount4a: 4000 dput(sb->s_root); 4001 sb->s_root = NULL; 4002 failed_mount4: 4003 ext4_msg(sb, KERN_ERR, "mount failed"); 4004 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); 4005 failed_mount_wq: 4006 if (sbi->s_journal) { 4007 jbd2_journal_destroy(sbi->s_journal); 4008 sbi->s_journal = NULL; 4009 } 4010 failed_mount3: 4011 del_timer(&sbi->s_err_report); 4012 if (sbi->s_flex_groups) 4013 ext4_kvfree(sbi->s_flex_groups); 4014 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4015 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4016 percpu_counter_destroy(&sbi->s_dirs_counter); 4017 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4018 if (sbi->s_mmp_tsk) 4019 kthread_stop(sbi->s_mmp_tsk); 4020 failed_mount2: 4021 for (i = 0; i < db_count; i++) 4022 brelse(sbi->s_group_desc[i]); 4023 ext4_kvfree(sbi->s_group_desc); 4024 failed_mount: 4025 if (sbi->s_chksum_driver) 4026 crypto_free_shash(sbi->s_chksum_driver); 4027 if (sbi->s_proc) { 4028 remove_proc_entry("options", sbi->s_proc); 4029 remove_proc_entry(sb->s_id, ext4_proc_root); 4030 } 4031 #ifdef CONFIG_QUOTA 4032 for (i = 0; i < MAXQUOTAS; i++) 4033 kfree(sbi->s_qf_names[i]); 4034 #endif 4035 ext4_blkdev_remove(sbi); 4036 brelse(bh); 4037 out_fail: 4038 sb->s_fs_info = NULL; 4039 kfree(sbi->s_blockgroup_lock); 4040 kfree(sbi); 4041 out_free_orig: 4042 kfree(orig_data); 4043 return ret; 4044 } 4045 4046 /* 4047 * Setup any per-fs journal parameters now. We'll do this both on 4048 * initial mount, once the journal has been initialised but before we've 4049 * done any recovery; and again on any subsequent remount. 4050 */ 4051 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4052 { 4053 struct ext4_sb_info *sbi = EXT4_SB(sb); 4054 4055 journal->j_commit_interval = sbi->s_commit_interval; 4056 journal->j_min_batch_time = sbi->s_min_batch_time; 4057 journal->j_max_batch_time = sbi->s_max_batch_time; 4058 4059 write_lock(&journal->j_state_lock); 4060 if (test_opt(sb, BARRIER)) 4061 journal->j_flags |= JBD2_BARRIER; 4062 else 4063 journal->j_flags &= ~JBD2_BARRIER; 4064 if (test_opt(sb, DATA_ERR_ABORT)) 4065 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4066 else 4067 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4068 write_unlock(&journal->j_state_lock); 4069 } 4070 4071 static journal_t *ext4_get_journal(struct super_block *sb, 4072 unsigned int journal_inum) 4073 { 4074 struct inode *journal_inode; 4075 journal_t *journal; 4076 4077 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4078 4079 /* First, test for the existence of a valid inode on disk. Bad 4080 * things happen if we iget() an unused inode, as the subsequent 4081 * iput() will try to delete it. */ 4082 4083 journal_inode = ext4_iget(sb, journal_inum); 4084 if (IS_ERR(journal_inode)) { 4085 ext4_msg(sb, KERN_ERR, "no journal found"); 4086 return NULL; 4087 } 4088 if (!journal_inode->i_nlink) { 4089 make_bad_inode(journal_inode); 4090 iput(journal_inode); 4091 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4092 return NULL; 4093 } 4094 4095 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4096 journal_inode, journal_inode->i_size); 4097 if (!S_ISREG(journal_inode->i_mode)) { 4098 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4099 iput(journal_inode); 4100 return NULL; 4101 } 4102 4103 journal = jbd2_journal_init_inode(journal_inode); 4104 if (!journal) { 4105 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4106 iput(journal_inode); 4107 return NULL; 4108 } 4109 journal->j_private = sb; 4110 ext4_init_journal_params(sb, journal); 4111 return journal; 4112 } 4113 4114 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4115 dev_t j_dev) 4116 { 4117 struct buffer_head *bh; 4118 journal_t *journal; 4119 ext4_fsblk_t start; 4120 ext4_fsblk_t len; 4121 int hblock, blocksize; 4122 ext4_fsblk_t sb_block; 4123 unsigned long offset; 4124 struct ext4_super_block *es; 4125 struct block_device *bdev; 4126 4127 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4128 4129 bdev = ext4_blkdev_get(j_dev, sb); 4130 if (bdev == NULL) 4131 return NULL; 4132 4133 blocksize = sb->s_blocksize; 4134 hblock = bdev_logical_block_size(bdev); 4135 if (blocksize < hblock) { 4136 ext4_msg(sb, KERN_ERR, 4137 "blocksize too small for journal device"); 4138 goto out_bdev; 4139 } 4140 4141 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4142 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4143 set_blocksize(bdev, blocksize); 4144 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4145 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4146 "external journal"); 4147 goto out_bdev; 4148 } 4149 4150 es = (struct ext4_super_block *) (bh->b_data + offset); 4151 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4152 !(le32_to_cpu(es->s_feature_incompat) & 4153 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4154 ext4_msg(sb, KERN_ERR, "external journal has " 4155 "bad superblock"); 4156 brelse(bh); 4157 goto out_bdev; 4158 } 4159 4160 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4161 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4162 brelse(bh); 4163 goto out_bdev; 4164 } 4165 4166 len = ext4_blocks_count(es); 4167 start = sb_block + 1; 4168 brelse(bh); /* we're done with the superblock */ 4169 4170 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4171 start, len, blocksize); 4172 if (!journal) { 4173 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4174 goto out_bdev; 4175 } 4176 journal->j_private = sb; 4177 ll_rw_block(READ, 1, &journal->j_sb_buffer); 4178 wait_on_buffer(journal->j_sb_buffer); 4179 if (!buffer_uptodate(journal->j_sb_buffer)) { 4180 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4181 goto out_journal; 4182 } 4183 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4184 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4185 "user (unsupported) - %d", 4186 be32_to_cpu(journal->j_superblock->s_nr_users)); 4187 goto out_journal; 4188 } 4189 EXT4_SB(sb)->journal_bdev = bdev; 4190 ext4_init_journal_params(sb, journal); 4191 return journal; 4192 4193 out_journal: 4194 jbd2_journal_destroy(journal); 4195 out_bdev: 4196 ext4_blkdev_put(bdev); 4197 return NULL; 4198 } 4199 4200 static int ext4_load_journal(struct super_block *sb, 4201 struct ext4_super_block *es, 4202 unsigned long journal_devnum) 4203 { 4204 journal_t *journal; 4205 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4206 dev_t journal_dev; 4207 int err = 0; 4208 int really_read_only; 4209 4210 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4211 4212 if (journal_devnum && 4213 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4214 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4215 "numbers have changed"); 4216 journal_dev = new_decode_dev(journal_devnum); 4217 } else 4218 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4219 4220 really_read_only = bdev_read_only(sb->s_bdev); 4221 4222 /* 4223 * Are we loading a blank journal or performing recovery after a 4224 * crash? For recovery, we need to check in advance whether we 4225 * can get read-write access to the device. 4226 */ 4227 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 4228 if (sb->s_flags & MS_RDONLY) { 4229 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4230 "required on readonly filesystem"); 4231 if (really_read_only) { 4232 ext4_msg(sb, KERN_ERR, "write access " 4233 "unavailable, cannot proceed"); 4234 return -EROFS; 4235 } 4236 ext4_msg(sb, KERN_INFO, "write access will " 4237 "be enabled during recovery"); 4238 } 4239 } 4240 4241 if (journal_inum && journal_dev) { 4242 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4243 "and inode journals!"); 4244 return -EINVAL; 4245 } 4246 4247 if (journal_inum) { 4248 if (!(journal = ext4_get_journal(sb, journal_inum))) 4249 return -EINVAL; 4250 } else { 4251 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4252 return -EINVAL; 4253 } 4254 4255 if (!(journal->j_flags & JBD2_BARRIER)) 4256 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4257 4258 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) 4259 err = jbd2_journal_wipe(journal, !really_read_only); 4260 if (!err) { 4261 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4262 if (save) 4263 memcpy(save, ((char *) es) + 4264 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4265 err = jbd2_journal_load(journal); 4266 if (save) 4267 memcpy(((char *) es) + EXT4_S_ERR_START, 4268 save, EXT4_S_ERR_LEN); 4269 kfree(save); 4270 } 4271 4272 if (err) { 4273 ext4_msg(sb, KERN_ERR, "error loading journal"); 4274 jbd2_journal_destroy(journal); 4275 return err; 4276 } 4277 4278 EXT4_SB(sb)->s_journal = journal; 4279 ext4_clear_journal_err(sb, es); 4280 4281 if (!really_read_only && journal_devnum && 4282 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4283 es->s_journal_dev = cpu_to_le32(journal_devnum); 4284 4285 /* Make sure we flush the recovery flag to disk. */ 4286 ext4_commit_super(sb, 1); 4287 } 4288 4289 return 0; 4290 } 4291 4292 static int ext4_commit_super(struct super_block *sb, int sync) 4293 { 4294 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4295 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4296 int error = 0; 4297 4298 if (!sbh || block_device_ejected(sb)) 4299 return error; 4300 if (buffer_write_io_error(sbh)) { 4301 /* 4302 * Oh, dear. A previous attempt to write the 4303 * superblock failed. This could happen because the 4304 * USB device was yanked out. Or it could happen to 4305 * be a transient write error and maybe the block will 4306 * be remapped. Nothing we can do but to retry the 4307 * write and hope for the best. 4308 */ 4309 ext4_msg(sb, KERN_ERR, "previous I/O error to " 4310 "superblock detected"); 4311 clear_buffer_write_io_error(sbh); 4312 set_buffer_uptodate(sbh); 4313 } 4314 /* 4315 * If the file system is mounted read-only, don't update the 4316 * superblock write time. This avoids updating the superblock 4317 * write time when we are mounting the root file system 4318 * read/only but we need to replay the journal; at that point, 4319 * for people who are east of GMT and who make their clock 4320 * tick in localtime for Windows bug-for-bug compatibility, 4321 * the clock is set in the future, and this will cause e2fsck 4322 * to complain and force a full file system check. 4323 */ 4324 if (!(sb->s_flags & MS_RDONLY)) 4325 es->s_wtime = cpu_to_le32(get_seconds()); 4326 if (sb->s_bdev->bd_part) 4327 es->s_kbytes_written = 4328 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4329 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4330 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4331 else 4332 es->s_kbytes_written = 4333 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4334 ext4_free_blocks_count_set(es, 4335 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4336 &EXT4_SB(sb)->s_freeclusters_counter))); 4337 es->s_free_inodes_count = 4338 cpu_to_le32(percpu_counter_sum_positive( 4339 &EXT4_SB(sb)->s_freeinodes_counter)); 4340 BUFFER_TRACE(sbh, "marking dirty"); 4341 ext4_superblock_csum_set(sb, es); 4342 mark_buffer_dirty(sbh); 4343 if (sync) { 4344 error = sync_dirty_buffer(sbh); 4345 if (error) 4346 return error; 4347 4348 error = buffer_write_io_error(sbh); 4349 if (error) { 4350 ext4_msg(sb, KERN_ERR, "I/O error while writing " 4351 "superblock"); 4352 clear_buffer_write_io_error(sbh); 4353 set_buffer_uptodate(sbh); 4354 } 4355 } 4356 return error; 4357 } 4358 4359 /* 4360 * Have we just finished recovery? If so, and if we are mounting (or 4361 * remounting) the filesystem readonly, then we will end up with a 4362 * consistent fs on disk. Record that fact. 4363 */ 4364 static void ext4_mark_recovery_complete(struct super_block *sb, 4365 struct ext4_super_block *es) 4366 { 4367 journal_t *journal = EXT4_SB(sb)->s_journal; 4368 4369 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 4370 BUG_ON(journal != NULL); 4371 return; 4372 } 4373 jbd2_journal_lock_updates(journal); 4374 if (jbd2_journal_flush(journal) < 0) 4375 goto out; 4376 4377 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && 4378 sb->s_flags & MS_RDONLY) { 4379 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4380 ext4_commit_super(sb, 1); 4381 } 4382 4383 out: 4384 jbd2_journal_unlock_updates(journal); 4385 } 4386 4387 /* 4388 * If we are mounting (or read-write remounting) a filesystem whose journal 4389 * has recorded an error from a previous lifetime, move that error to the 4390 * main filesystem now. 4391 */ 4392 static void ext4_clear_journal_err(struct super_block *sb, 4393 struct ext4_super_block *es) 4394 { 4395 journal_t *journal; 4396 int j_errno; 4397 const char *errstr; 4398 4399 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4400 4401 journal = EXT4_SB(sb)->s_journal; 4402 4403 /* 4404 * Now check for any error status which may have been recorded in the 4405 * journal by a prior ext4_error() or ext4_abort() 4406 */ 4407 4408 j_errno = jbd2_journal_errno(journal); 4409 if (j_errno) { 4410 char nbuf[16]; 4411 4412 errstr = ext4_decode_error(sb, j_errno, nbuf); 4413 ext4_warning(sb, "Filesystem error recorded " 4414 "from previous mount: %s", errstr); 4415 ext4_warning(sb, "Marking fs in need of filesystem check."); 4416 4417 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4418 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4419 ext4_commit_super(sb, 1); 4420 4421 jbd2_journal_clear_err(journal); 4422 } 4423 } 4424 4425 /* 4426 * Force the running and committing transactions to commit, 4427 * and wait on the commit. 4428 */ 4429 int ext4_force_commit(struct super_block *sb) 4430 { 4431 journal_t *journal; 4432 int ret = 0; 4433 4434 if (sb->s_flags & MS_RDONLY) 4435 return 0; 4436 4437 journal = EXT4_SB(sb)->s_journal; 4438 if (journal) 4439 ret = ext4_journal_force_commit(journal); 4440 4441 return ret; 4442 } 4443 4444 static int ext4_sync_fs(struct super_block *sb, int wait) 4445 { 4446 int ret = 0; 4447 tid_t target; 4448 struct ext4_sb_info *sbi = EXT4_SB(sb); 4449 4450 trace_ext4_sync_fs(sb, wait); 4451 flush_workqueue(sbi->dio_unwritten_wq); 4452 /* 4453 * Writeback quota in non-journalled quota case - journalled quota has 4454 * no dirty dquots 4455 */ 4456 dquot_writeback_dquots(sb, -1); 4457 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4458 if (wait) 4459 jbd2_log_wait_commit(sbi->s_journal, target); 4460 } 4461 return ret; 4462 } 4463 4464 /* 4465 * LVM calls this function before a (read-only) snapshot is created. This 4466 * gives us a chance to flush the journal completely and mark the fs clean. 4467 * 4468 * Note that only this function cannot bring a filesystem to be in a clean 4469 * state independently. It relies on upper layer to stop all data & metadata 4470 * modifications. 4471 */ 4472 static int ext4_freeze(struct super_block *sb) 4473 { 4474 int error = 0; 4475 journal_t *journal; 4476 4477 if (sb->s_flags & MS_RDONLY) 4478 return 0; 4479 4480 journal = EXT4_SB(sb)->s_journal; 4481 4482 /* Now we set up the journal barrier. */ 4483 jbd2_journal_lock_updates(journal); 4484 4485 /* 4486 * Don't clear the needs_recovery flag if we failed to flush 4487 * the journal. 4488 */ 4489 error = jbd2_journal_flush(journal); 4490 if (error < 0) 4491 goto out; 4492 4493 /* Journal blocked and flushed, clear needs_recovery flag. */ 4494 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4495 error = ext4_commit_super(sb, 1); 4496 out: 4497 /* we rely on upper layer to stop further updates */ 4498 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 4499 return error; 4500 } 4501 4502 /* 4503 * Called by LVM after the snapshot is done. We need to reset the RECOVER 4504 * flag here, even though the filesystem is not technically dirty yet. 4505 */ 4506 static int ext4_unfreeze(struct super_block *sb) 4507 { 4508 if (sb->s_flags & MS_RDONLY) 4509 return 0; 4510 4511 lock_super(sb); 4512 /* Reset the needs_recovery flag before the fs is unlocked. */ 4513 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4514 ext4_commit_super(sb, 1); 4515 unlock_super(sb); 4516 return 0; 4517 } 4518 4519 /* 4520 * Structure to save mount options for ext4_remount's benefit 4521 */ 4522 struct ext4_mount_options { 4523 unsigned long s_mount_opt; 4524 unsigned long s_mount_opt2; 4525 kuid_t s_resuid; 4526 kgid_t s_resgid; 4527 unsigned long s_commit_interval; 4528 u32 s_min_batch_time, s_max_batch_time; 4529 #ifdef CONFIG_QUOTA 4530 int s_jquota_fmt; 4531 char *s_qf_names[MAXQUOTAS]; 4532 #endif 4533 }; 4534 4535 static int ext4_remount(struct super_block *sb, int *flags, char *data) 4536 { 4537 struct ext4_super_block *es; 4538 struct ext4_sb_info *sbi = EXT4_SB(sb); 4539 unsigned long old_sb_flags; 4540 struct ext4_mount_options old_opts; 4541 int enable_quota = 0; 4542 ext4_group_t g; 4543 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4544 int err = 0; 4545 #ifdef CONFIG_QUOTA 4546 int i; 4547 #endif 4548 char *orig_data = kstrdup(data, GFP_KERNEL); 4549 4550 /* Store the original options */ 4551 lock_super(sb); 4552 old_sb_flags = sb->s_flags; 4553 old_opts.s_mount_opt = sbi->s_mount_opt; 4554 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4555 old_opts.s_resuid = sbi->s_resuid; 4556 old_opts.s_resgid = sbi->s_resgid; 4557 old_opts.s_commit_interval = sbi->s_commit_interval; 4558 old_opts.s_min_batch_time = sbi->s_min_batch_time; 4559 old_opts.s_max_batch_time = sbi->s_max_batch_time; 4560 #ifdef CONFIG_QUOTA 4561 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4562 for (i = 0; i < MAXQUOTAS; i++) 4563 old_opts.s_qf_names[i] = sbi->s_qf_names[i]; 4564 #endif 4565 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4566 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4567 4568 /* 4569 * Allow the "check" option to be passed as a remount option. 4570 */ 4571 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 4572 err = -EINVAL; 4573 goto restore_opts; 4574 } 4575 4576 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4577 ext4_abort(sb, "Abort forced by user"); 4578 4579 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4580 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4581 4582 es = sbi->s_es; 4583 4584 if (sbi->s_journal) { 4585 ext4_init_journal_params(sb, sbi->s_journal); 4586 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4587 } 4588 4589 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 4590 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4591 err = -EROFS; 4592 goto restore_opts; 4593 } 4594 4595 if (*flags & MS_RDONLY) { 4596 err = dquot_suspend(sb, -1); 4597 if (err < 0) 4598 goto restore_opts; 4599 4600 /* 4601 * First of all, the unconditional stuff we have to do 4602 * to disable replay of the journal when we next remount 4603 */ 4604 sb->s_flags |= MS_RDONLY; 4605 4606 /* 4607 * OK, test if we are remounting a valid rw partition 4608 * readonly, and if so set the rdonly flag and then 4609 * mark the partition as valid again. 4610 */ 4611 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4612 (sbi->s_mount_state & EXT4_VALID_FS)) 4613 es->s_state = cpu_to_le16(sbi->s_mount_state); 4614 4615 if (sbi->s_journal) 4616 ext4_mark_recovery_complete(sb, es); 4617 } else { 4618 /* Make sure we can mount this feature set readwrite */ 4619 if (!ext4_feature_set_ok(sb, 0)) { 4620 err = -EROFS; 4621 goto restore_opts; 4622 } 4623 /* 4624 * Make sure the group descriptor checksums 4625 * are sane. If they aren't, refuse to remount r/w. 4626 */ 4627 for (g = 0; g < sbi->s_groups_count; g++) { 4628 struct ext4_group_desc *gdp = 4629 ext4_get_group_desc(sb, g, NULL); 4630 4631 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 4632 ext4_msg(sb, KERN_ERR, 4633 "ext4_remount: Checksum for group %u failed (%u!=%u)", 4634 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), 4635 le16_to_cpu(gdp->bg_checksum)); 4636 err = -EINVAL; 4637 goto restore_opts; 4638 } 4639 } 4640 4641 /* 4642 * If we have an unprocessed orphan list hanging 4643 * around from a previously readonly bdev mount, 4644 * require a full umount/remount for now. 4645 */ 4646 if (es->s_last_orphan) { 4647 ext4_msg(sb, KERN_WARNING, "Couldn't " 4648 "remount RDWR because of unprocessed " 4649 "orphan inode list. Please " 4650 "umount/remount instead"); 4651 err = -EINVAL; 4652 goto restore_opts; 4653 } 4654 4655 /* 4656 * Mounting a RDONLY partition read-write, so reread 4657 * and store the current valid flag. (It may have 4658 * been changed by e2fsck since we originally mounted 4659 * the partition.) 4660 */ 4661 if (sbi->s_journal) 4662 ext4_clear_journal_err(sb, es); 4663 sbi->s_mount_state = le16_to_cpu(es->s_state); 4664 if (!ext4_setup_super(sb, es, 0)) 4665 sb->s_flags &= ~MS_RDONLY; 4666 if (EXT4_HAS_INCOMPAT_FEATURE(sb, 4667 EXT4_FEATURE_INCOMPAT_MMP)) 4668 if (ext4_multi_mount_protect(sb, 4669 le64_to_cpu(es->s_mmp_block))) { 4670 err = -EROFS; 4671 goto restore_opts; 4672 } 4673 enable_quota = 1; 4674 } 4675 } 4676 4677 /* 4678 * Reinitialize lazy itable initialization thread based on 4679 * current settings 4680 */ 4681 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4682 ext4_unregister_li_request(sb); 4683 else { 4684 ext4_group_t first_not_zeroed; 4685 first_not_zeroed = ext4_has_uninit_itable(sb); 4686 ext4_register_li_request(sb, first_not_zeroed); 4687 } 4688 4689 ext4_setup_system_zone(sb); 4690 if (sbi->s_journal == NULL) 4691 ext4_commit_super(sb, 1); 4692 4693 unlock_super(sb); 4694 #ifdef CONFIG_QUOTA 4695 /* Release old quota file names */ 4696 for (i = 0; i < MAXQUOTAS; i++) 4697 if (old_opts.s_qf_names[i] && 4698 old_opts.s_qf_names[i] != sbi->s_qf_names[i]) 4699 kfree(old_opts.s_qf_names[i]); 4700 if (enable_quota) { 4701 if (sb_any_quota_suspended(sb)) 4702 dquot_resume(sb, -1); 4703 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 4704 EXT4_FEATURE_RO_COMPAT_QUOTA)) { 4705 err = ext4_enable_quotas(sb); 4706 if (err) { 4707 lock_super(sb); 4708 goto restore_opts; 4709 } 4710 } 4711 } 4712 #endif 4713 4714 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4715 kfree(orig_data); 4716 return 0; 4717 4718 restore_opts: 4719 sb->s_flags = old_sb_flags; 4720 sbi->s_mount_opt = old_opts.s_mount_opt; 4721 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 4722 sbi->s_resuid = old_opts.s_resuid; 4723 sbi->s_resgid = old_opts.s_resgid; 4724 sbi->s_commit_interval = old_opts.s_commit_interval; 4725 sbi->s_min_batch_time = old_opts.s_min_batch_time; 4726 sbi->s_max_batch_time = old_opts.s_max_batch_time; 4727 #ifdef CONFIG_QUOTA 4728 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 4729 for (i = 0; i < MAXQUOTAS; i++) { 4730 if (sbi->s_qf_names[i] && 4731 old_opts.s_qf_names[i] != sbi->s_qf_names[i]) 4732 kfree(sbi->s_qf_names[i]); 4733 sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 4734 } 4735 #endif 4736 unlock_super(sb); 4737 kfree(orig_data); 4738 return err; 4739 } 4740 4741 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 4742 { 4743 struct super_block *sb = dentry->d_sb; 4744 struct ext4_sb_info *sbi = EXT4_SB(sb); 4745 struct ext4_super_block *es = sbi->s_es; 4746 ext4_fsblk_t overhead = 0; 4747 u64 fsid; 4748 s64 bfree; 4749 4750 if (!test_opt(sb, MINIX_DF)) 4751 overhead = sbi->s_overhead; 4752 4753 buf->f_type = EXT4_SUPER_MAGIC; 4754 buf->f_bsize = sb->s_blocksize; 4755 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead); 4756 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 4757 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 4758 /* prevent underflow in case that few free space is available */ 4759 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 4760 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); 4761 if (buf->f_bfree < ext4_r_blocks_count(es)) 4762 buf->f_bavail = 0; 4763 buf->f_files = le32_to_cpu(es->s_inodes_count); 4764 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 4765 buf->f_namelen = EXT4_NAME_LEN; 4766 fsid = le64_to_cpup((void *)es->s_uuid) ^ 4767 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 4768 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 4769 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 4770 4771 return 0; 4772 } 4773 4774 /* Helper function for writing quotas on sync - we need to start transaction 4775 * before quota file is locked for write. Otherwise the are possible deadlocks: 4776 * Process 1 Process 2 4777 * ext4_create() quota_sync() 4778 * jbd2_journal_start() write_dquot() 4779 * dquot_initialize() down(dqio_mutex) 4780 * down(dqio_mutex) jbd2_journal_start() 4781 * 4782 */ 4783 4784 #ifdef CONFIG_QUOTA 4785 4786 static inline struct inode *dquot_to_inode(struct dquot *dquot) 4787 { 4788 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; 4789 } 4790 4791 static int ext4_write_dquot(struct dquot *dquot) 4792 { 4793 int ret, err; 4794 handle_t *handle; 4795 struct inode *inode; 4796 4797 inode = dquot_to_inode(dquot); 4798 handle = ext4_journal_start(inode, 4799 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 4800 if (IS_ERR(handle)) 4801 return PTR_ERR(handle); 4802 ret = dquot_commit(dquot); 4803 err = ext4_journal_stop(handle); 4804 if (!ret) 4805 ret = err; 4806 return ret; 4807 } 4808 4809 static int ext4_acquire_dquot(struct dquot *dquot) 4810 { 4811 int ret, err; 4812 handle_t *handle; 4813 4814 handle = ext4_journal_start(dquot_to_inode(dquot), 4815 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 4816 if (IS_ERR(handle)) 4817 return PTR_ERR(handle); 4818 ret = dquot_acquire(dquot); 4819 err = ext4_journal_stop(handle); 4820 if (!ret) 4821 ret = err; 4822 return ret; 4823 } 4824 4825 static int ext4_release_dquot(struct dquot *dquot) 4826 { 4827 int ret, err; 4828 handle_t *handle; 4829 4830 handle = ext4_journal_start(dquot_to_inode(dquot), 4831 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 4832 if (IS_ERR(handle)) { 4833 /* Release dquot anyway to avoid endless cycle in dqput() */ 4834 dquot_release(dquot); 4835 return PTR_ERR(handle); 4836 } 4837 ret = dquot_release(dquot); 4838 err = ext4_journal_stop(handle); 4839 if (!ret) 4840 ret = err; 4841 return ret; 4842 } 4843 4844 static int ext4_mark_dquot_dirty(struct dquot *dquot) 4845 { 4846 /* Are we journaling quotas? */ 4847 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || 4848 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { 4849 dquot_mark_dquot_dirty(dquot); 4850 return ext4_write_dquot(dquot); 4851 } else { 4852 return dquot_mark_dquot_dirty(dquot); 4853 } 4854 } 4855 4856 static int ext4_write_info(struct super_block *sb, int type) 4857 { 4858 int ret, err; 4859 handle_t *handle; 4860 4861 /* Data block + inode block */ 4862 handle = ext4_journal_start(sb->s_root->d_inode, 2); 4863 if (IS_ERR(handle)) 4864 return PTR_ERR(handle); 4865 ret = dquot_commit_info(sb, type); 4866 err = ext4_journal_stop(handle); 4867 if (!ret) 4868 ret = err; 4869 return ret; 4870 } 4871 4872 /* 4873 * Turn on quotas during mount time - we need to find 4874 * the quota file and such... 4875 */ 4876 static int ext4_quota_on_mount(struct super_block *sb, int type) 4877 { 4878 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 4879 EXT4_SB(sb)->s_jquota_fmt, type); 4880 } 4881 4882 /* 4883 * Standard function to be called on quota_on 4884 */ 4885 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 4886 struct path *path) 4887 { 4888 int err; 4889 4890 if (!test_opt(sb, QUOTA)) 4891 return -EINVAL; 4892 4893 /* Quotafile not on the same filesystem? */ 4894 if (path->dentry->d_sb != sb) 4895 return -EXDEV; 4896 /* Journaling quota? */ 4897 if (EXT4_SB(sb)->s_qf_names[type]) { 4898 /* Quotafile not in fs root? */ 4899 if (path->dentry->d_parent != sb->s_root) 4900 ext4_msg(sb, KERN_WARNING, 4901 "Quota file not on filesystem root. " 4902 "Journaled quota will not work"); 4903 } 4904 4905 /* 4906 * When we journal data on quota file, we have to flush journal to see 4907 * all updates to the file when we bypass pagecache... 4908 */ 4909 if (EXT4_SB(sb)->s_journal && 4910 ext4_should_journal_data(path->dentry->d_inode)) { 4911 /* 4912 * We don't need to lock updates but journal_flush() could 4913 * otherwise be livelocked... 4914 */ 4915 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 4916 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 4917 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 4918 if (err) 4919 return err; 4920 } 4921 4922 return dquot_quota_on(sb, type, format_id, path); 4923 } 4924 4925 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 4926 unsigned int flags) 4927 { 4928 int err; 4929 struct inode *qf_inode; 4930 unsigned long qf_inums[MAXQUOTAS] = { 4931 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 4932 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 4933 }; 4934 4935 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)); 4936 4937 if (!qf_inums[type]) 4938 return -EPERM; 4939 4940 qf_inode = ext4_iget(sb, qf_inums[type]); 4941 if (IS_ERR(qf_inode)) { 4942 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 4943 return PTR_ERR(qf_inode); 4944 } 4945 4946 err = dquot_enable(qf_inode, type, format_id, flags); 4947 iput(qf_inode); 4948 4949 return err; 4950 } 4951 4952 /* Enable usage tracking for all quota types. */ 4953 static int ext4_enable_quotas(struct super_block *sb) 4954 { 4955 int type, err = 0; 4956 unsigned long qf_inums[MAXQUOTAS] = { 4957 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 4958 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 4959 }; 4960 4961 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 4962 for (type = 0; type < MAXQUOTAS; type++) { 4963 if (qf_inums[type]) { 4964 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 4965 DQUOT_USAGE_ENABLED); 4966 if (err) { 4967 ext4_warning(sb, 4968 "Failed to enable quota (type=%d) " 4969 "tracking. Please run e2fsck to fix.", 4970 type); 4971 return err; 4972 } 4973 } 4974 } 4975 return 0; 4976 } 4977 4978 /* 4979 * quota_on function that is used when QUOTA feature is set. 4980 */ 4981 static int ext4_quota_on_sysfile(struct super_block *sb, int type, 4982 int format_id) 4983 { 4984 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 4985 return -EINVAL; 4986 4987 /* 4988 * USAGE was enabled at mount time. Only need to enable LIMITS now. 4989 */ 4990 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED); 4991 } 4992 4993 static int ext4_quota_off(struct super_block *sb, int type) 4994 { 4995 struct inode *inode = sb_dqopt(sb)->files[type]; 4996 handle_t *handle; 4997 4998 /* Force all delayed allocation blocks to be allocated. 4999 * Caller already holds s_umount sem */ 5000 if (test_opt(sb, DELALLOC)) 5001 sync_filesystem(sb); 5002 5003 if (!inode) 5004 goto out; 5005 5006 /* Update modification times of quota files when userspace can 5007 * start looking at them */ 5008 handle = ext4_journal_start(inode, 1); 5009 if (IS_ERR(handle)) 5010 goto out; 5011 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 5012 ext4_mark_inode_dirty(handle, inode); 5013 ext4_journal_stop(handle); 5014 5015 out: 5016 return dquot_quota_off(sb, type); 5017 } 5018 5019 /* 5020 * quota_off function that is used when QUOTA feature is set. 5021 */ 5022 static int ext4_quota_off_sysfile(struct super_block *sb, int type) 5023 { 5024 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 5025 return -EINVAL; 5026 5027 /* Disable only the limits. */ 5028 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED); 5029 } 5030 5031 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5032 * acquiring the locks... As quota files are never truncated and quota code 5033 * itself serializes the operations (and no one else should touch the files) 5034 * we don't have to be afraid of races */ 5035 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5036 size_t len, loff_t off) 5037 { 5038 struct inode *inode = sb_dqopt(sb)->files[type]; 5039 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5040 int err = 0; 5041 int offset = off & (sb->s_blocksize - 1); 5042 int tocopy; 5043 size_t toread; 5044 struct buffer_head *bh; 5045 loff_t i_size = i_size_read(inode); 5046 5047 if (off > i_size) 5048 return 0; 5049 if (off+len > i_size) 5050 len = i_size-off; 5051 toread = len; 5052 while (toread > 0) { 5053 tocopy = sb->s_blocksize - offset < toread ? 5054 sb->s_blocksize - offset : toread; 5055 bh = ext4_bread(NULL, inode, blk, 0, &err); 5056 if (err) 5057 return err; 5058 if (!bh) /* A hole? */ 5059 memset(data, 0, tocopy); 5060 else 5061 memcpy(data, bh->b_data+offset, tocopy); 5062 brelse(bh); 5063 offset = 0; 5064 toread -= tocopy; 5065 data += tocopy; 5066 blk++; 5067 } 5068 return len; 5069 } 5070 5071 /* Write to quotafile (we know the transaction is already started and has 5072 * enough credits) */ 5073 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5074 const char *data, size_t len, loff_t off) 5075 { 5076 struct inode *inode = sb_dqopt(sb)->files[type]; 5077 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5078 int err = 0; 5079 int offset = off & (sb->s_blocksize - 1); 5080 struct buffer_head *bh; 5081 handle_t *handle = journal_current_handle(); 5082 5083 if (EXT4_SB(sb)->s_journal && !handle) { 5084 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5085 " cancelled because transaction is not started", 5086 (unsigned long long)off, (unsigned long long)len); 5087 return -EIO; 5088 } 5089 /* 5090 * Since we account only one data block in transaction credits, 5091 * then it is impossible to cross a block boundary. 5092 */ 5093 if (sb->s_blocksize - offset < len) { 5094 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5095 " cancelled because not block aligned", 5096 (unsigned long long)off, (unsigned long long)len); 5097 return -EIO; 5098 } 5099 5100 bh = ext4_bread(handle, inode, blk, 1, &err); 5101 if (!bh) 5102 goto out; 5103 err = ext4_journal_get_write_access(handle, bh); 5104 if (err) { 5105 brelse(bh); 5106 goto out; 5107 } 5108 lock_buffer(bh); 5109 memcpy(bh->b_data+offset, data, len); 5110 flush_dcache_page(bh->b_page); 5111 unlock_buffer(bh); 5112 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5113 brelse(bh); 5114 out: 5115 if (err) 5116 return err; 5117 if (inode->i_size < off + len) { 5118 i_size_write(inode, off + len); 5119 EXT4_I(inode)->i_disksize = inode->i_size; 5120 ext4_mark_inode_dirty(handle, inode); 5121 } 5122 return len; 5123 } 5124 5125 #endif 5126 5127 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5128 const char *dev_name, void *data) 5129 { 5130 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5131 } 5132 5133 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 5134 static inline void register_as_ext2(void) 5135 { 5136 int err = register_filesystem(&ext2_fs_type); 5137 if (err) 5138 printk(KERN_WARNING 5139 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 5140 } 5141 5142 static inline void unregister_as_ext2(void) 5143 { 5144 unregister_filesystem(&ext2_fs_type); 5145 } 5146 5147 static inline int ext2_feature_set_ok(struct super_block *sb) 5148 { 5149 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) 5150 return 0; 5151 if (sb->s_flags & MS_RDONLY) 5152 return 1; 5153 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) 5154 return 0; 5155 return 1; 5156 } 5157 MODULE_ALIAS("ext2"); 5158 #else 5159 static inline void register_as_ext2(void) { } 5160 static inline void unregister_as_ext2(void) { } 5161 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 5162 #endif 5163 5164 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 5165 static inline void register_as_ext3(void) 5166 { 5167 int err = register_filesystem(&ext3_fs_type); 5168 if (err) 5169 printk(KERN_WARNING 5170 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 5171 } 5172 5173 static inline void unregister_as_ext3(void) 5174 { 5175 unregister_filesystem(&ext3_fs_type); 5176 } 5177 5178 static inline int ext3_feature_set_ok(struct super_block *sb) 5179 { 5180 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) 5181 return 0; 5182 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 5183 return 0; 5184 if (sb->s_flags & MS_RDONLY) 5185 return 1; 5186 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) 5187 return 0; 5188 return 1; 5189 } 5190 MODULE_ALIAS("ext3"); 5191 #else 5192 static inline void register_as_ext3(void) { } 5193 static inline void unregister_as_ext3(void) { } 5194 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } 5195 #endif 5196 5197 static struct file_system_type ext4_fs_type = { 5198 .owner = THIS_MODULE, 5199 .name = "ext4", 5200 .mount = ext4_mount, 5201 .kill_sb = kill_block_super, 5202 .fs_flags = FS_REQUIRES_DEV, 5203 }; 5204 5205 static int __init ext4_init_feat_adverts(void) 5206 { 5207 struct ext4_features *ef; 5208 int ret = -ENOMEM; 5209 5210 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); 5211 if (!ef) 5212 goto out; 5213 5214 ef->f_kobj.kset = ext4_kset; 5215 init_completion(&ef->f_kobj_unregister); 5216 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, 5217 "features"); 5218 if (ret) { 5219 kfree(ef); 5220 goto out; 5221 } 5222 5223 ext4_feat = ef; 5224 ret = 0; 5225 out: 5226 return ret; 5227 } 5228 5229 static void ext4_exit_feat_adverts(void) 5230 { 5231 kobject_put(&ext4_feat->f_kobj); 5232 wait_for_completion(&ext4_feat->f_kobj_unregister); 5233 kfree(ext4_feat); 5234 } 5235 5236 /* Shared across all ext4 file systems */ 5237 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5238 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 5239 5240 static int __init ext4_init_fs(void) 5241 { 5242 int i, err; 5243 5244 ext4_li_info = NULL; 5245 mutex_init(&ext4_li_mtx); 5246 5247 ext4_check_flag_values(); 5248 5249 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { 5250 mutex_init(&ext4__aio_mutex[i]); 5251 init_waitqueue_head(&ext4__ioend_wq[i]); 5252 } 5253 5254 err = ext4_init_pageio(); 5255 if (err) 5256 return err; 5257 err = ext4_init_system_zone(); 5258 if (err) 5259 goto out6; 5260 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 5261 if (!ext4_kset) 5262 goto out5; 5263 ext4_proc_root = proc_mkdir("fs/ext4", NULL); 5264 5265 err = ext4_init_feat_adverts(); 5266 if (err) 5267 goto out4; 5268 5269 err = ext4_init_mballoc(); 5270 if (err) 5271 goto out3; 5272 5273 err = ext4_init_xattr(); 5274 if (err) 5275 goto out2; 5276 err = init_inodecache(); 5277 if (err) 5278 goto out1; 5279 register_as_ext3(); 5280 register_as_ext2(); 5281 err = register_filesystem(&ext4_fs_type); 5282 if (err) 5283 goto out; 5284 5285 return 0; 5286 out: 5287 unregister_as_ext2(); 5288 unregister_as_ext3(); 5289 destroy_inodecache(); 5290 out1: 5291 ext4_exit_xattr(); 5292 out2: 5293 ext4_exit_mballoc(); 5294 out3: 5295 ext4_exit_feat_adverts(); 5296 out4: 5297 if (ext4_proc_root) 5298 remove_proc_entry("fs/ext4", NULL); 5299 kset_unregister(ext4_kset); 5300 out5: 5301 ext4_exit_system_zone(); 5302 out6: 5303 ext4_exit_pageio(); 5304 return err; 5305 } 5306 5307 static void __exit ext4_exit_fs(void) 5308 { 5309 ext4_destroy_lazyinit_thread(); 5310 unregister_as_ext2(); 5311 unregister_as_ext3(); 5312 unregister_filesystem(&ext4_fs_type); 5313 destroy_inodecache(); 5314 ext4_exit_xattr(); 5315 ext4_exit_mballoc(); 5316 ext4_exit_feat_adverts(); 5317 remove_proc_entry("fs/ext4", NULL); 5318 kset_unregister(ext4_kset); 5319 ext4_exit_system_zone(); 5320 ext4_exit_pageio(); 5321 } 5322 5323 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 5324 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5325 MODULE_LICENSE("GPL"); 5326 module_init(ext4_init_fs) 5327 module_exit(ext4_exit_fs) 5328