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_commit_super(struct super_block *sb, int sync); 66 static void ext4_mark_recovery_complete(struct super_block *sb, 67 struct ext4_super_block *es); 68 static void ext4_clear_journal_err(struct super_block *sb, 69 struct ext4_super_block *es); 70 static int ext4_sync_fs(struct super_block *sb, int wait); 71 static const char *ext4_decode_error(struct super_block *sb, int errno, 72 char nbuf[16]); 73 static int ext4_remount(struct super_block *sb, int *flags, char *data); 74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 75 static int ext4_unfreeze(struct super_block *sb); 76 static void ext4_write_super(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 void *ext4_kvmalloc(size_t size, gfp_t flags) 115 { 116 void *ret; 117 118 ret = kmalloc(size, flags); 119 if (!ret) 120 ret = __vmalloc(size, flags, PAGE_KERNEL); 121 return ret; 122 } 123 124 void *ext4_kvzalloc(size_t size, gfp_t flags) 125 { 126 void *ret; 127 128 ret = kzalloc(size, flags); 129 if (!ret) 130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 131 return ret; 132 } 133 134 void ext4_kvfree(void *ptr) 135 { 136 if (is_vmalloc_addr(ptr)) 137 vfree(ptr); 138 else 139 kfree(ptr); 140 141 } 142 143 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 144 struct ext4_group_desc *bg) 145 { 146 return le32_to_cpu(bg->bg_block_bitmap_lo) | 147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 149 } 150 151 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 152 struct ext4_group_desc *bg) 153 { 154 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 157 } 158 159 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 160 struct ext4_group_desc *bg) 161 { 162 return le32_to_cpu(bg->bg_inode_table_lo) | 163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 165 } 166 167 __u32 ext4_free_group_clusters(struct super_block *sb, 168 struct ext4_group_desc *bg) 169 { 170 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 173 } 174 175 __u32 ext4_free_inodes_count(struct super_block *sb, 176 struct ext4_group_desc *bg) 177 { 178 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 181 } 182 183 __u32 ext4_used_dirs_count(struct super_block *sb, 184 struct ext4_group_desc *bg) 185 { 186 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 189 } 190 191 __u32 ext4_itable_unused_count(struct super_block *sb, 192 struct ext4_group_desc *bg) 193 { 194 return le16_to_cpu(bg->bg_itable_unused_lo) | 195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 197 } 198 199 void ext4_block_bitmap_set(struct super_block *sb, 200 struct ext4_group_desc *bg, ext4_fsblk_t blk) 201 { 202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 205 } 206 207 void ext4_inode_bitmap_set(struct super_block *sb, 208 struct ext4_group_desc *bg, ext4_fsblk_t blk) 209 { 210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 213 } 214 215 void ext4_inode_table_set(struct super_block *sb, 216 struct ext4_group_desc *bg, ext4_fsblk_t blk) 217 { 218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 221 } 222 223 void ext4_free_group_clusters_set(struct super_block *sb, 224 struct ext4_group_desc *bg, __u32 count) 225 { 226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 229 } 230 231 void ext4_free_inodes_set(struct super_block *sb, 232 struct ext4_group_desc *bg, __u32 count) 233 { 234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 237 } 238 239 void ext4_used_dirs_set(struct super_block *sb, 240 struct ext4_group_desc *bg, __u32 count) 241 { 242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 245 } 246 247 void ext4_itable_unused_set(struct super_block *sb, 248 struct ext4_group_desc *bg, __u32 count) 249 { 250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 253 } 254 255 256 /* Just increment the non-pointer handle value */ 257 static handle_t *ext4_get_nojournal(void) 258 { 259 handle_t *handle = current->journal_info; 260 unsigned long ref_cnt = (unsigned long)handle; 261 262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT); 263 264 ref_cnt++; 265 handle = (handle_t *)ref_cnt; 266 267 current->journal_info = handle; 268 return handle; 269 } 270 271 272 /* Decrement the non-pointer handle value */ 273 static void ext4_put_nojournal(handle_t *handle) 274 { 275 unsigned long ref_cnt = (unsigned long)handle; 276 277 BUG_ON(ref_cnt == 0); 278 279 ref_cnt--; 280 handle = (handle_t *)ref_cnt; 281 282 current->journal_info = handle; 283 } 284 285 /* 286 * Wrappers for jbd2_journal_start/end. 287 * 288 * The only special thing we need to do here is to make sure that all 289 * journal_end calls result in the superblock being marked dirty, so 290 * that sync() will call the filesystem's write_super callback if 291 * appropriate. 292 * 293 * To avoid j_barrier hold in userspace when a user calls freeze(), 294 * ext4 prevents a new handle from being started by s_frozen, which 295 * is in an upper layer. 296 */ 297 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) 298 { 299 journal_t *journal; 300 handle_t *handle; 301 302 trace_ext4_journal_start(sb, nblocks, _RET_IP_); 303 if (sb->s_flags & MS_RDONLY) 304 return ERR_PTR(-EROFS); 305 306 journal = EXT4_SB(sb)->s_journal; 307 handle = ext4_journal_current_handle(); 308 309 /* 310 * If a handle has been started, it should be allowed to 311 * finish, otherwise deadlock could happen between freeze 312 * and others(e.g. truncate) due to the restart of the 313 * journal handle if the filesystem is forzen and active 314 * handles are not stopped. 315 */ 316 if (!handle) 317 vfs_check_frozen(sb, SB_FREEZE_TRANS); 318 319 if (!journal) 320 return ext4_get_nojournal(); 321 /* 322 * Special case here: if the journal has aborted behind our 323 * backs (eg. EIO in the commit thread), then we still need to 324 * take the FS itself readonly cleanly. 325 */ 326 if (is_journal_aborted(journal)) { 327 ext4_abort(sb, "Detected aborted journal"); 328 return ERR_PTR(-EROFS); 329 } 330 return jbd2_journal_start(journal, nblocks); 331 } 332 333 /* 334 * The only special thing we need to do here is to make sure that all 335 * jbd2_journal_stop calls result in the superblock being marked dirty, so 336 * that sync() will call the filesystem's write_super callback if 337 * appropriate. 338 */ 339 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) 340 { 341 struct super_block *sb; 342 int err; 343 int rc; 344 345 if (!ext4_handle_valid(handle)) { 346 ext4_put_nojournal(handle); 347 return 0; 348 } 349 sb = handle->h_transaction->t_journal->j_private; 350 err = handle->h_err; 351 rc = jbd2_journal_stop(handle); 352 353 if (!err) 354 err = rc; 355 if (err) 356 __ext4_std_error(sb, where, line, err); 357 return err; 358 } 359 360 void ext4_journal_abort_handle(const char *caller, unsigned int line, 361 const char *err_fn, struct buffer_head *bh, 362 handle_t *handle, int err) 363 { 364 char nbuf[16]; 365 const char *errstr = ext4_decode_error(NULL, err, nbuf); 366 367 BUG_ON(!ext4_handle_valid(handle)); 368 369 if (bh) 370 BUFFER_TRACE(bh, "abort"); 371 372 if (!handle->h_err) 373 handle->h_err = err; 374 375 if (is_handle_aborted(handle)) 376 return; 377 378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n", 379 caller, line, errstr, err_fn); 380 381 jbd2_journal_abort_handle(handle); 382 } 383 384 static void __save_error_info(struct super_block *sb, const char *func, 385 unsigned int line) 386 { 387 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 388 389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 391 es->s_last_error_time = cpu_to_le32(get_seconds()); 392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 393 es->s_last_error_line = cpu_to_le32(line); 394 if (!es->s_first_error_time) { 395 es->s_first_error_time = es->s_last_error_time; 396 strncpy(es->s_first_error_func, func, 397 sizeof(es->s_first_error_func)); 398 es->s_first_error_line = cpu_to_le32(line); 399 es->s_first_error_ino = es->s_last_error_ino; 400 es->s_first_error_block = es->s_last_error_block; 401 } 402 /* 403 * Start the daily error reporting function if it hasn't been 404 * started already 405 */ 406 if (!es->s_error_count) 407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1); 409 } 410 411 static void save_error_info(struct super_block *sb, const char *func, 412 unsigned int line) 413 { 414 __save_error_info(sb, func, line); 415 ext4_commit_super(sb, 1); 416 } 417 418 /* 419 * The del_gendisk() function uninitializes the disk-specific data 420 * structures, including the bdi structure, without telling anyone 421 * else. Once this happens, any attempt to call mark_buffer_dirty() 422 * (for example, by ext4_commit_super), will cause a kernel OOPS. 423 * This is a kludge to prevent these oops until we can put in a proper 424 * hook in del_gendisk() to inform the VFS and file system layers. 425 */ 426 static int block_device_ejected(struct super_block *sb) 427 { 428 struct inode *bd_inode = sb->s_bdev->bd_inode; 429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info; 430 431 return bdi->dev == NULL; 432 } 433 434 435 /* Deal with the reporting of failure conditions on a filesystem such as 436 * inconsistencies detected or read IO failures. 437 * 438 * On ext2, we can store the error state of the filesystem in the 439 * superblock. That is not possible on ext4, because we may have other 440 * write ordering constraints on the superblock which prevent us from 441 * writing it out straight away; and given that the journal is about to 442 * be aborted, we can't rely on the current, or future, transactions to 443 * write out the superblock safely. 444 * 445 * We'll just use the jbd2_journal_abort() error code to record an error in 446 * the journal instead. On recovery, the journal will complain about 447 * that error until we've noted it down and cleared it. 448 */ 449 450 static void ext4_handle_error(struct super_block *sb) 451 { 452 if (sb->s_flags & MS_RDONLY) 453 return; 454 455 if (!test_opt(sb, ERRORS_CONT)) { 456 journal_t *journal = EXT4_SB(sb)->s_journal; 457 458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 459 if (journal) 460 jbd2_journal_abort(journal, -EIO); 461 } 462 if (test_opt(sb, ERRORS_RO)) { 463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 464 sb->s_flags |= MS_RDONLY; 465 } 466 if (test_opt(sb, ERRORS_PANIC)) 467 panic("EXT4-fs (device %s): panic forced after error\n", 468 sb->s_id); 469 } 470 471 void __ext4_error(struct super_block *sb, const char *function, 472 unsigned int line, const char *fmt, ...) 473 { 474 struct va_format vaf; 475 va_list args; 476 477 va_start(args, fmt); 478 vaf.fmt = fmt; 479 vaf.va = &args; 480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 481 sb->s_id, function, line, current->comm, &vaf); 482 va_end(args); 483 484 ext4_handle_error(sb); 485 } 486 487 void ext4_error_inode(struct inode *inode, const char *function, 488 unsigned int line, ext4_fsblk_t block, 489 const char *fmt, ...) 490 { 491 va_list args; 492 struct va_format vaf; 493 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 494 495 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 496 es->s_last_error_block = cpu_to_le64(block); 497 save_error_info(inode->i_sb, function, line); 498 va_start(args, fmt); 499 vaf.fmt = fmt; 500 vaf.va = &args; 501 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ", 502 inode->i_sb->s_id, function, line, inode->i_ino); 503 if (block) 504 printk(KERN_CONT "block %llu: ", block); 505 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf); 506 va_end(args); 507 508 ext4_handle_error(inode->i_sb); 509 } 510 511 void ext4_error_file(struct file *file, const char *function, 512 unsigned int line, ext4_fsblk_t block, 513 const char *fmt, ...) 514 { 515 va_list args; 516 struct va_format vaf; 517 struct ext4_super_block *es; 518 struct inode *inode = file->f_dentry->d_inode; 519 char pathname[80], *path; 520 521 es = EXT4_SB(inode->i_sb)->s_es; 522 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 523 save_error_info(inode->i_sb, function, line); 524 path = d_path(&(file->f_path), pathname, sizeof(pathname)); 525 if (IS_ERR(path)) 526 path = "(unknown)"; 527 printk(KERN_CRIT 528 "EXT4-fs error (device %s): %s:%d: inode #%lu: ", 529 inode->i_sb->s_id, function, line, inode->i_ino); 530 if (block) 531 printk(KERN_CONT "block %llu: ", block); 532 va_start(args, fmt); 533 vaf.fmt = fmt; 534 vaf.va = &args; 535 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf); 536 va_end(args); 537 538 ext4_handle_error(inode->i_sb); 539 } 540 541 static const char *ext4_decode_error(struct super_block *sb, int errno, 542 char nbuf[16]) 543 { 544 char *errstr = NULL; 545 546 switch (errno) { 547 case -EIO: 548 errstr = "IO failure"; 549 break; 550 case -ENOMEM: 551 errstr = "Out of memory"; 552 break; 553 case -EROFS: 554 if (!sb || (EXT4_SB(sb)->s_journal && 555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 556 errstr = "Journal has aborted"; 557 else 558 errstr = "Readonly filesystem"; 559 break; 560 default: 561 /* If the caller passed in an extra buffer for unknown 562 * errors, textualise them now. Else we just return 563 * NULL. */ 564 if (nbuf) { 565 /* Check for truncated error codes... */ 566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 567 errstr = nbuf; 568 } 569 break; 570 } 571 572 return errstr; 573 } 574 575 /* __ext4_std_error decodes expected errors from journaling functions 576 * automatically and invokes the appropriate error response. */ 577 578 void __ext4_std_error(struct super_block *sb, const char *function, 579 unsigned int line, int errno) 580 { 581 char nbuf[16]; 582 const char *errstr; 583 584 /* Special case: if the error is EROFS, and we're not already 585 * inside a transaction, then there's really no point in logging 586 * an error. */ 587 if (errno == -EROFS && journal_current_handle() == NULL && 588 (sb->s_flags & MS_RDONLY)) 589 return; 590 591 errstr = ext4_decode_error(sb, errno, nbuf); 592 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 593 sb->s_id, function, line, errstr); 594 save_error_info(sb, function, line); 595 596 ext4_handle_error(sb); 597 } 598 599 /* 600 * ext4_abort is a much stronger failure handler than ext4_error. The 601 * abort function may be used to deal with unrecoverable failures such 602 * as journal IO errors or ENOMEM at a critical moment in log management. 603 * 604 * We unconditionally force the filesystem into an ABORT|READONLY state, 605 * unless the error response on the fs has been set to panic in which 606 * case we take the easy way out and panic immediately. 607 */ 608 609 void __ext4_abort(struct super_block *sb, const char *function, 610 unsigned int line, const char *fmt, ...) 611 { 612 va_list args; 613 614 save_error_info(sb, function, line); 615 va_start(args, fmt); 616 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, 617 function, line); 618 vprintk(fmt, args); 619 printk("\n"); 620 va_end(args); 621 622 if ((sb->s_flags & MS_RDONLY) == 0) { 623 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 624 sb->s_flags |= MS_RDONLY; 625 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 626 if (EXT4_SB(sb)->s_journal) 627 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 628 save_error_info(sb, function, line); 629 } 630 if (test_opt(sb, ERRORS_PANIC)) 631 panic("EXT4-fs panic from previous error\n"); 632 } 633 634 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...) 635 { 636 struct va_format vaf; 637 va_list args; 638 639 va_start(args, fmt); 640 vaf.fmt = fmt; 641 vaf.va = &args; 642 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 643 va_end(args); 644 } 645 646 void __ext4_warning(struct super_block *sb, const char *function, 647 unsigned int line, const char *fmt, ...) 648 { 649 struct va_format vaf; 650 va_list args; 651 652 va_start(args, fmt); 653 vaf.fmt = fmt; 654 vaf.va = &args; 655 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 656 sb->s_id, function, line, &vaf); 657 va_end(args); 658 } 659 660 void __ext4_grp_locked_error(const char *function, unsigned int line, 661 struct super_block *sb, ext4_group_t grp, 662 unsigned long ino, ext4_fsblk_t block, 663 const char *fmt, ...) 664 __releases(bitlock) 665 __acquires(bitlock) 666 { 667 struct va_format vaf; 668 va_list args; 669 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 670 671 es->s_last_error_ino = cpu_to_le32(ino); 672 es->s_last_error_block = cpu_to_le64(block); 673 __save_error_info(sb, function, line); 674 675 va_start(args, fmt); 676 677 vaf.fmt = fmt; 678 vaf.va = &args; 679 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 680 sb->s_id, function, line, grp); 681 if (ino) 682 printk(KERN_CONT "inode %lu: ", ino); 683 if (block) 684 printk(KERN_CONT "block %llu:", (unsigned long long) block); 685 printk(KERN_CONT "%pV\n", &vaf); 686 va_end(args); 687 688 if (test_opt(sb, ERRORS_CONT)) { 689 ext4_commit_super(sb, 0); 690 return; 691 } 692 693 ext4_unlock_group(sb, grp); 694 ext4_handle_error(sb); 695 /* 696 * We only get here in the ERRORS_RO case; relocking the group 697 * may be dangerous, but nothing bad will happen since the 698 * filesystem will have already been marked read/only and the 699 * journal has been aborted. We return 1 as a hint to callers 700 * who might what to use the return value from 701 * ext4_grp_locked_error() to distinguish between the 702 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 703 * aggressively from the ext4 function in question, with a 704 * more appropriate error code. 705 */ 706 ext4_lock_group(sb, grp); 707 return; 708 } 709 710 void ext4_update_dynamic_rev(struct super_block *sb) 711 { 712 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 713 714 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 715 return; 716 717 ext4_warning(sb, 718 "updating to rev %d because of new feature flag, " 719 "running e2fsck is recommended", 720 EXT4_DYNAMIC_REV); 721 722 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 723 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 724 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 725 /* leave es->s_feature_*compat flags alone */ 726 /* es->s_uuid will be set by e2fsck if empty */ 727 728 /* 729 * The rest of the superblock fields should be zero, and if not it 730 * means they are likely already in use, so leave them alone. We 731 * can leave it up to e2fsck to clean up any inconsistencies there. 732 */ 733 } 734 735 /* 736 * Open the external journal device 737 */ 738 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 739 { 740 struct block_device *bdev; 741 char b[BDEVNAME_SIZE]; 742 743 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 744 if (IS_ERR(bdev)) 745 goto fail; 746 return bdev; 747 748 fail: 749 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 750 __bdevname(dev, b), PTR_ERR(bdev)); 751 return NULL; 752 } 753 754 /* 755 * Release the journal device 756 */ 757 static int ext4_blkdev_put(struct block_device *bdev) 758 { 759 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 760 } 761 762 static int ext4_blkdev_remove(struct ext4_sb_info *sbi) 763 { 764 struct block_device *bdev; 765 int ret = -ENODEV; 766 767 bdev = sbi->journal_bdev; 768 if (bdev) { 769 ret = ext4_blkdev_put(bdev); 770 sbi->journal_bdev = NULL; 771 } 772 return ret; 773 } 774 775 static inline struct inode *orphan_list_entry(struct list_head *l) 776 { 777 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 778 } 779 780 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 781 { 782 struct list_head *l; 783 784 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 785 le32_to_cpu(sbi->s_es->s_last_orphan)); 786 787 printk(KERN_ERR "sb_info orphan list:\n"); 788 list_for_each(l, &sbi->s_orphan) { 789 struct inode *inode = orphan_list_entry(l); 790 printk(KERN_ERR " " 791 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 792 inode->i_sb->s_id, inode->i_ino, inode, 793 inode->i_mode, inode->i_nlink, 794 NEXT_ORPHAN(inode)); 795 } 796 } 797 798 static void ext4_put_super(struct super_block *sb) 799 { 800 struct ext4_sb_info *sbi = EXT4_SB(sb); 801 struct ext4_super_block *es = sbi->s_es; 802 int i, err; 803 804 ext4_unregister_li_request(sb); 805 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 806 807 flush_workqueue(sbi->dio_unwritten_wq); 808 destroy_workqueue(sbi->dio_unwritten_wq); 809 810 lock_super(sb); 811 if (sb->s_dirt) 812 ext4_commit_super(sb, 1); 813 814 if (sbi->s_journal) { 815 err = jbd2_journal_destroy(sbi->s_journal); 816 sbi->s_journal = NULL; 817 if (err < 0) 818 ext4_abort(sb, "Couldn't clean up the journal"); 819 } 820 821 del_timer(&sbi->s_err_report); 822 ext4_release_system_zone(sb); 823 ext4_mb_release(sb); 824 ext4_ext_release(sb); 825 ext4_xattr_put_super(sb); 826 827 if (!(sb->s_flags & MS_RDONLY)) { 828 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 829 es->s_state = cpu_to_le16(sbi->s_mount_state); 830 ext4_commit_super(sb, 1); 831 } 832 if (sbi->s_proc) { 833 remove_proc_entry(sb->s_id, ext4_proc_root); 834 } 835 kobject_del(&sbi->s_kobj); 836 837 for (i = 0; i < sbi->s_gdb_count; i++) 838 brelse(sbi->s_group_desc[i]); 839 ext4_kvfree(sbi->s_group_desc); 840 ext4_kvfree(sbi->s_flex_groups); 841 percpu_counter_destroy(&sbi->s_freeclusters_counter); 842 percpu_counter_destroy(&sbi->s_freeinodes_counter); 843 percpu_counter_destroy(&sbi->s_dirs_counter); 844 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 845 brelse(sbi->s_sbh); 846 #ifdef CONFIG_QUOTA 847 for (i = 0; i < MAXQUOTAS; i++) 848 kfree(sbi->s_qf_names[i]); 849 #endif 850 851 /* Debugging code just in case the in-memory inode orphan list 852 * isn't empty. The on-disk one can be non-empty if we've 853 * detected an error and taken the fs readonly, but the 854 * in-memory list had better be clean by this point. */ 855 if (!list_empty(&sbi->s_orphan)) 856 dump_orphan_list(sb, sbi); 857 J_ASSERT(list_empty(&sbi->s_orphan)); 858 859 invalidate_bdev(sb->s_bdev); 860 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 861 /* 862 * Invalidate the journal device's buffers. We don't want them 863 * floating about in memory - the physical journal device may 864 * hotswapped, and it breaks the `ro-after' testing code. 865 */ 866 sync_blockdev(sbi->journal_bdev); 867 invalidate_bdev(sbi->journal_bdev); 868 ext4_blkdev_remove(sbi); 869 } 870 if (sbi->s_mmp_tsk) 871 kthread_stop(sbi->s_mmp_tsk); 872 sb->s_fs_info = NULL; 873 /* 874 * Now that we are completely done shutting down the 875 * superblock, we need to actually destroy the kobject. 876 */ 877 unlock_super(sb); 878 kobject_put(&sbi->s_kobj); 879 wait_for_completion(&sbi->s_kobj_unregister); 880 kfree(sbi->s_blockgroup_lock); 881 kfree(sbi); 882 } 883 884 static struct kmem_cache *ext4_inode_cachep; 885 886 /* 887 * Called inside transaction, so use GFP_NOFS 888 */ 889 static struct inode *ext4_alloc_inode(struct super_block *sb) 890 { 891 struct ext4_inode_info *ei; 892 893 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 894 if (!ei) 895 return NULL; 896 897 ei->vfs_inode.i_version = 1; 898 ei->vfs_inode.i_data.writeback_index = 0; 899 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); 900 INIT_LIST_HEAD(&ei->i_prealloc_list); 901 spin_lock_init(&ei->i_prealloc_lock); 902 ei->i_reserved_data_blocks = 0; 903 ei->i_reserved_meta_blocks = 0; 904 ei->i_allocated_meta_blocks = 0; 905 ei->i_da_metadata_calc_len = 0; 906 spin_lock_init(&(ei->i_block_reservation_lock)); 907 #ifdef CONFIG_QUOTA 908 ei->i_reserved_quota = 0; 909 #endif 910 ei->jinode = NULL; 911 INIT_LIST_HEAD(&ei->i_completed_io_list); 912 spin_lock_init(&ei->i_completed_io_lock); 913 ei->cur_aio_dio = NULL; 914 ei->i_sync_tid = 0; 915 ei->i_datasync_tid = 0; 916 atomic_set(&ei->i_ioend_count, 0); 917 atomic_set(&ei->i_aiodio_unwritten, 0); 918 919 return &ei->vfs_inode; 920 } 921 922 static int ext4_drop_inode(struct inode *inode) 923 { 924 int drop = generic_drop_inode(inode); 925 926 trace_ext4_drop_inode(inode, drop); 927 return drop; 928 } 929 930 static void ext4_i_callback(struct rcu_head *head) 931 { 932 struct inode *inode = container_of(head, struct inode, i_rcu); 933 INIT_LIST_HEAD(&inode->i_dentry); 934 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 935 } 936 937 static void ext4_destroy_inode(struct inode *inode) 938 { 939 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 940 ext4_msg(inode->i_sb, KERN_ERR, 941 "Inode %lu (%p): orphan list check failed!", 942 inode->i_ino, EXT4_I(inode)); 943 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 944 EXT4_I(inode), sizeof(struct ext4_inode_info), 945 true); 946 dump_stack(); 947 } 948 call_rcu(&inode->i_rcu, ext4_i_callback); 949 } 950 951 static void init_once(void *foo) 952 { 953 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 954 955 INIT_LIST_HEAD(&ei->i_orphan); 956 #ifdef CONFIG_EXT4_FS_XATTR 957 init_rwsem(&ei->xattr_sem); 958 #endif 959 init_rwsem(&ei->i_data_sem); 960 inode_init_once(&ei->vfs_inode); 961 } 962 963 static int init_inodecache(void) 964 { 965 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 966 sizeof(struct ext4_inode_info), 967 0, (SLAB_RECLAIM_ACCOUNT| 968 SLAB_MEM_SPREAD), 969 init_once); 970 if (ext4_inode_cachep == NULL) 971 return -ENOMEM; 972 return 0; 973 } 974 975 static void destroy_inodecache(void) 976 { 977 kmem_cache_destroy(ext4_inode_cachep); 978 } 979 980 void ext4_clear_inode(struct inode *inode) 981 { 982 invalidate_inode_buffers(inode); 983 end_writeback(inode); 984 dquot_drop(inode); 985 ext4_discard_preallocations(inode); 986 if (EXT4_I(inode)->jinode) { 987 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 988 EXT4_I(inode)->jinode); 989 jbd2_free_inode(EXT4_I(inode)->jinode); 990 EXT4_I(inode)->jinode = NULL; 991 } 992 } 993 994 static inline void ext4_show_quota_options(struct seq_file *seq, 995 struct super_block *sb) 996 { 997 #if defined(CONFIG_QUOTA) 998 struct ext4_sb_info *sbi = EXT4_SB(sb); 999 1000 if (sbi->s_jquota_fmt) { 1001 char *fmtname = ""; 1002 1003 switch (sbi->s_jquota_fmt) { 1004 case QFMT_VFS_OLD: 1005 fmtname = "vfsold"; 1006 break; 1007 case QFMT_VFS_V0: 1008 fmtname = "vfsv0"; 1009 break; 1010 case QFMT_VFS_V1: 1011 fmtname = "vfsv1"; 1012 break; 1013 } 1014 seq_printf(seq, ",jqfmt=%s", fmtname); 1015 } 1016 1017 if (sbi->s_qf_names[USRQUOTA]) 1018 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); 1019 1020 if (sbi->s_qf_names[GRPQUOTA]) 1021 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); 1022 1023 if (test_opt(sb, USRQUOTA)) 1024 seq_puts(seq, ",usrquota"); 1025 1026 if (test_opt(sb, GRPQUOTA)) 1027 seq_puts(seq, ",grpquota"); 1028 #endif 1029 } 1030 1031 /* 1032 * Show an option if 1033 * - it's set to a non-default value OR 1034 * - if the per-sb default is different from the global default 1035 */ 1036 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) 1037 { 1038 int def_errors; 1039 unsigned long def_mount_opts; 1040 struct super_block *sb = vfs->mnt_sb; 1041 struct ext4_sb_info *sbi = EXT4_SB(sb); 1042 struct ext4_super_block *es = sbi->s_es; 1043 1044 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 1045 def_errors = le16_to_cpu(es->s_errors); 1046 1047 if (sbi->s_sb_block != 1) 1048 seq_printf(seq, ",sb=%llu", sbi->s_sb_block); 1049 if (test_opt(sb, MINIX_DF)) 1050 seq_puts(seq, ",minixdf"); 1051 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS)) 1052 seq_puts(seq, ",grpid"); 1053 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS)) 1054 seq_puts(seq, ",nogrpid"); 1055 if (sbi->s_resuid != EXT4_DEF_RESUID || 1056 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) { 1057 seq_printf(seq, ",resuid=%u", sbi->s_resuid); 1058 } 1059 if (sbi->s_resgid != EXT4_DEF_RESGID || 1060 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) { 1061 seq_printf(seq, ",resgid=%u", sbi->s_resgid); 1062 } 1063 if (test_opt(sb, ERRORS_RO)) { 1064 if (def_errors == EXT4_ERRORS_PANIC || 1065 def_errors == EXT4_ERRORS_CONTINUE) { 1066 seq_puts(seq, ",errors=remount-ro"); 1067 } 1068 } 1069 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 1070 seq_puts(seq, ",errors=continue"); 1071 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 1072 seq_puts(seq, ",errors=panic"); 1073 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16)) 1074 seq_puts(seq, ",nouid32"); 1075 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) 1076 seq_puts(seq, ",debug"); 1077 #ifdef CONFIG_EXT4_FS_XATTR 1078 if (test_opt(sb, XATTR_USER)) 1079 seq_puts(seq, ",user_xattr"); 1080 if (!test_opt(sb, XATTR_USER)) 1081 seq_puts(seq, ",nouser_xattr"); 1082 #endif 1083 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1084 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) 1085 seq_puts(seq, ",acl"); 1086 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) 1087 seq_puts(seq, ",noacl"); 1088 #endif 1089 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 1090 seq_printf(seq, ",commit=%u", 1091 (unsigned) (sbi->s_commit_interval / HZ)); 1092 } 1093 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) { 1094 seq_printf(seq, ",min_batch_time=%u", 1095 (unsigned) sbi->s_min_batch_time); 1096 } 1097 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) { 1098 seq_printf(seq, ",max_batch_time=%u", 1099 (unsigned) sbi->s_min_batch_time); 1100 } 1101 1102 /* 1103 * We're changing the default of barrier mount option, so 1104 * let's always display its mount state so it's clear what its 1105 * status is. 1106 */ 1107 seq_puts(seq, ",barrier="); 1108 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); 1109 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) 1110 seq_puts(seq, ",journal_async_commit"); 1111 else if (test_opt(sb, JOURNAL_CHECKSUM)) 1112 seq_puts(seq, ",journal_checksum"); 1113 if (test_opt(sb, I_VERSION)) 1114 seq_puts(seq, ",i_version"); 1115 if (!test_opt(sb, DELALLOC) && 1116 !(def_mount_opts & EXT4_DEFM_NODELALLOC)) 1117 seq_puts(seq, ",nodelalloc"); 1118 1119 if (!test_opt(sb, MBLK_IO_SUBMIT)) 1120 seq_puts(seq, ",nomblk_io_submit"); 1121 if (sbi->s_stripe) 1122 seq_printf(seq, ",stripe=%lu", sbi->s_stripe); 1123 /* 1124 * journal mode get enabled in different ways 1125 * So just print the value even if we didn't specify it 1126 */ 1127 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 1128 seq_puts(seq, ",data=journal"); 1129 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 1130 seq_puts(seq, ",data=ordered"); 1131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 1132 seq_puts(seq, ",data=writeback"); 1133 1134 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 1135 seq_printf(seq, ",inode_readahead_blks=%u", 1136 sbi->s_inode_readahead_blks); 1137 1138 if (test_opt(sb, DATA_ERR_ABORT)) 1139 seq_puts(seq, ",data_err=abort"); 1140 1141 if (test_opt(sb, NO_AUTO_DA_ALLOC)) 1142 seq_puts(seq, ",noauto_da_alloc"); 1143 1144 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD)) 1145 seq_puts(seq, ",discard"); 1146 1147 if (test_opt(sb, NOLOAD)) 1148 seq_puts(seq, ",norecovery"); 1149 1150 if (test_opt(sb, DIOREAD_NOLOCK)) 1151 seq_puts(seq, ",dioread_nolock"); 1152 1153 if (test_opt(sb, BLOCK_VALIDITY) && 1154 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)) 1155 seq_puts(seq, ",block_validity"); 1156 1157 if (!test_opt(sb, INIT_INODE_TABLE)) 1158 seq_puts(seq, ",noinit_inode_table"); 1159 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT) 1160 seq_printf(seq, ",init_inode_table=%u", 1161 (unsigned) sbi->s_li_wait_mult); 1162 1163 ext4_show_quota_options(seq, sb); 1164 1165 return 0; 1166 } 1167 1168 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1169 u64 ino, u32 generation) 1170 { 1171 struct inode *inode; 1172 1173 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 1174 return ERR_PTR(-ESTALE); 1175 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 1176 return ERR_PTR(-ESTALE); 1177 1178 /* iget isn't really right if the inode is currently unallocated!! 1179 * 1180 * ext4_read_inode will return a bad_inode if the inode had been 1181 * deleted, so we should be safe. 1182 * 1183 * Currently we don't know the generation for parent directory, so 1184 * a generation of 0 means "accept any" 1185 */ 1186 inode = ext4_iget(sb, ino); 1187 if (IS_ERR(inode)) 1188 return ERR_CAST(inode); 1189 if (generation && inode->i_generation != generation) { 1190 iput(inode); 1191 return ERR_PTR(-ESTALE); 1192 } 1193 1194 return inode; 1195 } 1196 1197 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1198 int fh_len, int fh_type) 1199 { 1200 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1201 ext4_nfs_get_inode); 1202 } 1203 1204 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1205 int fh_len, int fh_type) 1206 { 1207 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1208 ext4_nfs_get_inode); 1209 } 1210 1211 /* 1212 * Try to release metadata pages (indirect blocks, directories) which are 1213 * mapped via the block device. Since these pages could have journal heads 1214 * which would prevent try_to_free_buffers() from freeing them, we must use 1215 * jbd2 layer's try_to_free_buffers() function to release them. 1216 */ 1217 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1218 gfp_t wait) 1219 { 1220 journal_t *journal = EXT4_SB(sb)->s_journal; 1221 1222 WARN_ON(PageChecked(page)); 1223 if (!page_has_buffers(page)) 1224 return 0; 1225 if (journal) 1226 return jbd2_journal_try_to_free_buffers(journal, page, 1227 wait & ~__GFP_WAIT); 1228 return try_to_free_buffers(page); 1229 } 1230 1231 #ifdef CONFIG_QUOTA 1232 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") 1233 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) 1234 1235 static int ext4_write_dquot(struct dquot *dquot); 1236 static int ext4_acquire_dquot(struct dquot *dquot); 1237 static int ext4_release_dquot(struct dquot *dquot); 1238 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1239 static int ext4_write_info(struct super_block *sb, int type); 1240 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1241 struct path *path); 1242 static int ext4_quota_off(struct super_block *sb, int type); 1243 static int ext4_quota_on_mount(struct super_block *sb, int type); 1244 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1245 size_t len, loff_t off); 1246 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1247 const char *data, size_t len, loff_t off); 1248 1249 static const struct dquot_operations ext4_quota_operations = { 1250 .get_reserved_space = ext4_get_reserved_space, 1251 .write_dquot = ext4_write_dquot, 1252 .acquire_dquot = ext4_acquire_dquot, 1253 .release_dquot = ext4_release_dquot, 1254 .mark_dirty = ext4_mark_dquot_dirty, 1255 .write_info = ext4_write_info, 1256 .alloc_dquot = dquot_alloc, 1257 .destroy_dquot = dquot_destroy, 1258 }; 1259 1260 static const struct quotactl_ops ext4_qctl_operations = { 1261 .quota_on = ext4_quota_on, 1262 .quota_off = ext4_quota_off, 1263 .quota_sync = dquot_quota_sync, 1264 .get_info = dquot_get_dqinfo, 1265 .set_info = dquot_set_dqinfo, 1266 .get_dqblk = dquot_get_dqblk, 1267 .set_dqblk = dquot_set_dqblk 1268 }; 1269 #endif 1270 1271 static const struct super_operations ext4_sops = { 1272 .alloc_inode = ext4_alloc_inode, 1273 .destroy_inode = ext4_destroy_inode, 1274 .write_inode = ext4_write_inode, 1275 .dirty_inode = ext4_dirty_inode, 1276 .drop_inode = ext4_drop_inode, 1277 .evict_inode = ext4_evict_inode, 1278 .put_super = ext4_put_super, 1279 .sync_fs = ext4_sync_fs, 1280 .freeze_fs = ext4_freeze, 1281 .unfreeze_fs = ext4_unfreeze, 1282 .statfs = ext4_statfs, 1283 .remount_fs = ext4_remount, 1284 .show_options = ext4_show_options, 1285 #ifdef CONFIG_QUOTA 1286 .quota_read = ext4_quota_read, 1287 .quota_write = ext4_quota_write, 1288 #endif 1289 .bdev_try_to_free_page = bdev_try_to_free_page, 1290 }; 1291 1292 static const struct super_operations ext4_nojournal_sops = { 1293 .alloc_inode = ext4_alloc_inode, 1294 .destroy_inode = ext4_destroy_inode, 1295 .write_inode = ext4_write_inode, 1296 .dirty_inode = ext4_dirty_inode, 1297 .drop_inode = ext4_drop_inode, 1298 .evict_inode = ext4_evict_inode, 1299 .write_super = ext4_write_super, 1300 .put_super = ext4_put_super, 1301 .statfs = ext4_statfs, 1302 .remount_fs = ext4_remount, 1303 .show_options = ext4_show_options, 1304 #ifdef CONFIG_QUOTA 1305 .quota_read = ext4_quota_read, 1306 .quota_write = ext4_quota_write, 1307 #endif 1308 .bdev_try_to_free_page = bdev_try_to_free_page, 1309 }; 1310 1311 static const struct export_operations ext4_export_ops = { 1312 .fh_to_dentry = ext4_fh_to_dentry, 1313 .fh_to_parent = ext4_fh_to_parent, 1314 .get_parent = ext4_get_parent, 1315 }; 1316 1317 enum { 1318 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1319 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1320 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, 1321 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1322 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh, 1323 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, 1324 Opt_journal_update, Opt_journal_dev, 1325 Opt_journal_checksum, Opt_journal_async_commit, 1326 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1327 Opt_data_err_abort, Opt_data_err_ignore, 1328 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1329 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1330 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, 1331 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version, 1332 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 1333 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1334 Opt_inode_readahead_blks, Opt_journal_ioprio, 1335 Opt_dioread_nolock, Opt_dioread_lock, 1336 Opt_discard, Opt_nodiscard, 1337 Opt_init_inode_table, Opt_noinit_inode_table, 1338 }; 1339 1340 static const match_table_t tokens = { 1341 {Opt_bsd_df, "bsddf"}, 1342 {Opt_minix_df, "minixdf"}, 1343 {Opt_grpid, "grpid"}, 1344 {Opt_grpid, "bsdgroups"}, 1345 {Opt_nogrpid, "nogrpid"}, 1346 {Opt_nogrpid, "sysvgroups"}, 1347 {Opt_resgid, "resgid=%u"}, 1348 {Opt_resuid, "resuid=%u"}, 1349 {Opt_sb, "sb=%u"}, 1350 {Opt_err_cont, "errors=continue"}, 1351 {Opt_err_panic, "errors=panic"}, 1352 {Opt_err_ro, "errors=remount-ro"}, 1353 {Opt_nouid32, "nouid32"}, 1354 {Opt_debug, "debug"}, 1355 {Opt_oldalloc, "oldalloc"}, 1356 {Opt_orlov, "orlov"}, 1357 {Opt_user_xattr, "user_xattr"}, 1358 {Opt_nouser_xattr, "nouser_xattr"}, 1359 {Opt_acl, "acl"}, 1360 {Opt_noacl, "noacl"}, 1361 {Opt_noload, "noload"}, 1362 {Opt_noload, "norecovery"}, 1363 {Opt_nobh, "nobh"}, 1364 {Opt_bh, "bh"}, 1365 {Opt_commit, "commit=%u"}, 1366 {Opt_min_batch_time, "min_batch_time=%u"}, 1367 {Opt_max_batch_time, "max_batch_time=%u"}, 1368 {Opt_journal_update, "journal=update"}, 1369 {Opt_journal_dev, "journal_dev=%u"}, 1370 {Opt_journal_checksum, "journal_checksum"}, 1371 {Opt_journal_async_commit, "journal_async_commit"}, 1372 {Opt_abort, "abort"}, 1373 {Opt_data_journal, "data=journal"}, 1374 {Opt_data_ordered, "data=ordered"}, 1375 {Opt_data_writeback, "data=writeback"}, 1376 {Opt_data_err_abort, "data_err=abort"}, 1377 {Opt_data_err_ignore, "data_err=ignore"}, 1378 {Opt_offusrjquota, "usrjquota="}, 1379 {Opt_usrjquota, "usrjquota=%s"}, 1380 {Opt_offgrpjquota, "grpjquota="}, 1381 {Opt_grpjquota, "grpjquota=%s"}, 1382 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1383 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1384 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1385 {Opt_grpquota, "grpquota"}, 1386 {Opt_noquota, "noquota"}, 1387 {Opt_quota, "quota"}, 1388 {Opt_usrquota, "usrquota"}, 1389 {Opt_barrier, "barrier=%u"}, 1390 {Opt_barrier, "barrier"}, 1391 {Opt_nobarrier, "nobarrier"}, 1392 {Opt_i_version, "i_version"}, 1393 {Opt_stripe, "stripe=%u"}, 1394 {Opt_resize, "resize"}, 1395 {Opt_delalloc, "delalloc"}, 1396 {Opt_nodelalloc, "nodelalloc"}, 1397 {Opt_mblk_io_submit, "mblk_io_submit"}, 1398 {Opt_nomblk_io_submit, "nomblk_io_submit"}, 1399 {Opt_block_validity, "block_validity"}, 1400 {Opt_noblock_validity, "noblock_validity"}, 1401 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1402 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1403 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1404 {Opt_auto_da_alloc, "auto_da_alloc"}, 1405 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1406 {Opt_dioread_nolock, "dioread_nolock"}, 1407 {Opt_dioread_lock, "dioread_lock"}, 1408 {Opt_discard, "discard"}, 1409 {Opt_nodiscard, "nodiscard"}, 1410 {Opt_init_inode_table, "init_itable=%u"}, 1411 {Opt_init_inode_table, "init_itable"}, 1412 {Opt_noinit_inode_table, "noinit_itable"}, 1413 {Opt_err, NULL}, 1414 }; 1415 1416 static ext4_fsblk_t get_sb_block(void **data) 1417 { 1418 ext4_fsblk_t sb_block; 1419 char *options = (char *) *data; 1420 1421 if (!options || strncmp(options, "sb=", 3) != 0) 1422 return 1; /* Default location */ 1423 1424 options += 3; 1425 /* TODO: use simple_strtoll with >32bit ext4 */ 1426 sb_block = simple_strtoul(options, &options, 0); 1427 if (*options && *options != ',') { 1428 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1429 (char *) *data); 1430 return 1; 1431 } 1432 if (*options == ',') 1433 options++; 1434 *data = (void *) options; 1435 1436 return sb_block; 1437 } 1438 1439 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1440 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1441 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1442 1443 #ifdef CONFIG_QUOTA 1444 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1445 { 1446 struct ext4_sb_info *sbi = EXT4_SB(sb); 1447 char *qname; 1448 1449 if (sb_any_quota_loaded(sb) && 1450 !sbi->s_qf_names[qtype]) { 1451 ext4_msg(sb, KERN_ERR, 1452 "Cannot change journaled " 1453 "quota options when quota turned on"); 1454 return 0; 1455 } 1456 qname = match_strdup(args); 1457 if (!qname) { 1458 ext4_msg(sb, KERN_ERR, 1459 "Not enough memory for storing quotafile name"); 1460 return 0; 1461 } 1462 if (sbi->s_qf_names[qtype] && 1463 strcmp(sbi->s_qf_names[qtype], qname)) { 1464 ext4_msg(sb, KERN_ERR, 1465 "%s quota file already specified", QTYPE2NAME(qtype)); 1466 kfree(qname); 1467 return 0; 1468 } 1469 sbi->s_qf_names[qtype] = qname; 1470 if (strchr(sbi->s_qf_names[qtype], '/')) { 1471 ext4_msg(sb, KERN_ERR, 1472 "quotafile must be on filesystem root"); 1473 kfree(sbi->s_qf_names[qtype]); 1474 sbi->s_qf_names[qtype] = NULL; 1475 return 0; 1476 } 1477 set_opt(sb, QUOTA); 1478 return 1; 1479 } 1480 1481 static int clear_qf_name(struct super_block *sb, int qtype) 1482 { 1483 1484 struct ext4_sb_info *sbi = EXT4_SB(sb); 1485 1486 if (sb_any_quota_loaded(sb) && 1487 sbi->s_qf_names[qtype]) { 1488 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1489 " when quota turned on"); 1490 return 0; 1491 } 1492 /* 1493 * The space will be released later when all options are confirmed 1494 * to be correct 1495 */ 1496 sbi->s_qf_names[qtype] = NULL; 1497 return 1; 1498 } 1499 #endif 1500 1501 static int parse_options(char *options, struct super_block *sb, 1502 unsigned long *journal_devnum, 1503 unsigned int *journal_ioprio, 1504 ext4_fsblk_t *n_blocks_count, int is_remount) 1505 { 1506 struct ext4_sb_info *sbi = EXT4_SB(sb); 1507 char *p; 1508 substring_t args[MAX_OPT_ARGS]; 1509 int data_opt = 0; 1510 int option; 1511 #ifdef CONFIG_QUOTA 1512 int qfmt; 1513 #endif 1514 1515 if (!options) 1516 return 1; 1517 1518 while ((p = strsep(&options, ",")) != NULL) { 1519 int token; 1520 if (!*p) 1521 continue; 1522 1523 /* 1524 * Initialize args struct so we know whether arg was 1525 * found; some options take optional arguments. 1526 */ 1527 args[0].to = args[0].from = NULL; 1528 token = match_token(p, tokens, args); 1529 switch (token) { 1530 case Opt_bsd_df: 1531 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); 1532 clear_opt(sb, MINIX_DF); 1533 break; 1534 case Opt_minix_df: 1535 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); 1536 set_opt(sb, MINIX_DF); 1537 1538 break; 1539 case Opt_grpid: 1540 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); 1541 set_opt(sb, GRPID); 1542 1543 break; 1544 case Opt_nogrpid: 1545 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); 1546 clear_opt(sb, GRPID); 1547 1548 break; 1549 case Opt_resuid: 1550 if (match_int(&args[0], &option)) 1551 return 0; 1552 sbi->s_resuid = option; 1553 break; 1554 case Opt_resgid: 1555 if (match_int(&args[0], &option)) 1556 return 0; 1557 sbi->s_resgid = option; 1558 break; 1559 case Opt_sb: 1560 /* handled by get_sb_block() instead of here */ 1561 /* *sb_block = match_int(&args[0]); */ 1562 break; 1563 case Opt_err_panic: 1564 clear_opt(sb, ERRORS_CONT); 1565 clear_opt(sb, ERRORS_RO); 1566 set_opt(sb, ERRORS_PANIC); 1567 break; 1568 case Opt_err_ro: 1569 clear_opt(sb, ERRORS_CONT); 1570 clear_opt(sb, ERRORS_PANIC); 1571 set_opt(sb, ERRORS_RO); 1572 break; 1573 case Opt_err_cont: 1574 clear_opt(sb, ERRORS_RO); 1575 clear_opt(sb, ERRORS_PANIC); 1576 set_opt(sb, ERRORS_CONT); 1577 break; 1578 case Opt_nouid32: 1579 set_opt(sb, NO_UID32); 1580 break; 1581 case Opt_debug: 1582 set_opt(sb, DEBUG); 1583 break; 1584 case Opt_oldalloc: 1585 ext4_msg(sb, KERN_WARNING, 1586 "Ignoring deprecated oldalloc option"); 1587 break; 1588 case Opt_orlov: 1589 ext4_msg(sb, KERN_WARNING, 1590 "Ignoring deprecated orlov option"); 1591 break; 1592 #ifdef CONFIG_EXT4_FS_XATTR 1593 case Opt_user_xattr: 1594 set_opt(sb, XATTR_USER); 1595 break; 1596 case Opt_nouser_xattr: 1597 clear_opt(sb, XATTR_USER); 1598 break; 1599 #else 1600 case Opt_user_xattr: 1601 case Opt_nouser_xattr: 1602 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported"); 1603 break; 1604 #endif 1605 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1606 case Opt_acl: 1607 set_opt(sb, POSIX_ACL); 1608 break; 1609 case Opt_noacl: 1610 clear_opt(sb, POSIX_ACL); 1611 break; 1612 #else 1613 case Opt_acl: 1614 case Opt_noacl: 1615 ext4_msg(sb, KERN_ERR, "(no)acl options not supported"); 1616 break; 1617 #endif 1618 case Opt_journal_update: 1619 /* @@@ FIXME */ 1620 /* Eventually we will want to be able to create 1621 a journal file here. For now, only allow the 1622 user to specify an existing inode to be the 1623 journal file. */ 1624 if (is_remount) { 1625 ext4_msg(sb, KERN_ERR, 1626 "Cannot specify journal on remount"); 1627 return 0; 1628 } 1629 set_opt(sb, UPDATE_JOURNAL); 1630 break; 1631 case Opt_journal_dev: 1632 if (is_remount) { 1633 ext4_msg(sb, KERN_ERR, 1634 "Cannot specify journal on remount"); 1635 return 0; 1636 } 1637 if (match_int(&args[0], &option)) 1638 return 0; 1639 *journal_devnum = option; 1640 break; 1641 case Opt_journal_checksum: 1642 set_opt(sb, JOURNAL_CHECKSUM); 1643 break; 1644 case Opt_journal_async_commit: 1645 set_opt(sb, JOURNAL_ASYNC_COMMIT); 1646 set_opt(sb, JOURNAL_CHECKSUM); 1647 break; 1648 case Opt_noload: 1649 set_opt(sb, NOLOAD); 1650 break; 1651 case Opt_commit: 1652 if (match_int(&args[0], &option)) 1653 return 0; 1654 if (option < 0) 1655 return 0; 1656 if (option == 0) 1657 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1658 sbi->s_commit_interval = HZ * option; 1659 break; 1660 case Opt_max_batch_time: 1661 if (match_int(&args[0], &option)) 1662 return 0; 1663 if (option < 0) 1664 return 0; 1665 if (option == 0) 1666 option = EXT4_DEF_MAX_BATCH_TIME; 1667 sbi->s_max_batch_time = option; 1668 break; 1669 case Opt_min_batch_time: 1670 if (match_int(&args[0], &option)) 1671 return 0; 1672 if (option < 0) 1673 return 0; 1674 sbi->s_min_batch_time = option; 1675 break; 1676 case Opt_data_journal: 1677 data_opt = EXT4_MOUNT_JOURNAL_DATA; 1678 goto datacheck; 1679 case Opt_data_ordered: 1680 data_opt = EXT4_MOUNT_ORDERED_DATA; 1681 goto datacheck; 1682 case Opt_data_writeback: 1683 data_opt = EXT4_MOUNT_WRITEBACK_DATA; 1684 datacheck: 1685 if (is_remount) { 1686 if (test_opt(sb, DATA_FLAGS) != data_opt) { 1687 ext4_msg(sb, KERN_ERR, 1688 "Cannot change data mode on remount"); 1689 return 0; 1690 } 1691 } else { 1692 clear_opt(sb, DATA_FLAGS); 1693 sbi->s_mount_opt |= data_opt; 1694 } 1695 break; 1696 case Opt_data_err_abort: 1697 set_opt(sb, DATA_ERR_ABORT); 1698 break; 1699 case Opt_data_err_ignore: 1700 clear_opt(sb, DATA_ERR_ABORT); 1701 break; 1702 #ifdef CONFIG_QUOTA 1703 case Opt_usrjquota: 1704 if (!set_qf_name(sb, USRQUOTA, &args[0])) 1705 return 0; 1706 break; 1707 case Opt_grpjquota: 1708 if (!set_qf_name(sb, GRPQUOTA, &args[0])) 1709 return 0; 1710 break; 1711 case Opt_offusrjquota: 1712 if (!clear_qf_name(sb, USRQUOTA)) 1713 return 0; 1714 break; 1715 case Opt_offgrpjquota: 1716 if (!clear_qf_name(sb, GRPQUOTA)) 1717 return 0; 1718 break; 1719 1720 case Opt_jqfmt_vfsold: 1721 qfmt = QFMT_VFS_OLD; 1722 goto set_qf_format; 1723 case Opt_jqfmt_vfsv0: 1724 qfmt = QFMT_VFS_V0; 1725 goto set_qf_format; 1726 case Opt_jqfmt_vfsv1: 1727 qfmt = QFMT_VFS_V1; 1728 set_qf_format: 1729 if (sb_any_quota_loaded(sb) && 1730 sbi->s_jquota_fmt != qfmt) { 1731 ext4_msg(sb, KERN_ERR, "Cannot change " 1732 "journaled quota options when " 1733 "quota turned on"); 1734 return 0; 1735 } 1736 sbi->s_jquota_fmt = qfmt; 1737 break; 1738 case Opt_quota: 1739 case Opt_usrquota: 1740 set_opt(sb, QUOTA); 1741 set_opt(sb, USRQUOTA); 1742 break; 1743 case Opt_grpquota: 1744 set_opt(sb, QUOTA); 1745 set_opt(sb, GRPQUOTA); 1746 break; 1747 case Opt_noquota: 1748 if (sb_any_quota_loaded(sb)) { 1749 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1750 "options when quota turned on"); 1751 return 0; 1752 } 1753 clear_opt(sb, QUOTA); 1754 clear_opt(sb, USRQUOTA); 1755 clear_opt(sb, GRPQUOTA); 1756 break; 1757 #else 1758 case Opt_quota: 1759 case Opt_usrquota: 1760 case Opt_grpquota: 1761 ext4_msg(sb, KERN_ERR, 1762 "quota options not supported"); 1763 break; 1764 case Opt_usrjquota: 1765 case Opt_grpjquota: 1766 case Opt_offusrjquota: 1767 case Opt_offgrpjquota: 1768 case Opt_jqfmt_vfsold: 1769 case Opt_jqfmt_vfsv0: 1770 case Opt_jqfmt_vfsv1: 1771 ext4_msg(sb, KERN_ERR, 1772 "journaled quota options not supported"); 1773 break; 1774 case Opt_noquota: 1775 break; 1776 #endif 1777 case Opt_abort: 1778 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1779 break; 1780 case Opt_nobarrier: 1781 clear_opt(sb, BARRIER); 1782 break; 1783 case Opt_barrier: 1784 if (args[0].from) { 1785 if (match_int(&args[0], &option)) 1786 return 0; 1787 } else 1788 option = 1; /* No argument, default to 1 */ 1789 if (option) 1790 set_opt(sb, BARRIER); 1791 else 1792 clear_opt(sb, BARRIER); 1793 break; 1794 case Opt_ignore: 1795 break; 1796 case Opt_resize: 1797 if (!is_remount) { 1798 ext4_msg(sb, KERN_ERR, 1799 "resize option only available " 1800 "for remount"); 1801 return 0; 1802 } 1803 if (match_int(&args[0], &option) != 0) 1804 return 0; 1805 *n_blocks_count = option; 1806 break; 1807 case Opt_nobh: 1808 ext4_msg(sb, KERN_WARNING, 1809 "Ignoring deprecated nobh option"); 1810 break; 1811 case Opt_bh: 1812 ext4_msg(sb, KERN_WARNING, 1813 "Ignoring deprecated bh option"); 1814 break; 1815 case Opt_i_version: 1816 set_opt(sb, I_VERSION); 1817 sb->s_flags |= MS_I_VERSION; 1818 break; 1819 case Opt_nodelalloc: 1820 clear_opt(sb, DELALLOC); 1821 clear_opt2(sb, EXPLICIT_DELALLOC); 1822 break; 1823 case Opt_mblk_io_submit: 1824 set_opt(sb, MBLK_IO_SUBMIT); 1825 break; 1826 case Opt_nomblk_io_submit: 1827 clear_opt(sb, MBLK_IO_SUBMIT); 1828 break; 1829 case Opt_stripe: 1830 if (match_int(&args[0], &option)) 1831 return 0; 1832 if (option < 0) 1833 return 0; 1834 sbi->s_stripe = option; 1835 break; 1836 case Opt_delalloc: 1837 set_opt(sb, DELALLOC); 1838 set_opt2(sb, EXPLICIT_DELALLOC); 1839 break; 1840 case Opt_block_validity: 1841 set_opt(sb, BLOCK_VALIDITY); 1842 break; 1843 case Opt_noblock_validity: 1844 clear_opt(sb, BLOCK_VALIDITY); 1845 break; 1846 case Opt_inode_readahead_blks: 1847 if (match_int(&args[0], &option)) 1848 return 0; 1849 if (option < 0 || option > (1 << 30)) 1850 return 0; 1851 if (option && !is_power_of_2(option)) { 1852 ext4_msg(sb, KERN_ERR, 1853 "EXT4-fs: inode_readahead_blks" 1854 " must be a power of 2"); 1855 return 0; 1856 } 1857 sbi->s_inode_readahead_blks = option; 1858 break; 1859 case Opt_journal_ioprio: 1860 if (match_int(&args[0], &option)) 1861 return 0; 1862 if (option < 0 || option > 7) 1863 break; 1864 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 1865 option); 1866 break; 1867 case Opt_noauto_da_alloc: 1868 set_opt(sb, NO_AUTO_DA_ALLOC); 1869 break; 1870 case Opt_auto_da_alloc: 1871 if (args[0].from) { 1872 if (match_int(&args[0], &option)) 1873 return 0; 1874 } else 1875 option = 1; /* No argument, default to 1 */ 1876 if (option) 1877 clear_opt(sb, NO_AUTO_DA_ALLOC); 1878 else 1879 set_opt(sb,NO_AUTO_DA_ALLOC); 1880 break; 1881 case Opt_discard: 1882 set_opt(sb, DISCARD); 1883 break; 1884 case Opt_nodiscard: 1885 clear_opt(sb, DISCARD); 1886 break; 1887 case Opt_dioread_nolock: 1888 set_opt(sb, DIOREAD_NOLOCK); 1889 break; 1890 case Opt_dioread_lock: 1891 clear_opt(sb, DIOREAD_NOLOCK); 1892 break; 1893 case Opt_init_inode_table: 1894 set_opt(sb, INIT_INODE_TABLE); 1895 if (args[0].from) { 1896 if (match_int(&args[0], &option)) 1897 return 0; 1898 } else 1899 option = EXT4_DEF_LI_WAIT_MULT; 1900 if (option < 0) 1901 return 0; 1902 sbi->s_li_wait_mult = option; 1903 break; 1904 case Opt_noinit_inode_table: 1905 clear_opt(sb, INIT_INODE_TABLE); 1906 break; 1907 default: 1908 ext4_msg(sb, KERN_ERR, 1909 "Unrecognized mount option \"%s\" " 1910 "or missing value", p); 1911 return 0; 1912 } 1913 } 1914 #ifdef CONFIG_QUOTA 1915 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1916 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1917 clear_opt(sb, USRQUOTA); 1918 1919 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1920 clear_opt(sb, GRPQUOTA); 1921 1922 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1923 ext4_msg(sb, KERN_ERR, "old and new quota " 1924 "format mixing"); 1925 return 0; 1926 } 1927 1928 if (!sbi->s_jquota_fmt) { 1929 ext4_msg(sb, KERN_ERR, "journaled quota format " 1930 "not specified"); 1931 return 0; 1932 } 1933 } else { 1934 if (sbi->s_jquota_fmt) { 1935 ext4_msg(sb, KERN_ERR, "journaled quota format " 1936 "specified with no journaling " 1937 "enabled"); 1938 return 0; 1939 } 1940 } 1941 #endif 1942 return 1; 1943 } 1944 1945 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 1946 int read_only) 1947 { 1948 struct ext4_sb_info *sbi = EXT4_SB(sb); 1949 int res = 0; 1950 1951 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 1952 ext4_msg(sb, KERN_ERR, "revision level too high, " 1953 "forcing read-only mode"); 1954 res = MS_RDONLY; 1955 } 1956 if (read_only) 1957 goto done; 1958 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1960 "running e2fsck is recommended"); 1961 else if ((sbi->s_mount_state & EXT4_ERROR_FS)) 1962 ext4_msg(sb, KERN_WARNING, 1963 "warning: mounting fs with errors, " 1964 "running e2fsck is recommended"); 1965 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 1966 le16_to_cpu(es->s_mnt_count) >= 1967 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 1968 ext4_msg(sb, KERN_WARNING, 1969 "warning: maximal mount count reached, " 1970 "running e2fsck is recommended"); 1971 else if (le32_to_cpu(es->s_checkinterval) && 1972 (le32_to_cpu(es->s_lastcheck) + 1973 le32_to_cpu(es->s_checkinterval) <= get_seconds())) 1974 ext4_msg(sb, KERN_WARNING, 1975 "warning: checktime reached, " 1976 "running e2fsck is recommended"); 1977 if (!sbi->s_journal) 1978 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 1979 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1980 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1981 le16_add_cpu(&es->s_mnt_count, 1); 1982 es->s_mtime = cpu_to_le32(get_seconds()); 1983 ext4_update_dynamic_rev(sb); 1984 if (sbi->s_journal) 1985 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1986 1987 ext4_commit_super(sb, 1); 1988 done: 1989 if (test_opt(sb, DEBUG)) 1990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1992 sb->s_blocksize, 1993 sbi->s_groups_count, 1994 EXT4_BLOCKS_PER_GROUP(sb), 1995 EXT4_INODES_PER_GROUP(sb), 1996 sbi->s_mount_opt, sbi->s_mount_opt2); 1997 1998 cleancache_init_fs(sb); 1999 return res; 2000 } 2001 2002 static int ext4_fill_flex_info(struct super_block *sb) 2003 { 2004 struct ext4_sb_info *sbi = EXT4_SB(sb); 2005 struct ext4_group_desc *gdp = NULL; 2006 ext4_group_t flex_group_count; 2007 ext4_group_t flex_group; 2008 int groups_per_flex = 0; 2009 size_t size; 2010 int i; 2011 2012 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2013 groups_per_flex = 1 << sbi->s_log_groups_per_flex; 2014 2015 if (groups_per_flex < 2) { 2016 sbi->s_log_groups_per_flex = 0; 2017 return 1; 2018 } 2019 2020 /* We allocate both existing and potentially added groups */ 2021 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + 2022 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << 2023 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; 2024 size = flex_group_count * sizeof(struct flex_groups); 2025 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL); 2026 if (sbi->s_flex_groups == NULL) { 2027 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups", 2028 flex_group_count); 2029 goto failed; 2030 } 2031 2032 for (i = 0; i < sbi->s_groups_count; i++) { 2033 gdp = ext4_get_group_desc(sb, i, NULL); 2034 2035 flex_group = ext4_flex_group(sbi, i); 2036 atomic_add(ext4_free_inodes_count(sb, gdp), 2037 &sbi->s_flex_groups[flex_group].free_inodes); 2038 atomic_add(ext4_free_group_clusters(sb, gdp), 2039 &sbi->s_flex_groups[flex_group].free_clusters); 2040 atomic_add(ext4_used_dirs_count(sb, gdp), 2041 &sbi->s_flex_groups[flex_group].used_dirs); 2042 } 2043 2044 return 1; 2045 failed: 2046 return 0; 2047 } 2048 2049 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, 2050 struct ext4_group_desc *gdp) 2051 { 2052 __u16 crc = 0; 2053 2054 if (sbi->s_es->s_feature_ro_compat & 2055 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { 2056 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2057 __le32 le_group = cpu_to_le32(block_group); 2058 2059 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2060 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2061 crc = crc16(crc, (__u8 *)gdp, offset); 2062 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2063 /* for checksum of struct ext4_group_desc do the rest...*/ 2064 if ((sbi->s_es->s_feature_incompat & 2065 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && 2066 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2067 crc = crc16(crc, (__u8 *)gdp + offset, 2068 le16_to_cpu(sbi->s_es->s_desc_size) - 2069 offset); 2070 } 2071 2072 return cpu_to_le16(crc); 2073 } 2074 2075 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group, 2076 struct ext4_group_desc *gdp) 2077 { 2078 if ((sbi->s_es->s_feature_ro_compat & 2079 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) && 2080 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp))) 2081 return 0; 2082 2083 return 1; 2084 } 2085 2086 /* Called at mount-time, super-block is locked */ 2087 static int ext4_check_descriptors(struct super_block *sb, 2088 ext4_group_t *first_not_zeroed) 2089 { 2090 struct ext4_sb_info *sbi = EXT4_SB(sb); 2091 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2092 ext4_fsblk_t last_block; 2093 ext4_fsblk_t block_bitmap; 2094 ext4_fsblk_t inode_bitmap; 2095 ext4_fsblk_t inode_table; 2096 int flexbg_flag = 0; 2097 ext4_group_t i, grp = sbi->s_groups_count; 2098 2099 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 2100 flexbg_flag = 1; 2101 2102 ext4_debug("Checking group descriptors"); 2103 2104 for (i = 0; i < sbi->s_groups_count; i++) { 2105 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2106 2107 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2108 last_block = ext4_blocks_count(sbi->s_es) - 1; 2109 else 2110 last_block = first_block + 2111 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2112 2113 if ((grp == sbi->s_groups_count) && 2114 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2115 grp = i; 2116 2117 block_bitmap = ext4_block_bitmap(sb, gdp); 2118 if (block_bitmap < first_block || block_bitmap > last_block) { 2119 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2120 "Block bitmap for group %u not in group " 2121 "(block %llu)!", i, block_bitmap); 2122 return 0; 2123 } 2124 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2125 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2126 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2127 "Inode bitmap for group %u not in group " 2128 "(block %llu)!", i, inode_bitmap); 2129 return 0; 2130 } 2131 inode_table = ext4_inode_table(sb, gdp); 2132 if (inode_table < first_block || 2133 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2134 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2135 "Inode table for group %u not in group " 2136 "(block %llu)!", i, inode_table); 2137 return 0; 2138 } 2139 ext4_lock_group(sb, i); 2140 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { 2141 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2142 "Checksum for group %u failed (%u!=%u)", 2143 i, le16_to_cpu(ext4_group_desc_csum(sbi, i, 2144 gdp)), le16_to_cpu(gdp->bg_checksum)); 2145 if (!(sb->s_flags & MS_RDONLY)) { 2146 ext4_unlock_group(sb, i); 2147 return 0; 2148 } 2149 } 2150 ext4_unlock_group(sb, i); 2151 if (!flexbg_flag) 2152 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2153 } 2154 if (NULL != first_not_zeroed) 2155 *first_not_zeroed = grp; 2156 2157 ext4_free_blocks_count_set(sbi->s_es, 2158 EXT4_C2B(sbi, ext4_count_free_clusters(sb))); 2159 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2160 return 1; 2161 } 2162 2163 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2164 * the superblock) which were deleted from all directories, but held open by 2165 * a process at the time of a crash. We walk the list and try to delete these 2166 * inodes at recovery time (only with a read-write filesystem). 2167 * 2168 * In order to keep the orphan inode chain consistent during traversal (in 2169 * case of crash during recovery), we link each inode into the superblock 2170 * orphan list_head and handle it the same way as an inode deletion during 2171 * normal operation (which journals the operations for us). 2172 * 2173 * We only do an iget() and an iput() on each inode, which is very safe if we 2174 * accidentally point at an in-use or already deleted inode. The worst that 2175 * can happen in this case is that we get a "bit already cleared" message from 2176 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2177 * e2fsck was run on this filesystem, and it must have already done the orphan 2178 * inode cleanup for us, so we can safely abort without any further action. 2179 */ 2180 static void ext4_orphan_cleanup(struct super_block *sb, 2181 struct ext4_super_block *es) 2182 { 2183 unsigned int s_flags = sb->s_flags; 2184 int nr_orphans = 0, nr_truncates = 0; 2185 #ifdef CONFIG_QUOTA 2186 int i; 2187 #endif 2188 if (!es->s_last_orphan) { 2189 jbd_debug(4, "no orphan inodes to clean up\n"); 2190 return; 2191 } 2192 2193 if (bdev_read_only(sb->s_bdev)) { 2194 ext4_msg(sb, KERN_ERR, "write access " 2195 "unavailable, skipping orphan cleanup"); 2196 return; 2197 } 2198 2199 /* Check if feature set would not allow a r/w mount */ 2200 if (!ext4_feature_set_ok(sb, 0)) { 2201 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2202 "unknown ROCOMPAT features"); 2203 return; 2204 } 2205 2206 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2207 if (es->s_last_orphan) 2208 jbd_debug(1, "Errors on filesystem, " 2209 "clearing orphan list.\n"); 2210 es->s_last_orphan = 0; 2211 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2212 return; 2213 } 2214 2215 if (s_flags & MS_RDONLY) { 2216 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2217 sb->s_flags &= ~MS_RDONLY; 2218 } 2219 #ifdef CONFIG_QUOTA 2220 /* Needed for iput() to work correctly and not trash data */ 2221 sb->s_flags |= MS_ACTIVE; 2222 /* Turn on quotas so that they are updated correctly */ 2223 for (i = 0; i < MAXQUOTAS; i++) { 2224 if (EXT4_SB(sb)->s_qf_names[i]) { 2225 int ret = ext4_quota_on_mount(sb, i); 2226 if (ret < 0) 2227 ext4_msg(sb, KERN_ERR, 2228 "Cannot turn on journaled " 2229 "quota: error %d", ret); 2230 } 2231 } 2232 #endif 2233 2234 while (es->s_last_orphan) { 2235 struct inode *inode; 2236 2237 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2238 if (IS_ERR(inode)) { 2239 es->s_last_orphan = 0; 2240 break; 2241 } 2242 2243 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2244 dquot_initialize(inode); 2245 if (inode->i_nlink) { 2246 ext4_msg(sb, KERN_DEBUG, 2247 "%s: truncating inode %lu to %lld bytes", 2248 __func__, inode->i_ino, inode->i_size); 2249 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2250 inode->i_ino, inode->i_size); 2251 ext4_truncate(inode); 2252 nr_truncates++; 2253 } else { 2254 ext4_msg(sb, KERN_DEBUG, 2255 "%s: deleting unreferenced inode %lu", 2256 __func__, inode->i_ino); 2257 jbd_debug(2, "deleting unreferenced inode %lu\n", 2258 inode->i_ino); 2259 nr_orphans++; 2260 } 2261 iput(inode); /* The delete magic happens here! */ 2262 } 2263 2264 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2265 2266 if (nr_orphans) 2267 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2268 PLURAL(nr_orphans)); 2269 if (nr_truncates) 2270 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2271 PLURAL(nr_truncates)); 2272 #ifdef CONFIG_QUOTA 2273 /* Turn quotas off */ 2274 for (i = 0; i < MAXQUOTAS; i++) { 2275 if (sb_dqopt(sb)->files[i]) 2276 dquot_quota_off(sb, i); 2277 } 2278 #endif 2279 sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2280 } 2281 2282 /* 2283 * Maximal extent format file size. 2284 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2285 * extent format containers, within a sector_t, and within i_blocks 2286 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2287 * so that won't be a limiting factor. 2288 * 2289 * However there is other limiting factor. We do store extents in the form 2290 * of starting block and length, hence the resulting length of the extent 2291 * covering maximum file size must fit into on-disk format containers as 2292 * well. Given that length is always by 1 unit bigger than max unit (because 2293 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2294 * 2295 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2296 */ 2297 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2298 { 2299 loff_t res; 2300 loff_t upper_limit = MAX_LFS_FILESIZE; 2301 2302 /* small i_blocks in vfs inode? */ 2303 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2304 /* 2305 * CONFIG_LBDAF is not enabled implies the inode 2306 * i_block represent total blocks in 512 bytes 2307 * 32 == size of vfs inode i_blocks * 8 2308 */ 2309 upper_limit = (1LL << 32) - 1; 2310 2311 /* total blocks in file system block size */ 2312 upper_limit >>= (blkbits - 9); 2313 upper_limit <<= blkbits; 2314 } 2315 2316 /* 2317 * 32-bit extent-start container, ee_block. We lower the maxbytes 2318 * by one fs block, so ee_len can cover the extent of maximum file 2319 * size 2320 */ 2321 res = (1LL << 32) - 1; 2322 res <<= blkbits; 2323 2324 /* Sanity check against vm- & vfs- imposed limits */ 2325 if (res > upper_limit) 2326 res = upper_limit; 2327 2328 return res; 2329 } 2330 2331 /* 2332 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2333 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2334 * We need to be 1 filesystem block less than the 2^48 sector limit. 2335 */ 2336 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2337 { 2338 loff_t res = EXT4_NDIR_BLOCKS; 2339 int meta_blocks; 2340 loff_t upper_limit; 2341 /* This is calculated to be the largest file size for a dense, block 2342 * mapped file such that the file's total number of 512-byte sectors, 2343 * including data and all indirect blocks, does not exceed (2^48 - 1). 2344 * 2345 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2346 * number of 512-byte sectors of the file. 2347 */ 2348 2349 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2350 /* 2351 * !has_huge_files or CONFIG_LBDAF not enabled implies that 2352 * the inode i_block field represents total file blocks in 2353 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 2354 */ 2355 upper_limit = (1LL << 32) - 1; 2356 2357 /* total blocks in file system block size */ 2358 upper_limit >>= (bits - 9); 2359 2360 } else { 2361 /* 2362 * We use 48 bit ext4_inode i_blocks 2363 * With EXT4_HUGE_FILE_FL set the i_blocks 2364 * represent total number of blocks in 2365 * file system block size 2366 */ 2367 upper_limit = (1LL << 48) - 1; 2368 2369 } 2370 2371 /* indirect blocks */ 2372 meta_blocks = 1; 2373 /* double indirect blocks */ 2374 meta_blocks += 1 + (1LL << (bits-2)); 2375 /* tripple indirect blocks */ 2376 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2377 2378 upper_limit -= meta_blocks; 2379 upper_limit <<= bits; 2380 2381 res += 1LL << (bits-2); 2382 res += 1LL << (2*(bits-2)); 2383 res += 1LL << (3*(bits-2)); 2384 res <<= bits; 2385 if (res > upper_limit) 2386 res = upper_limit; 2387 2388 if (res > MAX_LFS_FILESIZE) 2389 res = MAX_LFS_FILESIZE; 2390 2391 return res; 2392 } 2393 2394 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2395 ext4_fsblk_t logical_sb_block, int nr) 2396 { 2397 struct ext4_sb_info *sbi = EXT4_SB(sb); 2398 ext4_group_t bg, first_meta_bg; 2399 int has_super = 0; 2400 2401 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2402 2403 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || 2404 nr < first_meta_bg) 2405 return logical_sb_block + nr + 1; 2406 bg = sbi->s_desc_per_block * nr; 2407 if (ext4_bg_has_super(sb, bg)) 2408 has_super = 1; 2409 2410 return (has_super + ext4_group_first_block_no(sb, bg)); 2411 } 2412 2413 /** 2414 * ext4_get_stripe_size: Get the stripe size. 2415 * @sbi: In memory super block info 2416 * 2417 * If we have specified it via mount option, then 2418 * use the mount option value. If the value specified at mount time is 2419 * greater than the blocks per group use the super block value. 2420 * If the super block value is greater than blocks per group return 0. 2421 * Allocator needs it be less than blocks per group. 2422 * 2423 */ 2424 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2425 { 2426 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2427 unsigned long stripe_width = 2428 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2429 int ret; 2430 2431 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2432 ret = sbi->s_stripe; 2433 else if (stripe_width <= sbi->s_blocks_per_group) 2434 ret = stripe_width; 2435 else if (stride <= sbi->s_blocks_per_group) 2436 ret = stride; 2437 else 2438 ret = 0; 2439 2440 /* 2441 * If the stripe width is 1, this makes no sense and 2442 * we set it to 0 to turn off stripe handling code. 2443 */ 2444 if (ret <= 1) 2445 ret = 0; 2446 2447 return ret; 2448 } 2449 2450 /* sysfs supprt */ 2451 2452 struct ext4_attr { 2453 struct attribute attr; 2454 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); 2455 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, 2456 const char *, size_t); 2457 int offset; 2458 }; 2459 2460 static int parse_strtoul(const char *buf, 2461 unsigned long max, unsigned long *value) 2462 { 2463 char *endp; 2464 2465 *value = simple_strtoul(skip_spaces(buf), &endp, 0); 2466 endp = skip_spaces(endp); 2467 if (*endp || *value > max) 2468 return -EINVAL; 2469 2470 return 0; 2471 } 2472 2473 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, 2474 struct ext4_sb_info *sbi, 2475 char *buf) 2476 { 2477 return snprintf(buf, PAGE_SIZE, "%llu\n", 2478 (s64) EXT4_C2B(sbi, 2479 percpu_counter_sum(&sbi->s_dirtyclusters_counter))); 2480 } 2481 2482 static ssize_t session_write_kbytes_show(struct ext4_attr *a, 2483 struct ext4_sb_info *sbi, char *buf) 2484 { 2485 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2486 2487 if (!sb->s_bdev->bd_part) 2488 return snprintf(buf, PAGE_SIZE, "0\n"); 2489 return snprintf(buf, PAGE_SIZE, "%lu\n", 2490 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2491 sbi->s_sectors_written_start) >> 1); 2492 } 2493 2494 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, 2495 struct ext4_sb_info *sbi, char *buf) 2496 { 2497 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2498 2499 if (!sb->s_bdev->bd_part) 2500 return snprintf(buf, PAGE_SIZE, "0\n"); 2501 return snprintf(buf, PAGE_SIZE, "%llu\n", 2502 (unsigned long long)(sbi->s_kbytes_written + 2503 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2504 EXT4_SB(sb)->s_sectors_written_start) >> 1))); 2505 } 2506 2507 static ssize_t extent_cache_hits_show(struct ext4_attr *a, 2508 struct ext4_sb_info *sbi, char *buf) 2509 { 2510 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits); 2511 } 2512 2513 static ssize_t extent_cache_misses_show(struct ext4_attr *a, 2514 struct ext4_sb_info *sbi, char *buf) 2515 { 2516 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses); 2517 } 2518 2519 static ssize_t inode_readahead_blks_store(struct ext4_attr *a, 2520 struct ext4_sb_info *sbi, 2521 const char *buf, size_t count) 2522 { 2523 unsigned long t; 2524 2525 if (parse_strtoul(buf, 0x40000000, &t)) 2526 return -EINVAL; 2527 2528 if (t && !is_power_of_2(t)) 2529 return -EINVAL; 2530 2531 sbi->s_inode_readahead_blks = t; 2532 return count; 2533 } 2534 2535 static ssize_t sbi_ui_show(struct ext4_attr *a, 2536 struct ext4_sb_info *sbi, char *buf) 2537 { 2538 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); 2539 2540 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 2541 } 2542 2543 static ssize_t sbi_ui_store(struct ext4_attr *a, 2544 struct ext4_sb_info *sbi, 2545 const char *buf, size_t count) 2546 { 2547 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); 2548 unsigned long t; 2549 2550 if (parse_strtoul(buf, 0xffffffff, &t)) 2551 return -EINVAL; 2552 *ui = t; 2553 return count; 2554 } 2555 2556 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ 2557 static struct ext4_attr ext4_attr_##_name = { \ 2558 .attr = {.name = __stringify(_name), .mode = _mode }, \ 2559 .show = _show, \ 2560 .store = _store, \ 2561 .offset = offsetof(struct ext4_sb_info, _elname), \ 2562 } 2563 #define EXT4_ATTR(name, mode, show, store) \ 2564 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) 2565 2566 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) 2567 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) 2568 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) 2569 #define EXT4_RW_ATTR_SBI_UI(name, elname) \ 2570 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) 2571 #define ATTR_LIST(name) &ext4_attr_##name.attr 2572 2573 EXT4_RO_ATTR(delayed_allocation_blocks); 2574 EXT4_RO_ATTR(session_write_kbytes); 2575 EXT4_RO_ATTR(lifetime_write_kbytes); 2576 EXT4_RO_ATTR(extent_cache_hits); 2577 EXT4_RO_ATTR(extent_cache_misses); 2578 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, 2579 inode_readahead_blks_store, s_inode_readahead_blks); 2580 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); 2581 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); 2582 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); 2583 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); 2584 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); 2585 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); 2586 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); 2587 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); 2588 2589 static struct attribute *ext4_attrs[] = { 2590 ATTR_LIST(delayed_allocation_blocks), 2591 ATTR_LIST(session_write_kbytes), 2592 ATTR_LIST(lifetime_write_kbytes), 2593 ATTR_LIST(extent_cache_hits), 2594 ATTR_LIST(extent_cache_misses), 2595 ATTR_LIST(inode_readahead_blks), 2596 ATTR_LIST(inode_goal), 2597 ATTR_LIST(mb_stats), 2598 ATTR_LIST(mb_max_to_scan), 2599 ATTR_LIST(mb_min_to_scan), 2600 ATTR_LIST(mb_order2_req), 2601 ATTR_LIST(mb_stream_req), 2602 ATTR_LIST(mb_group_prealloc), 2603 ATTR_LIST(max_writeback_mb_bump), 2604 NULL, 2605 }; 2606 2607 /* Features this copy of ext4 supports */ 2608 EXT4_INFO_ATTR(lazy_itable_init); 2609 EXT4_INFO_ATTR(batched_discard); 2610 2611 static struct attribute *ext4_feat_attrs[] = { 2612 ATTR_LIST(lazy_itable_init), 2613 ATTR_LIST(batched_discard), 2614 NULL, 2615 }; 2616 2617 static ssize_t ext4_attr_show(struct kobject *kobj, 2618 struct attribute *attr, char *buf) 2619 { 2620 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2621 s_kobj); 2622 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2623 2624 return a->show ? a->show(a, sbi, buf) : 0; 2625 } 2626 2627 static ssize_t ext4_attr_store(struct kobject *kobj, 2628 struct attribute *attr, 2629 const char *buf, size_t len) 2630 { 2631 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2632 s_kobj); 2633 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 2634 2635 return a->store ? a->store(a, sbi, buf, len) : 0; 2636 } 2637 2638 static void ext4_sb_release(struct kobject *kobj) 2639 { 2640 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 2641 s_kobj); 2642 complete(&sbi->s_kobj_unregister); 2643 } 2644 2645 static const struct sysfs_ops ext4_attr_ops = { 2646 .show = ext4_attr_show, 2647 .store = ext4_attr_store, 2648 }; 2649 2650 static struct kobj_type ext4_ktype = { 2651 .default_attrs = ext4_attrs, 2652 .sysfs_ops = &ext4_attr_ops, 2653 .release = ext4_sb_release, 2654 }; 2655 2656 static void ext4_feat_release(struct kobject *kobj) 2657 { 2658 complete(&ext4_feat->f_kobj_unregister); 2659 } 2660 2661 static struct kobj_type ext4_feat_ktype = { 2662 .default_attrs = ext4_feat_attrs, 2663 .sysfs_ops = &ext4_attr_ops, 2664 .release = ext4_feat_release, 2665 }; 2666 2667 /* 2668 * Check whether this filesystem can be mounted based on 2669 * the features present and the RDONLY/RDWR mount requested. 2670 * Returns 1 if this filesystem can be mounted as requested, 2671 * 0 if it cannot be. 2672 */ 2673 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2674 { 2675 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { 2676 ext4_msg(sb, KERN_ERR, 2677 "Couldn't mount because of " 2678 "unsupported optional features (%x)", 2679 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2680 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2681 return 0; 2682 } 2683 2684 if (readonly) 2685 return 1; 2686 2687 /* Check that feature set is OK for a read-write mount */ 2688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { 2689 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2690 "unsupported optional features (%x)", 2691 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2692 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2693 return 0; 2694 } 2695 /* 2696 * Large file size enabled file system can only be mounted 2697 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2698 */ 2699 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { 2700 if (sizeof(blkcnt_t) < sizeof(u64)) { 2701 ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2702 "cannot be mounted RDWR without " 2703 "CONFIG_LBDAF"); 2704 return 0; 2705 } 2706 } 2707 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) && 2708 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { 2709 ext4_msg(sb, KERN_ERR, 2710 "Can't support bigalloc feature without " 2711 "extents feature\n"); 2712 return 0; 2713 } 2714 return 1; 2715 } 2716 2717 /* 2718 * This function is called once a day if we have errors logged 2719 * on the file system 2720 */ 2721 static void print_daily_error_info(unsigned long arg) 2722 { 2723 struct super_block *sb = (struct super_block *) arg; 2724 struct ext4_sb_info *sbi; 2725 struct ext4_super_block *es; 2726 2727 sbi = EXT4_SB(sb); 2728 es = sbi->s_es; 2729 2730 if (es->s_error_count) 2731 ext4_msg(sb, KERN_NOTICE, "error count: %u", 2732 le32_to_cpu(es->s_error_count)); 2733 if (es->s_first_error_time) { 2734 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", 2735 sb->s_id, le32_to_cpu(es->s_first_error_time), 2736 (int) sizeof(es->s_first_error_func), 2737 es->s_first_error_func, 2738 le32_to_cpu(es->s_first_error_line)); 2739 if (es->s_first_error_ino) 2740 printk(": inode %u", 2741 le32_to_cpu(es->s_first_error_ino)); 2742 if (es->s_first_error_block) 2743 printk(": block %llu", (unsigned long long) 2744 le64_to_cpu(es->s_first_error_block)); 2745 printk("\n"); 2746 } 2747 if (es->s_last_error_time) { 2748 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", 2749 sb->s_id, le32_to_cpu(es->s_last_error_time), 2750 (int) sizeof(es->s_last_error_func), 2751 es->s_last_error_func, 2752 le32_to_cpu(es->s_last_error_line)); 2753 if (es->s_last_error_ino) 2754 printk(": inode %u", 2755 le32_to_cpu(es->s_last_error_ino)); 2756 if (es->s_last_error_block) 2757 printk(": block %llu", (unsigned long long) 2758 le64_to_cpu(es->s_last_error_block)); 2759 printk("\n"); 2760 } 2761 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 2762 } 2763 2764 /* Find next suitable group and run ext4_init_inode_table */ 2765 static int ext4_run_li_request(struct ext4_li_request *elr) 2766 { 2767 struct ext4_group_desc *gdp = NULL; 2768 ext4_group_t group, ngroups; 2769 struct super_block *sb; 2770 unsigned long timeout = 0; 2771 int ret = 0; 2772 2773 sb = elr->lr_super; 2774 ngroups = EXT4_SB(sb)->s_groups_count; 2775 2776 for (group = elr->lr_next_group; group < ngroups; group++) { 2777 gdp = ext4_get_group_desc(sb, group, NULL); 2778 if (!gdp) { 2779 ret = 1; 2780 break; 2781 } 2782 2783 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2784 break; 2785 } 2786 2787 if (group == ngroups) 2788 ret = 1; 2789 2790 if (!ret) { 2791 timeout = jiffies; 2792 ret = ext4_init_inode_table(sb, group, 2793 elr->lr_timeout ? 0 : 1); 2794 if (elr->lr_timeout == 0) { 2795 timeout = (jiffies - timeout) * 2796 elr->lr_sbi->s_li_wait_mult; 2797 elr->lr_timeout = timeout; 2798 } 2799 elr->lr_next_sched = jiffies + elr->lr_timeout; 2800 elr->lr_next_group = group + 1; 2801 } 2802 2803 return ret; 2804 } 2805 2806 /* 2807 * Remove lr_request from the list_request and free the 2808 * request structure. Should be called with li_list_mtx held 2809 */ 2810 static void ext4_remove_li_request(struct ext4_li_request *elr) 2811 { 2812 struct ext4_sb_info *sbi; 2813 2814 if (!elr) 2815 return; 2816 2817 sbi = elr->lr_sbi; 2818 2819 list_del(&elr->lr_request); 2820 sbi->s_li_request = NULL; 2821 kfree(elr); 2822 } 2823 2824 static void ext4_unregister_li_request(struct super_block *sb) 2825 { 2826 mutex_lock(&ext4_li_mtx); 2827 if (!ext4_li_info) { 2828 mutex_unlock(&ext4_li_mtx); 2829 return; 2830 } 2831 2832 mutex_lock(&ext4_li_info->li_list_mtx); 2833 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2834 mutex_unlock(&ext4_li_info->li_list_mtx); 2835 mutex_unlock(&ext4_li_mtx); 2836 } 2837 2838 static struct task_struct *ext4_lazyinit_task; 2839 2840 /* 2841 * This is the function where ext4lazyinit thread lives. It walks 2842 * through the request list searching for next scheduled filesystem. 2843 * When such a fs is found, run the lazy initialization request 2844 * (ext4_rn_li_request) and keep track of the time spend in this 2845 * function. Based on that time we compute next schedule time of 2846 * the request. When walking through the list is complete, compute 2847 * next waking time and put itself into sleep. 2848 */ 2849 static int ext4_lazyinit_thread(void *arg) 2850 { 2851 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2852 struct list_head *pos, *n; 2853 struct ext4_li_request *elr; 2854 unsigned long next_wakeup, cur; 2855 2856 BUG_ON(NULL == eli); 2857 2858 cont_thread: 2859 while (true) { 2860 next_wakeup = MAX_JIFFY_OFFSET; 2861 2862 mutex_lock(&eli->li_list_mtx); 2863 if (list_empty(&eli->li_request_list)) { 2864 mutex_unlock(&eli->li_list_mtx); 2865 goto exit_thread; 2866 } 2867 2868 list_for_each_safe(pos, n, &eli->li_request_list) { 2869 elr = list_entry(pos, struct ext4_li_request, 2870 lr_request); 2871 2872 if (time_after_eq(jiffies, elr->lr_next_sched)) { 2873 if (ext4_run_li_request(elr) != 0) { 2874 /* error, remove the lazy_init job */ 2875 ext4_remove_li_request(elr); 2876 continue; 2877 } 2878 } 2879 2880 if (time_before(elr->lr_next_sched, next_wakeup)) 2881 next_wakeup = elr->lr_next_sched; 2882 } 2883 mutex_unlock(&eli->li_list_mtx); 2884 2885 if (freezing(current)) 2886 refrigerator(); 2887 2888 cur = jiffies; 2889 if ((time_after_eq(cur, next_wakeup)) || 2890 (MAX_JIFFY_OFFSET == next_wakeup)) { 2891 cond_resched(); 2892 continue; 2893 } 2894 2895 schedule_timeout_interruptible(next_wakeup - cur); 2896 2897 if (kthread_should_stop()) { 2898 ext4_clear_request_list(); 2899 goto exit_thread; 2900 } 2901 } 2902 2903 exit_thread: 2904 /* 2905 * It looks like the request list is empty, but we need 2906 * to check it under the li_list_mtx lock, to prevent any 2907 * additions into it, and of course we should lock ext4_li_mtx 2908 * to atomically free the list and ext4_li_info, because at 2909 * this point another ext4 filesystem could be registering 2910 * new one. 2911 */ 2912 mutex_lock(&ext4_li_mtx); 2913 mutex_lock(&eli->li_list_mtx); 2914 if (!list_empty(&eli->li_request_list)) { 2915 mutex_unlock(&eli->li_list_mtx); 2916 mutex_unlock(&ext4_li_mtx); 2917 goto cont_thread; 2918 } 2919 mutex_unlock(&eli->li_list_mtx); 2920 kfree(ext4_li_info); 2921 ext4_li_info = NULL; 2922 mutex_unlock(&ext4_li_mtx); 2923 2924 return 0; 2925 } 2926 2927 static void ext4_clear_request_list(void) 2928 { 2929 struct list_head *pos, *n; 2930 struct ext4_li_request *elr; 2931 2932 mutex_lock(&ext4_li_info->li_list_mtx); 2933 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 2934 elr = list_entry(pos, struct ext4_li_request, 2935 lr_request); 2936 ext4_remove_li_request(elr); 2937 } 2938 mutex_unlock(&ext4_li_info->li_list_mtx); 2939 } 2940 2941 static int ext4_run_lazyinit_thread(void) 2942 { 2943 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 2944 ext4_li_info, "ext4lazyinit"); 2945 if (IS_ERR(ext4_lazyinit_task)) { 2946 int err = PTR_ERR(ext4_lazyinit_task); 2947 ext4_clear_request_list(); 2948 kfree(ext4_li_info); 2949 ext4_li_info = NULL; 2950 printk(KERN_CRIT "EXT4: error %d creating inode table " 2951 "initialization thread\n", 2952 err); 2953 return err; 2954 } 2955 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 2956 return 0; 2957 } 2958 2959 /* 2960 * Check whether it make sense to run itable init. thread or not. 2961 * If there is at least one uninitialized inode table, return 2962 * corresponding group number, else the loop goes through all 2963 * groups and return total number of groups. 2964 */ 2965 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 2966 { 2967 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 2968 struct ext4_group_desc *gdp = NULL; 2969 2970 for (group = 0; group < ngroups; group++) { 2971 gdp = ext4_get_group_desc(sb, group, NULL); 2972 if (!gdp) 2973 continue; 2974 2975 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2976 break; 2977 } 2978 2979 return group; 2980 } 2981 2982 static int ext4_li_info_new(void) 2983 { 2984 struct ext4_lazy_init *eli = NULL; 2985 2986 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 2987 if (!eli) 2988 return -ENOMEM; 2989 2990 INIT_LIST_HEAD(&eli->li_request_list); 2991 mutex_init(&eli->li_list_mtx); 2992 2993 eli->li_state |= EXT4_LAZYINIT_QUIT; 2994 2995 ext4_li_info = eli; 2996 2997 return 0; 2998 } 2999 3000 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3001 ext4_group_t start) 3002 { 3003 struct ext4_sb_info *sbi = EXT4_SB(sb); 3004 struct ext4_li_request *elr; 3005 unsigned long rnd; 3006 3007 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3008 if (!elr) 3009 return NULL; 3010 3011 elr->lr_super = sb; 3012 elr->lr_sbi = sbi; 3013 elr->lr_next_group = start; 3014 3015 /* 3016 * Randomize first schedule time of the request to 3017 * spread the inode table initialization requests 3018 * better. 3019 */ 3020 get_random_bytes(&rnd, sizeof(rnd)); 3021 elr->lr_next_sched = jiffies + (unsigned long)rnd % 3022 (EXT4_DEF_LI_MAX_START_DELAY * HZ); 3023 3024 return elr; 3025 } 3026 3027 static int ext4_register_li_request(struct super_block *sb, 3028 ext4_group_t first_not_zeroed) 3029 { 3030 struct ext4_sb_info *sbi = EXT4_SB(sb); 3031 struct ext4_li_request *elr; 3032 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3033 int ret = 0; 3034 3035 if (sbi->s_li_request != NULL) { 3036 /* 3037 * Reset timeout so it can be computed again, because 3038 * s_li_wait_mult might have changed. 3039 */ 3040 sbi->s_li_request->lr_timeout = 0; 3041 return 0; 3042 } 3043 3044 if (first_not_zeroed == ngroups || 3045 (sb->s_flags & MS_RDONLY) || 3046 !test_opt(sb, INIT_INODE_TABLE)) 3047 return 0; 3048 3049 elr = ext4_li_request_new(sb, first_not_zeroed); 3050 if (!elr) 3051 return -ENOMEM; 3052 3053 mutex_lock(&ext4_li_mtx); 3054 3055 if (NULL == ext4_li_info) { 3056 ret = ext4_li_info_new(); 3057 if (ret) 3058 goto out; 3059 } 3060 3061 mutex_lock(&ext4_li_info->li_list_mtx); 3062 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3063 mutex_unlock(&ext4_li_info->li_list_mtx); 3064 3065 sbi->s_li_request = elr; 3066 /* 3067 * set elr to NULL here since it has been inserted to 3068 * the request_list and the removal and free of it is 3069 * handled by ext4_clear_request_list from now on. 3070 */ 3071 elr = NULL; 3072 3073 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3074 ret = ext4_run_lazyinit_thread(); 3075 if (ret) 3076 goto out; 3077 } 3078 out: 3079 mutex_unlock(&ext4_li_mtx); 3080 if (ret) 3081 kfree(elr); 3082 return ret; 3083 } 3084 3085 /* 3086 * We do not need to lock anything since this is called on 3087 * module unload. 3088 */ 3089 static void ext4_destroy_lazyinit_thread(void) 3090 { 3091 /* 3092 * If thread exited earlier 3093 * there's nothing to be done. 3094 */ 3095 if (!ext4_li_info || !ext4_lazyinit_task) 3096 return; 3097 3098 kthread_stop(ext4_lazyinit_task); 3099 } 3100 3101 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3102 __releases(kernel_lock) 3103 __acquires(kernel_lock) 3104 { 3105 char *orig_data = kstrdup(data, GFP_KERNEL); 3106 struct buffer_head *bh; 3107 struct ext4_super_block *es = NULL; 3108 struct ext4_sb_info *sbi; 3109 ext4_fsblk_t block; 3110 ext4_fsblk_t sb_block = get_sb_block(&data); 3111 ext4_fsblk_t logical_sb_block; 3112 unsigned long offset = 0; 3113 unsigned long journal_devnum = 0; 3114 unsigned long def_mount_opts; 3115 struct inode *root; 3116 char *cp; 3117 const char *descr; 3118 int ret = -ENOMEM; 3119 int blocksize, clustersize; 3120 unsigned int db_count; 3121 unsigned int i; 3122 int needs_recovery, has_huge_files, has_bigalloc; 3123 __u64 blocks_count; 3124 int err; 3125 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3126 ext4_group_t first_not_zeroed; 3127 3128 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3129 if (!sbi) 3130 goto out_free_orig; 3131 3132 sbi->s_blockgroup_lock = 3133 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3134 if (!sbi->s_blockgroup_lock) { 3135 kfree(sbi); 3136 goto out_free_orig; 3137 } 3138 sb->s_fs_info = sbi; 3139 sbi->s_mount_opt = 0; 3140 sbi->s_resuid = EXT4_DEF_RESUID; 3141 sbi->s_resgid = EXT4_DEF_RESGID; 3142 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3143 sbi->s_sb_block = sb_block; 3144 if (sb->s_bdev->bd_part) 3145 sbi->s_sectors_written_start = 3146 part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3147 3148 /* Cleanup superblock name */ 3149 for (cp = sb->s_id; (cp = strchr(cp, '/'));) 3150 *cp = '!'; 3151 3152 ret = -EINVAL; 3153 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3154 if (!blocksize) { 3155 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3156 goto out_fail; 3157 } 3158 3159 /* 3160 * The ext4 superblock will not be buffer aligned for other than 1kB 3161 * block sizes. We need to calculate the offset from buffer start. 3162 */ 3163 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3164 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3165 offset = do_div(logical_sb_block, blocksize); 3166 } else { 3167 logical_sb_block = sb_block; 3168 } 3169 3170 if (!(bh = sb_bread(sb, logical_sb_block))) { 3171 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3172 goto out_fail; 3173 } 3174 /* 3175 * Note: s_es must be initialized as soon as possible because 3176 * some ext4 macro-instructions depend on its value 3177 */ 3178 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); 3179 sbi->s_es = es; 3180 sb->s_magic = le16_to_cpu(es->s_magic); 3181 if (sb->s_magic != EXT4_SUPER_MAGIC) 3182 goto cantfind_ext4; 3183 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3184 3185 /* Set defaults before we parse the mount options */ 3186 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3187 set_opt(sb, INIT_INODE_TABLE); 3188 if (def_mount_opts & EXT4_DEFM_DEBUG) 3189 set_opt(sb, DEBUG); 3190 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) { 3191 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups", 3192 "2.6.38"); 3193 set_opt(sb, GRPID); 3194 } 3195 if (def_mount_opts & EXT4_DEFM_UID16) 3196 set_opt(sb, NO_UID32); 3197 /* xattr user namespace & acls are now defaulted on */ 3198 #ifdef CONFIG_EXT4_FS_XATTR 3199 set_opt(sb, XATTR_USER); 3200 #endif 3201 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3202 set_opt(sb, POSIX_ACL); 3203 #endif 3204 set_opt(sb, MBLK_IO_SUBMIT); 3205 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3206 set_opt(sb, JOURNAL_DATA); 3207 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3208 set_opt(sb, ORDERED_DATA); 3209 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3210 set_opt(sb, WRITEBACK_DATA); 3211 3212 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3213 set_opt(sb, ERRORS_PANIC); 3214 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3215 set_opt(sb, ERRORS_CONT); 3216 else 3217 set_opt(sb, ERRORS_RO); 3218 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) 3219 set_opt(sb, BLOCK_VALIDITY); 3220 if (def_mount_opts & EXT4_DEFM_DISCARD) 3221 set_opt(sb, DISCARD); 3222 3223 sbi->s_resuid = le16_to_cpu(es->s_def_resuid); 3224 sbi->s_resgid = le16_to_cpu(es->s_def_resgid); 3225 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3226 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3227 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3228 3229 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3230 set_opt(sb, BARRIER); 3231 3232 /* 3233 * enable delayed allocation by default 3234 * Use -o nodelalloc to turn it off 3235 */ 3236 if (!IS_EXT3_SB(sb) && 3237 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3238 set_opt(sb, DELALLOC); 3239 3240 /* 3241 * set default s_li_wait_mult for lazyinit, for the case there is 3242 * no mount option specified. 3243 */ 3244 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3245 3246 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3247 &journal_devnum, &journal_ioprio, NULL, 0)) { 3248 ext4_msg(sb, KERN_WARNING, 3249 "failed to parse options in superblock: %s", 3250 sbi->s_es->s_mount_opts); 3251 } 3252 if (!parse_options((char *) data, sb, &journal_devnum, 3253 &journal_ioprio, NULL, 0)) 3254 goto failed_mount; 3255 3256 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3257 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3258 "with data=journal disables delayed " 3259 "allocation and O_DIRECT support!\n"); 3260 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3261 ext4_msg(sb, KERN_ERR, "can't mount with " 3262 "both data=journal and delalloc"); 3263 goto failed_mount; 3264 } 3265 if (test_opt(sb, DIOREAD_NOLOCK)) { 3266 ext4_msg(sb, KERN_ERR, "can't mount with " 3267 "both data=journal and delalloc"); 3268 goto failed_mount; 3269 } 3270 if (test_opt(sb, DELALLOC)) 3271 clear_opt(sb, DELALLOC); 3272 } 3273 3274 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3275 if (test_opt(sb, DIOREAD_NOLOCK)) { 3276 if (blocksize < PAGE_SIZE) { 3277 ext4_msg(sb, KERN_ERR, "can't mount with " 3278 "dioread_nolock if block size != PAGE_SIZE"); 3279 goto failed_mount; 3280 } 3281 } 3282 3283 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3284 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3285 3286 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3287 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || 3288 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || 3289 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) 3290 ext4_msg(sb, KERN_WARNING, 3291 "feature flags set on rev 0 fs, " 3292 "running e2fsck is recommended"); 3293 3294 if (IS_EXT2_SB(sb)) { 3295 if (ext2_feature_set_ok(sb)) 3296 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3297 "using the ext4 subsystem"); 3298 else { 3299 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3300 "to feature incompatibilities"); 3301 goto failed_mount; 3302 } 3303 } 3304 3305 if (IS_EXT3_SB(sb)) { 3306 if (ext3_feature_set_ok(sb)) 3307 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3308 "using the ext4 subsystem"); 3309 else { 3310 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3311 "to feature incompatibilities"); 3312 goto failed_mount; 3313 } 3314 } 3315 3316 /* 3317 * Check feature flags regardless of the revision level, since we 3318 * previously didn't change the revision level when setting the flags, 3319 * so there is a chance incompat flags are set on a rev 0 filesystem. 3320 */ 3321 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3322 goto failed_mount; 3323 3324 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3325 blocksize > EXT4_MAX_BLOCK_SIZE) { 3326 ext4_msg(sb, KERN_ERR, 3327 "Unsupported filesystem blocksize %d", blocksize); 3328 goto failed_mount; 3329 } 3330 3331 if (sb->s_blocksize != blocksize) { 3332 /* Validate the filesystem blocksize */ 3333 if (!sb_set_blocksize(sb, blocksize)) { 3334 ext4_msg(sb, KERN_ERR, "bad block size %d", 3335 blocksize); 3336 goto failed_mount; 3337 } 3338 3339 brelse(bh); 3340 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3341 offset = do_div(logical_sb_block, blocksize); 3342 bh = sb_bread(sb, logical_sb_block); 3343 if (!bh) { 3344 ext4_msg(sb, KERN_ERR, 3345 "Can't read superblock on 2nd try"); 3346 goto failed_mount; 3347 } 3348 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); 3349 sbi->s_es = es; 3350 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3351 ext4_msg(sb, KERN_ERR, 3352 "Magic mismatch, very weird!"); 3353 goto failed_mount; 3354 } 3355 } 3356 3357 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3358 EXT4_FEATURE_RO_COMPAT_HUGE_FILE); 3359 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3360 has_huge_files); 3361 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3362 3363 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3364 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3365 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3366 } else { 3367 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3368 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3369 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3370 (!is_power_of_2(sbi->s_inode_size)) || 3371 (sbi->s_inode_size > blocksize)) { 3372 ext4_msg(sb, KERN_ERR, 3373 "unsupported inode size: %d", 3374 sbi->s_inode_size); 3375 goto failed_mount; 3376 } 3377 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3378 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3379 } 3380 3381 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3382 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { 3383 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 3384 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3385 !is_power_of_2(sbi->s_desc_size)) { 3386 ext4_msg(sb, KERN_ERR, 3387 "unsupported descriptor size %lu", 3388 sbi->s_desc_size); 3389 goto failed_mount; 3390 } 3391 } else 3392 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 3393 3394 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3395 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3396 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3397 goto cantfind_ext4; 3398 3399 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3400 if (sbi->s_inodes_per_block == 0) 3401 goto cantfind_ext4; 3402 sbi->s_itb_per_group = sbi->s_inodes_per_group / 3403 sbi->s_inodes_per_block; 3404 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3405 sbi->s_sbh = bh; 3406 sbi->s_mount_state = le16_to_cpu(es->s_state); 3407 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3408 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 3409 3410 for (i = 0; i < 4; i++) 3411 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3412 sbi->s_def_hash_version = es->s_def_hash_version; 3413 i = le32_to_cpu(es->s_flags); 3414 if (i & EXT2_FLAGS_UNSIGNED_HASH) 3415 sbi->s_hash_unsigned = 3; 3416 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3417 #ifdef __CHAR_UNSIGNED__ 3418 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3419 sbi->s_hash_unsigned = 3; 3420 #else 3421 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3422 #endif 3423 sb->s_dirt = 1; 3424 } 3425 3426 /* Handle clustersize */ 3427 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3428 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3429 EXT4_FEATURE_RO_COMPAT_BIGALLOC); 3430 if (has_bigalloc) { 3431 if (clustersize < blocksize) { 3432 ext4_msg(sb, KERN_ERR, 3433 "cluster size (%d) smaller than " 3434 "block size (%d)", clustersize, blocksize); 3435 goto failed_mount; 3436 } 3437 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3438 le32_to_cpu(es->s_log_block_size); 3439 sbi->s_clusters_per_group = 3440 le32_to_cpu(es->s_clusters_per_group); 3441 if (sbi->s_clusters_per_group > blocksize * 8) { 3442 ext4_msg(sb, KERN_ERR, 3443 "#clusters per group too big: %lu", 3444 sbi->s_clusters_per_group); 3445 goto failed_mount; 3446 } 3447 if (sbi->s_blocks_per_group != 3448 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3449 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3450 "clusters per group (%lu) inconsistent", 3451 sbi->s_blocks_per_group, 3452 sbi->s_clusters_per_group); 3453 goto failed_mount; 3454 } 3455 } else { 3456 if (clustersize != blocksize) { 3457 ext4_warning(sb, "fragment/cluster size (%d) != " 3458 "block size (%d)", clustersize, 3459 blocksize); 3460 clustersize = blocksize; 3461 } 3462 if (sbi->s_blocks_per_group > blocksize * 8) { 3463 ext4_msg(sb, KERN_ERR, 3464 "#blocks per group too big: %lu", 3465 sbi->s_blocks_per_group); 3466 goto failed_mount; 3467 } 3468 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3469 sbi->s_cluster_bits = 0; 3470 } 3471 sbi->s_cluster_ratio = clustersize / blocksize; 3472 3473 if (sbi->s_inodes_per_group > blocksize * 8) { 3474 ext4_msg(sb, KERN_ERR, 3475 "#inodes per group too big: %lu", 3476 sbi->s_inodes_per_group); 3477 goto failed_mount; 3478 } 3479 3480 /* 3481 * Test whether we have more sectors than will fit in sector_t, 3482 * and whether the max offset is addressable by the page cache. 3483 */ 3484 err = generic_check_addressable(sb->s_blocksize_bits, 3485 ext4_blocks_count(es)); 3486 if (err) { 3487 ext4_msg(sb, KERN_ERR, "filesystem" 3488 " too large to mount safely on this system"); 3489 if (sizeof(sector_t) < 8) 3490 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3491 ret = err; 3492 goto failed_mount; 3493 } 3494 3495 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3496 goto cantfind_ext4; 3497 3498 /* check blocks count against device size */ 3499 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 3500 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3501 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3502 "exceeds size of device (%llu blocks)", 3503 ext4_blocks_count(es), blocks_count); 3504 goto failed_mount; 3505 } 3506 3507 /* 3508 * It makes no sense for the first data block to be beyond the end 3509 * of the filesystem. 3510 */ 3511 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3512 ext4_msg(sb, KERN_WARNING, "bad geometry: first data" 3513 "block %u is beyond end of filesystem (%llu)", 3514 le32_to_cpu(es->s_first_data_block), 3515 ext4_blocks_count(es)); 3516 goto failed_mount; 3517 } 3518 blocks_count = (ext4_blocks_count(es) - 3519 le32_to_cpu(es->s_first_data_block) + 3520 EXT4_BLOCKS_PER_GROUP(sb) - 1); 3521 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 3522 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3523 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 3524 "(block count %llu, first data block %u, " 3525 "blocks per group %lu)", sbi->s_groups_count, 3526 ext4_blocks_count(es), 3527 le32_to_cpu(es->s_first_data_block), 3528 EXT4_BLOCKS_PER_GROUP(sb)); 3529 goto failed_mount; 3530 } 3531 sbi->s_groups_count = blocks_count; 3532 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3533 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3534 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3535 EXT4_DESC_PER_BLOCK(sb); 3536 sbi->s_group_desc = ext4_kvmalloc(db_count * 3537 sizeof(struct buffer_head *), 3538 GFP_KERNEL); 3539 if (sbi->s_group_desc == NULL) { 3540 ext4_msg(sb, KERN_ERR, "not enough memory"); 3541 goto failed_mount; 3542 } 3543 3544 if (ext4_proc_root) 3545 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); 3546 3547 bgl_lock_init(sbi->s_blockgroup_lock); 3548 3549 for (i = 0; i < db_count; i++) { 3550 block = descriptor_loc(sb, logical_sb_block, i); 3551 sbi->s_group_desc[i] = sb_bread(sb, block); 3552 if (!sbi->s_group_desc[i]) { 3553 ext4_msg(sb, KERN_ERR, 3554 "can't read group descriptor %d", i); 3555 db_count = i; 3556 goto failed_mount2; 3557 } 3558 } 3559 if (!ext4_check_descriptors(sb, &first_not_zeroed)) { 3560 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3561 goto failed_mount2; 3562 } 3563 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 3564 if (!ext4_fill_flex_info(sb)) { 3565 ext4_msg(sb, KERN_ERR, 3566 "unable to initialize " 3567 "flex_bg meta info!"); 3568 goto failed_mount2; 3569 } 3570 3571 sbi->s_gdb_count = db_count; 3572 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3573 spin_lock_init(&sbi->s_next_gen_lock); 3574 3575 init_timer(&sbi->s_err_report); 3576 sbi->s_err_report.function = print_daily_error_info; 3577 sbi->s_err_report.data = (unsigned long) sb; 3578 3579 err = percpu_counter_init(&sbi->s_freeclusters_counter, 3580 ext4_count_free_clusters(sb)); 3581 if (!err) { 3582 err = percpu_counter_init(&sbi->s_freeinodes_counter, 3583 ext4_count_free_inodes(sb)); 3584 } 3585 if (!err) { 3586 err = percpu_counter_init(&sbi->s_dirs_counter, 3587 ext4_count_dirs(sb)); 3588 } 3589 if (!err) { 3590 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0); 3591 } 3592 if (err) { 3593 ext4_msg(sb, KERN_ERR, "insufficient memory"); 3594 goto failed_mount3; 3595 } 3596 3597 sbi->s_stripe = ext4_get_stripe_size(sbi); 3598 sbi->s_max_writeback_mb_bump = 128; 3599 3600 /* 3601 * set up enough so that it can read an inode 3602 */ 3603 if (!test_opt(sb, NOLOAD) && 3604 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 3605 sb->s_op = &ext4_sops; 3606 else 3607 sb->s_op = &ext4_nojournal_sops; 3608 sb->s_export_op = &ext4_export_ops; 3609 sb->s_xattr = ext4_xattr_handlers; 3610 #ifdef CONFIG_QUOTA 3611 sb->s_qcop = &ext4_qctl_operations; 3612 sb->dq_op = &ext4_quota_operations; 3613 #endif 3614 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3615 3616 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 3617 mutex_init(&sbi->s_orphan_lock); 3618 sbi->s_resize_flags = 0; 3619 3620 sb->s_root = NULL; 3621 3622 needs_recovery = (es->s_last_orphan != 0 || 3623 EXT4_HAS_INCOMPAT_FEATURE(sb, 3624 EXT4_FEATURE_INCOMPAT_RECOVER)); 3625 3626 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) && 3627 !(sb->s_flags & MS_RDONLY)) 3628 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 3629 goto failed_mount3; 3630 3631 /* 3632 * The first inode we look at is the journal inode. Don't try 3633 * root first: it may be modified in the journal! 3634 */ 3635 if (!test_opt(sb, NOLOAD) && 3636 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 3637 if (ext4_load_journal(sb, es, journal_devnum)) 3638 goto failed_mount3; 3639 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 3640 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 3641 ext4_msg(sb, KERN_ERR, "required journal recovery " 3642 "suppressed and not mounted read-only"); 3643 goto failed_mount_wq; 3644 } else { 3645 clear_opt(sb, DATA_FLAGS); 3646 sbi->s_journal = NULL; 3647 needs_recovery = 0; 3648 goto no_journal; 3649 } 3650 3651 if (ext4_blocks_count(es) > 0xffffffffULL && 3652 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3653 JBD2_FEATURE_INCOMPAT_64BIT)) { 3654 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3655 goto failed_mount_wq; 3656 } 3657 3658 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3659 jbd2_journal_set_features(sbi->s_journal, 3660 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3661 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3662 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3663 jbd2_journal_set_features(sbi->s_journal, 3664 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0); 3665 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3666 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3667 } else { 3668 jbd2_journal_clear_features(sbi->s_journal, 3669 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3670 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3671 } 3672 3673 /* We have now updated the journal if required, so we can 3674 * validate the data journaling mode. */ 3675 switch (test_opt(sb, DATA_FLAGS)) { 3676 case 0: 3677 /* No mode set, assume a default based on the journal 3678 * capabilities: ORDERED_DATA if the journal can 3679 * cope, else JOURNAL_DATA 3680 */ 3681 if (jbd2_journal_check_available_features 3682 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 3683 set_opt(sb, ORDERED_DATA); 3684 else 3685 set_opt(sb, JOURNAL_DATA); 3686 break; 3687 3688 case EXT4_MOUNT_ORDERED_DATA: 3689 case EXT4_MOUNT_WRITEBACK_DATA: 3690 if (!jbd2_journal_check_available_features 3691 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 3692 ext4_msg(sb, KERN_ERR, "Journal does not support " 3693 "requested data journaling mode"); 3694 goto failed_mount_wq; 3695 } 3696 default: 3697 break; 3698 } 3699 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 3700 3701 /* 3702 * The journal may have updated the bg summary counts, so we 3703 * need to update the global counters. 3704 */ 3705 percpu_counter_set(&sbi->s_freeclusters_counter, 3706 ext4_count_free_clusters(sb)); 3707 percpu_counter_set(&sbi->s_freeinodes_counter, 3708 ext4_count_free_inodes(sb)); 3709 percpu_counter_set(&sbi->s_dirs_counter, 3710 ext4_count_dirs(sb)); 3711 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0); 3712 3713 no_journal: 3714 /* 3715 * The maximum number of concurrent works can be high and 3716 * concurrency isn't really necessary. Limit it to 1. 3717 */ 3718 EXT4_SB(sb)->dio_unwritten_wq = 3719 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 3720 if (!EXT4_SB(sb)->dio_unwritten_wq) { 3721 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); 3722 goto failed_mount_wq; 3723 } 3724 3725 /* 3726 * The jbd2_journal_load will have done any necessary log recovery, 3727 * so we can safely mount the rest of the filesystem now. 3728 */ 3729 3730 root = ext4_iget(sb, EXT4_ROOT_INO); 3731 if (IS_ERR(root)) { 3732 ext4_msg(sb, KERN_ERR, "get root inode failed"); 3733 ret = PTR_ERR(root); 3734 root = NULL; 3735 goto failed_mount4; 3736 } 3737 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 3738 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 3739 goto failed_mount4; 3740 } 3741 sb->s_root = d_alloc_root(root); 3742 if (!sb->s_root) { 3743 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 3744 ret = -ENOMEM; 3745 goto failed_mount4; 3746 } 3747 3748 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY); 3749 3750 /* determine the minimum size of new large inodes, if present */ 3751 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 3752 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3753 EXT4_GOOD_OLD_INODE_SIZE; 3754 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3755 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { 3756 if (sbi->s_want_extra_isize < 3757 le16_to_cpu(es->s_want_extra_isize)) 3758 sbi->s_want_extra_isize = 3759 le16_to_cpu(es->s_want_extra_isize); 3760 if (sbi->s_want_extra_isize < 3761 le16_to_cpu(es->s_min_extra_isize)) 3762 sbi->s_want_extra_isize = 3763 le16_to_cpu(es->s_min_extra_isize); 3764 } 3765 } 3766 /* Check if enough inode space is available */ 3767 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 3768 sbi->s_inode_size) { 3769 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3770 EXT4_GOOD_OLD_INODE_SIZE; 3771 ext4_msg(sb, KERN_INFO, "required extra inode space not" 3772 "available"); 3773 } 3774 3775 err = ext4_setup_system_zone(sb); 3776 if (err) { 3777 ext4_msg(sb, KERN_ERR, "failed to initialize system " 3778 "zone (%d)", err); 3779 goto failed_mount4; 3780 } 3781 3782 ext4_ext_init(sb); 3783 err = ext4_mb_init(sb, needs_recovery); 3784 if (err) { 3785 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 3786 err); 3787 goto failed_mount5; 3788 } 3789 3790 err = ext4_register_li_request(sb, first_not_zeroed); 3791 if (err) 3792 goto failed_mount6; 3793 3794 sbi->s_kobj.kset = ext4_kset; 3795 init_completion(&sbi->s_kobj_unregister); 3796 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 3797 "%s", sb->s_id); 3798 if (err) 3799 goto failed_mount7; 3800 3801 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 3802 ext4_orphan_cleanup(sb, es); 3803 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 3804 if (needs_recovery) { 3805 ext4_msg(sb, KERN_INFO, "recovery complete"); 3806 ext4_mark_recovery_complete(sb, es); 3807 } 3808 if (EXT4_SB(sb)->s_journal) { 3809 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 3810 descr = " journalled data mode"; 3811 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 3812 descr = " ordered data mode"; 3813 else 3814 descr = " writeback data mode"; 3815 } else 3816 descr = "out journal"; 3817 3818 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 3819 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 3820 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 3821 3822 if (es->s_error_count) 3823 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 3824 3825 kfree(orig_data); 3826 return 0; 3827 3828 cantfind_ext4: 3829 if (!silent) 3830 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 3831 goto failed_mount; 3832 3833 failed_mount7: 3834 ext4_unregister_li_request(sb); 3835 failed_mount6: 3836 ext4_ext_release(sb); 3837 failed_mount5: 3838 ext4_mb_release(sb); 3839 ext4_release_system_zone(sb); 3840 failed_mount4: 3841 iput(root); 3842 sb->s_root = NULL; 3843 ext4_msg(sb, KERN_ERR, "mount failed"); 3844 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); 3845 failed_mount_wq: 3846 if (sbi->s_journal) { 3847 jbd2_journal_destroy(sbi->s_journal); 3848 sbi->s_journal = NULL; 3849 } 3850 failed_mount3: 3851 del_timer(&sbi->s_err_report); 3852 if (sbi->s_flex_groups) 3853 ext4_kvfree(sbi->s_flex_groups); 3854 percpu_counter_destroy(&sbi->s_freeclusters_counter); 3855 percpu_counter_destroy(&sbi->s_freeinodes_counter); 3856 percpu_counter_destroy(&sbi->s_dirs_counter); 3857 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 3858 if (sbi->s_mmp_tsk) 3859 kthread_stop(sbi->s_mmp_tsk); 3860 failed_mount2: 3861 for (i = 0; i < db_count; i++) 3862 brelse(sbi->s_group_desc[i]); 3863 ext4_kvfree(sbi->s_group_desc); 3864 failed_mount: 3865 if (sbi->s_proc) { 3866 remove_proc_entry(sb->s_id, ext4_proc_root); 3867 } 3868 #ifdef CONFIG_QUOTA 3869 for (i = 0; i < MAXQUOTAS; i++) 3870 kfree(sbi->s_qf_names[i]); 3871 #endif 3872 ext4_blkdev_remove(sbi); 3873 brelse(bh); 3874 out_fail: 3875 sb->s_fs_info = NULL; 3876 kfree(sbi->s_blockgroup_lock); 3877 kfree(sbi); 3878 out_free_orig: 3879 kfree(orig_data); 3880 return ret; 3881 } 3882 3883 /* 3884 * Setup any per-fs journal parameters now. We'll do this both on 3885 * initial mount, once the journal has been initialised but before we've 3886 * done any recovery; and again on any subsequent remount. 3887 */ 3888 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 3889 { 3890 struct ext4_sb_info *sbi = EXT4_SB(sb); 3891 3892 journal->j_commit_interval = sbi->s_commit_interval; 3893 journal->j_min_batch_time = sbi->s_min_batch_time; 3894 journal->j_max_batch_time = sbi->s_max_batch_time; 3895 3896 write_lock(&journal->j_state_lock); 3897 if (test_opt(sb, BARRIER)) 3898 journal->j_flags |= JBD2_BARRIER; 3899 else 3900 journal->j_flags &= ~JBD2_BARRIER; 3901 if (test_opt(sb, DATA_ERR_ABORT)) 3902 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 3903 else 3904 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 3905 write_unlock(&journal->j_state_lock); 3906 } 3907 3908 static journal_t *ext4_get_journal(struct super_block *sb, 3909 unsigned int journal_inum) 3910 { 3911 struct inode *journal_inode; 3912 journal_t *journal; 3913 3914 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 3915 3916 /* First, test for the existence of a valid inode on disk. Bad 3917 * things happen if we iget() an unused inode, as the subsequent 3918 * iput() will try to delete it. */ 3919 3920 journal_inode = ext4_iget(sb, journal_inum); 3921 if (IS_ERR(journal_inode)) { 3922 ext4_msg(sb, KERN_ERR, "no journal found"); 3923 return NULL; 3924 } 3925 if (!journal_inode->i_nlink) { 3926 make_bad_inode(journal_inode); 3927 iput(journal_inode); 3928 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 3929 return NULL; 3930 } 3931 3932 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 3933 journal_inode, journal_inode->i_size); 3934 if (!S_ISREG(journal_inode->i_mode)) { 3935 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 3936 iput(journal_inode); 3937 return NULL; 3938 } 3939 3940 journal = jbd2_journal_init_inode(journal_inode); 3941 if (!journal) { 3942 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 3943 iput(journal_inode); 3944 return NULL; 3945 } 3946 journal->j_private = sb; 3947 ext4_init_journal_params(sb, journal); 3948 return journal; 3949 } 3950 3951 static journal_t *ext4_get_dev_journal(struct super_block *sb, 3952 dev_t j_dev) 3953 { 3954 struct buffer_head *bh; 3955 journal_t *journal; 3956 ext4_fsblk_t start; 3957 ext4_fsblk_t len; 3958 int hblock, blocksize; 3959 ext4_fsblk_t sb_block; 3960 unsigned long offset; 3961 struct ext4_super_block *es; 3962 struct block_device *bdev; 3963 3964 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 3965 3966 bdev = ext4_blkdev_get(j_dev, sb); 3967 if (bdev == NULL) 3968 return NULL; 3969 3970 blocksize = sb->s_blocksize; 3971 hblock = bdev_logical_block_size(bdev); 3972 if (blocksize < hblock) { 3973 ext4_msg(sb, KERN_ERR, 3974 "blocksize too small for journal device"); 3975 goto out_bdev; 3976 } 3977 3978 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 3979 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 3980 set_blocksize(bdev, blocksize); 3981 if (!(bh = __bread(bdev, sb_block, blocksize))) { 3982 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 3983 "external journal"); 3984 goto out_bdev; 3985 } 3986 3987 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); 3988 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 3989 !(le32_to_cpu(es->s_feature_incompat) & 3990 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 3991 ext4_msg(sb, KERN_ERR, "external journal has " 3992 "bad superblock"); 3993 brelse(bh); 3994 goto out_bdev; 3995 } 3996 3997 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 3998 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 3999 brelse(bh); 4000 goto out_bdev; 4001 } 4002 4003 len = ext4_blocks_count(es); 4004 start = sb_block + 1; 4005 brelse(bh); /* we're done with the superblock */ 4006 4007 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4008 start, len, blocksize); 4009 if (!journal) { 4010 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4011 goto out_bdev; 4012 } 4013 journal->j_private = sb; 4014 ll_rw_block(READ, 1, &journal->j_sb_buffer); 4015 wait_on_buffer(journal->j_sb_buffer); 4016 if (!buffer_uptodate(journal->j_sb_buffer)) { 4017 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4018 goto out_journal; 4019 } 4020 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4021 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4022 "user (unsupported) - %d", 4023 be32_to_cpu(journal->j_superblock->s_nr_users)); 4024 goto out_journal; 4025 } 4026 EXT4_SB(sb)->journal_bdev = bdev; 4027 ext4_init_journal_params(sb, journal); 4028 return journal; 4029 4030 out_journal: 4031 jbd2_journal_destroy(journal); 4032 out_bdev: 4033 ext4_blkdev_put(bdev); 4034 return NULL; 4035 } 4036 4037 static int ext4_load_journal(struct super_block *sb, 4038 struct ext4_super_block *es, 4039 unsigned long journal_devnum) 4040 { 4041 journal_t *journal; 4042 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4043 dev_t journal_dev; 4044 int err = 0; 4045 int really_read_only; 4046 4047 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4048 4049 if (journal_devnum && 4050 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4051 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4052 "numbers have changed"); 4053 journal_dev = new_decode_dev(journal_devnum); 4054 } else 4055 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4056 4057 really_read_only = bdev_read_only(sb->s_bdev); 4058 4059 /* 4060 * Are we loading a blank journal or performing recovery after a 4061 * crash? For recovery, we need to check in advance whether we 4062 * can get read-write access to the device. 4063 */ 4064 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 4065 if (sb->s_flags & MS_RDONLY) { 4066 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4067 "required on readonly filesystem"); 4068 if (really_read_only) { 4069 ext4_msg(sb, KERN_ERR, "write access " 4070 "unavailable, cannot proceed"); 4071 return -EROFS; 4072 } 4073 ext4_msg(sb, KERN_INFO, "write access will " 4074 "be enabled during recovery"); 4075 } 4076 } 4077 4078 if (journal_inum && journal_dev) { 4079 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4080 "and inode journals!"); 4081 return -EINVAL; 4082 } 4083 4084 if (journal_inum) { 4085 if (!(journal = ext4_get_journal(sb, journal_inum))) 4086 return -EINVAL; 4087 } else { 4088 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4089 return -EINVAL; 4090 } 4091 4092 if (!(journal->j_flags & JBD2_BARRIER)) 4093 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4094 4095 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { 4096 err = jbd2_journal_update_format(journal); 4097 if (err) { 4098 ext4_msg(sb, KERN_ERR, "error updating journal"); 4099 jbd2_journal_destroy(journal); 4100 return err; 4101 } 4102 } 4103 4104 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) 4105 err = jbd2_journal_wipe(journal, !really_read_only); 4106 if (!err) { 4107 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4108 if (save) 4109 memcpy(save, ((char *) es) + 4110 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4111 err = jbd2_journal_load(journal); 4112 if (save) 4113 memcpy(((char *) es) + EXT4_S_ERR_START, 4114 save, EXT4_S_ERR_LEN); 4115 kfree(save); 4116 } 4117 4118 if (err) { 4119 ext4_msg(sb, KERN_ERR, "error loading journal"); 4120 jbd2_journal_destroy(journal); 4121 return err; 4122 } 4123 4124 EXT4_SB(sb)->s_journal = journal; 4125 ext4_clear_journal_err(sb, es); 4126 4127 if (!really_read_only && journal_devnum && 4128 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4129 es->s_journal_dev = cpu_to_le32(journal_devnum); 4130 4131 /* Make sure we flush the recovery flag to disk. */ 4132 ext4_commit_super(sb, 1); 4133 } 4134 4135 return 0; 4136 } 4137 4138 static int ext4_commit_super(struct super_block *sb, int sync) 4139 { 4140 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4141 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4142 int error = 0; 4143 4144 if (!sbh || block_device_ejected(sb)) 4145 return error; 4146 if (buffer_write_io_error(sbh)) { 4147 /* 4148 * Oh, dear. A previous attempt to write the 4149 * superblock failed. This could happen because the 4150 * USB device was yanked out. Or it could happen to 4151 * be a transient write error and maybe the block will 4152 * be remapped. Nothing we can do but to retry the 4153 * write and hope for the best. 4154 */ 4155 ext4_msg(sb, KERN_ERR, "previous I/O error to " 4156 "superblock detected"); 4157 clear_buffer_write_io_error(sbh); 4158 set_buffer_uptodate(sbh); 4159 } 4160 /* 4161 * If the file system is mounted read-only, don't update the 4162 * superblock write time. This avoids updating the superblock 4163 * write time when we are mounting the root file system 4164 * read/only but we need to replay the journal; at that point, 4165 * for people who are east of GMT and who make their clock 4166 * tick in localtime for Windows bug-for-bug compatibility, 4167 * the clock is set in the future, and this will cause e2fsck 4168 * to complain and force a full file system check. 4169 */ 4170 if (!(sb->s_flags & MS_RDONLY)) 4171 es->s_wtime = cpu_to_le32(get_seconds()); 4172 if (sb->s_bdev->bd_part) 4173 es->s_kbytes_written = 4174 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4175 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4176 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4177 else 4178 es->s_kbytes_written = 4179 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4180 ext4_free_blocks_count_set(es, 4181 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4182 &EXT4_SB(sb)->s_freeclusters_counter))); 4183 es->s_free_inodes_count = 4184 cpu_to_le32(percpu_counter_sum_positive( 4185 &EXT4_SB(sb)->s_freeinodes_counter)); 4186 sb->s_dirt = 0; 4187 BUFFER_TRACE(sbh, "marking dirty"); 4188 mark_buffer_dirty(sbh); 4189 if (sync) { 4190 error = sync_dirty_buffer(sbh); 4191 if (error) 4192 return error; 4193 4194 error = buffer_write_io_error(sbh); 4195 if (error) { 4196 ext4_msg(sb, KERN_ERR, "I/O error while writing " 4197 "superblock"); 4198 clear_buffer_write_io_error(sbh); 4199 set_buffer_uptodate(sbh); 4200 } 4201 } 4202 return error; 4203 } 4204 4205 /* 4206 * Have we just finished recovery? If so, and if we are mounting (or 4207 * remounting) the filesystem readonly, then we will end up with a 4208 * consistent fs on disk. Record that fact. 4209 */ 4210 static void ext4_mark_recovery_complete(struct super_block *sb, 4211 struct ext4_super_block *es) 4212 { 4213 journal_t *journal = EXT4_SB(sb)->s_journal; 4214 4215 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 4216 BUG_ON(journal != NULL); 4217 return; 4218 } 4219 jbd2_journal_lock_updates(journal); 4220 if (jbd2_journal_flush(journal) < 0) 4221 goto out; 4222 4223 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && 4224 sb->s_flags & MS_RDONLY) { 4225 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4226 ext4_commit_super(sb, 1); 4227 } 4228 4229 out: 4230 jbd2_journal_unlock_updates(journal); 4231 } 4232 4233 /* 4234 * If we are mounting (or read-write remounting) a filesystem whose journal 4235 * has recorded an error from a previous lifetime, move that error to the 4236 * main filesystem now. 4237 */ 4238 static void ext4_clear_journal_err(struct super_block *sb, 4239 struct ext4_super_block *es) 4240 { 4241 journal_t *journal; 4242 int j_errno; 4243 const char *errstr; 4244 4245 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 4246 4247 journal = EXT4_SB(sb)->s_journal; 4248 4249 /* 4250 * Now check for any error status which may have been recorded in the 4251 * journal by a prior ext4_error() or ext4_abort() 4252 */ 4253 4254 j_errno = jbd2_journal_errno(journal); 4255 if (j_errno) { 4256 char nbuf[16]; 4257 4258 errstr = ext4_decode_error(sb, j_errno, nbuf); 4259 ext4_warning(sb, "Filesystem error recorded " 4260 "from previous mount: %s", errstr); 4261 ext4_warning(sb, "Marking fs in need of filesystem check."); 4262 4263 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4264 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4265 ext4_commit_super(sb, 1); 4266 4267 jbd2_journal_clear_err(journal); 4268 } 4269 } 4270 4271 /* 4272 * Force the running and committing transactions to commit, 4273 * and wait on the commit. 4274 */ 4275 int ext4_force_commit(struct super_block *sb) 4276 { 4277 journal_t *journal; 4278 int ret = 0; 4279 4280 if (sb->s_flags & MS_RDONLY) 4281 return 0; 4282 4283 journal = EXT4_SB(sb)->s_journal; 4284 if (journal) { 4285 vfs_check_frozen(sb, SB_FREEZE_TRANS); 4286 ret = ext4_journal_force_commit(journal); 4287 } 4288 4289 return ret; 4290 } 4291 4292 static void ext4_write_super(struct super_block *sb) 4293 { 4294 lock_super(sb); 4295 ext4_commit_super(sb, 1); 4296 unlock_super(sb); 4297 } 4298 4299 static int ext4_sync_fs(struct super_block *sb, int wait) 4300 { 4301 int ret = 0; 4302 tid_t target; 4303 struct ext4_sb_info *sbi = EXT4_SB(sb); 4304 4305 trace_ext4_sync_fs(sb, wait); 4306 flush_workqueue(sbi->dio_unwritten_wq); 4307 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4308 if (wait) 4309 jbd2_log_wait_commit(sbi->s_journal, target); 4310 } 4311 return ret; 4312 } 4313 4314 /* 4315 * LVM calls this function before a (read-only) snapshot is created. This 4316 * gives us a chance to flush the journal completely and mark the fs clean. 4317 * 4318 * Note that only this function cannot bring a filesystem to be in a clean 4319 * state independently, because ext4 prevents a new handle from being started 4320 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from 4321 * the upper layer. 4322 */ 4323 static int ext4_freeze(struct super_block *sb) 4324 { 4325 int error = 0; 4326 journal_t *journal; 4327 4328 if (sb->s_flags & MS_RDONLY) 4329 return 0; 4330 4331 journal = EXT4_SB(sb)->s_journal; 4332 4333 /* Now we set up the journal barrier. */ 4334 jbd2_journal_lock_updates(journal); 4335 4336 /* 4337 * Don't clear the needs_recovery flag if we failed to flush 4338 * the journal. 4339 */ 4340 error = jbd2_journal_flush(journal); 4341 if (error < 0) 4342 goto out; 4343 4344 /* Journal blocked and flushed, clear needs_recovery flag. */ 4345 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4346 error = ext4_commit_super(sb, 1); 4347 out: 4348 /* we rely on s_frozen to stop further updates */ 4349 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 4350 return error; 4351 } 4352 4353 /* 4354 * Called by LVM after the snapshot is done. We need to reset the RECOVER 4355 * flag here, even though the filesystem is not technically dirty yet. 4356 */ 4357 static int ext4_unfreeze(struct super_block *sb) 4358 { 4359 if (sb->s_flags & MS_RDONLY) 4360 return 0; 4361 4362 lock_super(sb); 4363 /* Reset the needs_recovery flag before the fs is unlocked. */ 4364 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4365 ext4_commit_super(sb, 1); 4366 unlock_super(sb); 4367 return 0; 4368 } 4369 4370 /* 4371 * Structure to save mount options for ext4_remount's benefit 4372 */ 4373 struct ext4_mount_options { 4374 unsigned long s_mount_opt; 4375 unsigned long s_mount_opt2; 4376 uid_t s_resuid; 4377 gid_t s_resgid; 4378 unsigned long s_commit_interval; 4379 u32 s_min_batch_time, s_max_batch_time; 4380 #ifdef CONFIG_QUOTA 4381 int s_jquota_fmt; 4382 char *s_qf_names[MAXQUOTAS]; 4383 #endif 4384 }; 4385 4386 static int ext4_remount(struct super_block *sb, int *flags, char *data) 4387 { 4388 struct ext4_super_block *es; 4389 struct ext4_sb_info *sbi = EXT4_SB(sb); 4390 ext4_fsblk_t n_blocks_count = 0; 4391 unsigned long old_sb_flags; 4392 struct ext4_mount_options old_opts; 4393 int enable_quota = 0; 4394 ext4_group_t g; 4395 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4396 int err = 0; 4397 #ifdef CONFIG_QUOTA 4398 int i; 4399 #endif 4400 char *orig_data = kstrdup(data, GFP_KERNEL); 4401 4402 /* Store the original options */ 4403 lock_super(sb); 4404 old_sb_flags = sb->s_flags; 4405 old_opts.s_mount_opt = sbi->s_mount_opt; 4406 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4407 old_opts.s_resuid = sbi->s_resuid; 4408 old_opts.s_resgid = sbi->s_resgid; 4409 old_opts.s_commit_interval = sbi->s_commit_interval; 4410 old_opts.s_min_batch_time = sbi->s_min_batch_time; 4411 old_opts.s_max_batch_time = sbi->s_max_batch_time; 4412 #ifdef CONFIG_QUOTA 4413 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4414 for (i = 0; i < MAXQUOTAS; i++) 4415 old_opts.s_qf_names[i] = sbi->s_qf_names[i]; 4416 #endif 4417 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4418 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4419 4420 /* 4421 * Allow the "check" option to be passed as a remount option. 4422 */ 4423 if (!parse_options(data, sb, NULL, &journal_ioprio, 4424 &n_blocks_count, 1)) { 4425 err = -EINVAL; 4426 goto restore_opts; 4427 } 4428 4429 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4430 ext4_abort(sb, "Abort forced by user"); 4431 4432 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4433 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4434 4435 es = sbi->s_es; 4436 4437 if (sbi->s_journal) { 4438 ext4_init_journal_params(sb, sbi->s_journal); 4439 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4440 } 4441 4442 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || 4443 n_blocks_count > ext4_blocks_count(es)) { 4444 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4445 err = -EROFS; 4446 goto restore_opts; 4447 } 4448 4449 if (*flags & MS_RDONLY) { 4450 err = dquot_suspend(sb, -1); 4451 if (err < 0) 4452 goto restore_opts; 4453 4454 /* 4455 * First of all, the unconditional stuff we have to do 4456 * to disable replay of the journal when we next remount 4457 */ 4458 sb->s_flags |= MS_RDONLY; 4459 4460 /* 4461 * OK, test if we are remounting a valid rw partition 4462 * readonly, and if so set the rdonly flag and then 4463 * mark the partition as valid again. 4464 */ 4465 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4466 (sbi->s_mount_state & EXT4_VALID_FS)) 4467 es->s_state = cpu_to_le16(sbi->s_mount_state); 4468 4469 if (sbi->s_journal) 4470 ext4_mark_recovery_complete(sb, es); 4471 } else { 4472 /* Make sure we can mount this feature set readwrite */ 4473 if (!ext4_feature_set_ok(sb, 0)) { 4474 err = -EROFS; 4475 goto restore_opts; 4476 } 4477 /* 4478 * Make sure the group descriptor checksums 4479 * are sane. If they aren't, refuse to remount r/w. 4480 */ 4481 for (g = 0; g < sbi->s_groups_count; g++) { 4482 struct ext4_group_desc *gdp = 4483 ext4_get_group_desc(sb, g, NULL); 4484 4485 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { 4486 ext4_msg(sb, KERN_ERR, 4487 "ext4_remount: Checksum for group %u failed (%u!=%u)", 4488 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), 4489 le16_to_cpu(gdp->bg_checksum)); 4490 err = -EINVAL; 4491 goto restore_opts; 4492 } 4493 } 4494 4495 /* 4496 * If we have an unprocessed orphan list hanging 4497 * around from a previously readonly bdev mount, 4498 * require a full umount/remount for now. 4499 */ 4500 if (es->s_last_orphan) { 4501 ext4_msg(sb, KERN_WARNING, "Couldn't " 4502 "remount RDWR because of unprocessed " 4503 "orphan inode list. Please " 4504 "umount/remount instead"); 4505 err = -EINVAL; 4506 goto restore_opts; 4507 } 4508 4509 /* 4510 * Mounting a RDONLY partition read-write, so reread 4511 * and store the current valid flag. (It may have 4512 * been changed by e2fsck since we originally mounted 4513 * the partition.) 4514 */ 4515 if (sbi->s_journal) 4516 ext4_clear_journal_err(sb, es); 4517 sbi->s_mount_state = le16_to_cpu(es->s_state); 4518 if ((err = ext4_group_extend(sb, es, n_blocks_count))) 4519 goto restore_opts; 4520 if (!ext4_setup_super(sb, es, 0)) 4521 sb->s_flags &= ~MS_RDONLY; 4522 if (EXT4_HAS_INCOMPAT_FEATURE(sb, 4523 EXT4_FEATURE_INCOMPAT_MMP)) 4524 if (ext4_multi_mount_protect(sb, 4525 le64_to_cpu(es->s_mmp_block))) { 4526 err = -EROFS; 4527 goto restore_opts; 4528 } 4529 enable_quota = 1; 4530 } 4531 } 4532 4533 /* 4534 * Reinitialize lazy itable initialization thread based on 4535 * current settings 4536 */ 4537 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4538 ext4_unregister_li_request(sb); 4539 else { 4540 ext4_group_t first_not_zeroed; 4541 first_not_zeroed = ext4_has_uninit_itable(sb); 4542 ext4_register_li_request(sb, first_not_zeroed); 4543 } 4544 4545 ext4_setup_system_zone(sb); 4546 if (sbi->s_journal == NULL) 4547 ext4_commit_super(sb, 1); 4548 4549 #ifdef CONFIG_QUOTA 4550 /* Release old quota file names */ 4551 for (i = 0; i < MAXQUOTAS; i++) 4552 if (old_opts.s_qf_names[i] && 4553 old_opts.s_qf_names[i] != sbi->s_qf_names[i]) 4554 kfree(old_opts.s_qf_names[i]); 4555 #endif 4556 unlock_super(sb); 4557 if (enable_quota) 4558 dquot_resume(sb, -1); 4559 4560 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4561 kfree(orig_data); 4562 return 0; 4563 4564 restore_opts: 4565 sb->s_flags = old_sb_flags; 4566 sbi->s_mount_opt = old_opts.s_mount_opt; 4567 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 4568 sbi->s_resuid = old_opts.s_resuid; 4569 sbi->s_resgid = old_opts.s_resgid; 4570 sbi->s_commit_interval = old_opts.s_commit_interval; 4571 sbi->s_min_batch_time = old_opts.s_min_batch_time; 4572 sbi->s_max_batch_time = old_opts.s_max_batch_time; 4573 #ifdef CONFIG_QUOTA 4574 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 4575 for (i = 0; i < MAXQUOTAS; i++) { 4576 if (sbi->s_qf_names[i] && 4577 old_opts.s_qf_names[i] != sbi->s_qf_names[i]) 4578 kfree(sbi->s_qf_names[i]); 4579 sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 4580 } 4581 #endif 4582 unlock_super(sb); 4583 kfree(orig_data); 4584 return err; 4585 } 4586 4587 /* 4588 * Note: calculating the overhead so we can be compatible with 4589 * historical BSD practice is quite difficult in the face of 4590 * clusters/bigalloc. This is because multiple metadata blocks from 4591 * different block group can end up in the same allocation cluster. 4592 * Calculating the exact overhead in the face of clustered allocation 4593 * requires either O(all block bitmaps) in memory or O(number of block 4594 * groups**2) in time. We will still calculate the superblock for 4595 * older file systems --- and if we come across with a bigalloc file 4596 * system with zero in s_overhead_clusters the estimate will be close to 4597 * correct especially for very large cluster sizes --- but for newer 4598 * file systems, it's better to calculate this figure once at mkfs 4599 * time, and store it in the superblock. If the superblock value is 4600 * present (even for non-bigalloc file systems), we will use it. 4601 */ 4602 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 4603 { 4604 struct super_block *sb = dentry->d_sb; 4605 struct ext4_sb_info *sbi = EXT4_SB(sb); 4606 struct ext4_super_block *es = sbi->s_es; 4607 struct ext4_group_desc *gdp; 4608 u64 fsid; 4609 s64 bfree; 4610 4611 if (test_opt(sb, MINIX_DF)) { 4612 sbi->s_overhead_last = 0; 4613 } else if (es->s_overhead_clusters) { 4614 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters); 4615 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { 4616 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4617 ext4_fsblk_t overhead = 0; 4618 4619 /* 4620 * Compute the overhead (FS structures). This is constant 4621 * for a given filesystem unless the number of block groups 4622 * changes so we cache the previous value until it does. 4623 */ 4624 4625 /* 4626 * All of the blocks before first_data_block are 4627 * overhead 4628 */ 4629 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 4630 4631 /* 4632 * Add the overhead found in each block group 4633 */ 4634 for (i = 0; i < ngroups; i++) { 4635 gdp = ext4_get_group_desc(sb, i, NULL); 4636 overhead += ext4_num_overhead_clusters(sb, i, gdp); 4637 cond_resched(); 4638 } 4639 sbi->s_overhead_last = overhead; 4640 smp_wmb(); 4641 sbi->s_blocks_last = ext4_blocks_count(es); 4642 } 4643 4644 buf->f_type = EXT4_SUPER_MAGIC; 4645 buf->f_bsize = sb->s_blocksize; 4646 buf->f_blocks = (ext4_blocks_count(es) - 4647 EXT4_C2B(sbi, sbi->s_overhead_last)); 4648 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 4649 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 4650 /* prevent underflow in case that few free space is available */ 4651 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 4652 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); 4653 if (buf->f_bfree < ext4_r_blocks_count(es)) 4654 buf->f_bavail = 0; 4655 buf->f_files = le32_to_cpu(es->s_inodes_count); 4656 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 4657 buf->f_namelen = EXT4_NAME_LEN; 4658 fsid = le64_to_cpup((void *)es->s_uuid) ^ 4659 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 4660 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 4661 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 4662 4663 return 0; 4664 } 4665 4666 /* Helper function for writing quotas on sync - we need to start transaction 4667 * before quota file is locked for write. Otherwise the are possible deadlocks: 4668 * Process 1 Process 2 4669 * ext4_create() quota_sync() 4670 * jbd2_journal_start() write_dquot() 4671 * dquot_initialize() down(dqio_mutex) 4672 * down(dqio_mutex) jbd2_journal_start() 4673 * 4674 */ 4675 4676 #ifdef CONFIG_QUOTA 4677 4678 static inline struct inode *dquot_to_inode(struct dquot *dquot) 4679 { 4680 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; 4681 } 4682 4683 static int ext4_write_dquot(struct dquot *dquot) 4684 { 4685 int ret, err; 4686 handle_t *handle; 4687 struct inode *inode; 4688 4689 inode = dquot_to_inode(dquot); 4690 handle = ext4_journal_start(inode, 4691 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 4692 if (IS_ERR(handle)) 4693 return PTR_ERR(handle); 4694 ret = dquot_commit(dquot); 4695 err = ext4_journal_stop(handle); 4696 if (!ret) 4697 ret = err; 4698 return ret; 4699 } 4700 4701 static int ext4_acquire_dquot(struct dquot *dquot) 4702 { 4703 int ret, err; 4704 handle_t *handle; 4705 4706 handle = ext4_journal_start(dquot_to_inode(dquot), 4707 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 4708 if (IS_ERR(handle)) 4709 return PTR_ERR(handle); 4710 ret = dquot_acquire(dquot); 4711 err = ext4_journal_stop(handle); 4712 if (!ret) 4713 ret = err; 4714 return ret; 4715 } 4716 4717 static int ext4_release_dquot(struct dquot *dquot) 4718 { 4719 int ret, err; 4720 handle_t *handle; 4721 4722 handle = ext4_journal_start(dquot_to_inode(dquot), 4723 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 4724 if (IS_ERR(handle)) { 4725 /* Release dquot anyway to avoid endless cycle in dqput() */ 4726 dquot_release(dquot); 4727 return PTR_ERR(handle); 4728 } 4729 ret = dquot_release(dquot); 4730 err = ext4_journal_stop(handle); 4731 if (!ret) 4732 ret = err; 4733 return ret; 4734 } 4735 4736 static int ext4_mark_dquot_dirty(struct dquot *dquot) 4737 { 4738 /* Are we journaling quotas? */ 4739 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || 4740 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { 4741 dquot_mark_dquot_dirty(dquot); 4742 return ext4_write_dquot(dquot); 4743 } else { 4744 return dquot_mark_dquot_dirty(dquot); 4745 } 4746 } 4747 4748 static int ext4_write_info(struct super_block *sb, int type) 4749 { 4750 int ret, err; 4751 handle_t *handle; 4752 4753 /* Data block + inode block */ 4754 handle = ext4_journal_start(sb->s_root->d_inode, 2); 4755 if (IS_ERR(handle)) 4756 return PTR_ERR(handle); 4757 ret = dquot_commit_info(sb, type); 4758 err = ext4_journal_stop(handle); 4759 if (!ret) 4760 ret = err; 4761 return ret; 4762 } 4763 4764 /* 4765 * Turn on quotas during mount time - we need to find 4766 * the quota file and such... 4767 */ 4768 static int ext4_quota_on_mount(struct super_block *sb, int type) 4769 { 4770 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 4771 EXT4_SB(sb)->s_jquota_fmt, type); 4772 } 4773 4774 /* 4775 * Standard function to be called on quota_on 4776 */ 4777 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 4778 struct path *path) 4779 { 4780 int err; 4781 4782 if (!test_opt(sb, QUOTA)) 4783 return -EINVAL; 4784 4785 /* Quotafile not on the same filesystem? */ 4786 if (path->mnt->mnt_sb != sb) 4787 return -EXDEV; 4788 /* Journaling quota? */ 4789 if (EXT4_SB(sb)->s_qf_names[type]) { 4790 /* Quotafile not in fs root? */ 4791 if (path->dentry->d_parent != sb->s_root) 4792 ext4_msg(sb, KERN_WARNING, 4793 "Quota file not on filesystem root. " 4794 "Journaled quota will not work"); 4795 } 4796 4797 /* 4798 * When we journal data on quota file, we have to flush journal to see 4799 * all updates to the file when we bypass pagecache... 4800 */ 4801 if (EXT4_SB(sb)->s_journal && 4802 ext4_should_journal_data(path->dentry->d_inode)) { 4803 /* 4804 * We don't need to lock updates but journal_flush() could 4805 * otherwise be livelocked... 4806 */ 4807 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 4808 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 4809 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 4810 if (err) 4811 return err; 4812 } 4813 4814 return dquot_quota_on(sb, type, format_id, path); 4815 } 4816 4817 static int ext4_quota_off(struct super_block *sb, int type) 4818 { 4819 struct inode *inode = sb_dqopt(sb)->files[type]; 4820 handle_t *handle; 4821 4822 /* Force all delayed allocation blocks to be allocated. 4823 * Caller already holds s_umount sem */ 4824 if (test_opt(sb, DELALLOC)) 4825 sync_filesystem(sb); 4826 4827 if (!inode) 4828 goto out; 4829 4830 /* Update modification times of quota files when userspace can 4831 * start looking at them */ 4832 handle = ext4_journal_start(inode, 1); 4833 if (IS_ERR(handle)) 4834 goto out; 4835 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 4836 ext4_mark_inode_dirty(handle, inode); 4837 ext4_journal_stop(handle); 4838 4839 out: 4840 return dquot_quota_off(sb, type); 4841 } 4842 4843 /* Read data from quotafile - avoid pagecache and such because we cannot afford 4844 * acquiring the locks... As quota files are never truncated and quota code 4845 * itself serializes the operations (and no one else should touch the files) 4846 * we don't have to be afraid of races */ 4847 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 4848 size_t len, loff_t off) 4849 { 4850 struct inode *inode = sb_dqopt(sb)->files[type]; 4851 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 4852 int err = 0; 4853 int offset = off & (sb->s_blocksize - 1); 4854 int tocopy; 4855 size_t toread; 4856 struct buffer_head *bh; 4857 loff_t i_size = i_size_read(inode); 4858 4859 if (off > i_size) 4860 return 0; 4861 if (off+len > i_size) 4862 len = i_size-off; 4863 toread = len; 4864 while (toread > 0) { 4865 tocopy = sb->s_blocksize - offset < toread ? 4866 sb->s_blocksize - offset : toread; 4867 bh = ext4_bread(NULL, inode, blk, 0, &err); 4868 if (err) 4869 return err; 4870 if (!bh) /* A hole? */ 4871 memset(data, 0, tocopy); 4872 else 4873 memcpy(data, bh->b_data+offset, tocopy); 4874 brelse(bh); 4875 offset = 0; 4876 toread -= tocopy; 4877 data += tocopy; 4878 blk++; 4879 } 4880 return len; 4881 } 4882 4883 /* Write to quotafile (we know the transaction is already started and has 4884 * enough credits) */ 4885 static ssize_t ext4_quota_write(struct super_block *sb, int type, 4886 const char *data, size_t len, loff_t off) 4887 { 4888 struct inode *inode = sb_dqopt(sb)->files[type]; 4889 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 4890 int err = 0; 4891 int offset = off & (sb->s_blocksize - 1); 4892 struct buffer_head *bh; 4893 handle_t *handle = journal_current_handle(); 4894 4895 if (EXT4_SB(sb)->s_journal && !handle) { 4896 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 4897 " cancelled because transaction is not started", 4898 (unsigned long long)off, (unsigned long long)len); 4899 return -EIO; 4900 } 4901 /* 4902 * Since we account only one data block in transaction credits, 4903 * then it is impossible to cross a block boundary. 4904 */ 4905 if (sb->s_blocksize - offset < len) { 4906 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 4907 " cancelled because not block aligned", 4908 (unsigned long long)off, (unsigned long long)len); 4909 return -EIO; 4910 } 4911 4912 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); 4913 bh = ext4_bread(handle, inode, blk, 1, &err); 4914 if (!bh) 4915 goto out; 4916 err = ext4_journal_get_write_access(handle, bh); 4917 if (err) { 4918 brelse(bh); 4919 goto out; 4920 } 4921 lock_buffer(bh); 4922 memcpy(bh->b_data+offset, data, len); 4923 flush_dcache_page(bh->b_page); 4924 unlock_buffer(bh); 4925 err = ext4_handle_dirty_metadata(handle, NULL, bh); 4926 brelse(bh); 4927 out: 4928 if (err) { 4929 mutex_unlock(&inode->i_mutex); 4930 return err; 4931 } 4932 if (inode->i_size < off + len) { 4933 i_size_write(inode, off + len); 4934 EXT4_I(inode)->i_disksize = inode->i_size; 4935 ext4_mark_inode_dirty(handle, inode); 4936 } 4937 mutex_unlock(&inode->i_mutex); 4938 return len; 4939 } 4940 4941 #endif 4942 4943 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 4944 const char *dev_name, void *data) 4945 { 4946 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 4947 } 4948 4949 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 4950 static inline void register_as_ext2(void) 4951 { 4952 int err = register_filesystem(&ext2_fs_type); 4953 if (err) 4954 printk(KERN_WARNING 4955 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 4956 } 4957 4958 static inline void unregister_as_ext2(void) 4959 { 4960 unregister_filesystem(&ext2_fs_type); 4961 } 4962 4963 static inline int ext2_feature_set_ok(struct super_block *sb) 4964 { 4965 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) 4966 return 0; 4967 if (sb->s_flags & MS_RDONLY) 4968 return 1; 4969 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) 4970 return 0; 4971 return 1; 4972 } 4973 MODULE_ALIAS("ext2"); 4974 #else 4975 static inline void register_as_ext2(void) { } 4976 static inline void unregister_as_ext2(void) { } 4977 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 4978 #endif 4979 4980 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 4981 static inline void register_as_ext3(void) 4982 { 4983 int err = register_filesystem(&ext3_fs_type); 4984 if (err) 4985 printk(KERN_WARNING 4986 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 4987 } 4988 4989 static inline void unregister_as_ext3(void) 4990 { 4991 unregister_filesystem(&ext3_fs_type); 4992 } 4993 4994 static inline int ext3_feature_set_ok(struct super_block *sb) 4995 { 4996 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) 4997 return 0; 4998 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 4999 return 0; 5000 if (sb->s_flags & MS_RDONLY) 5001 return 1; 5002 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) 5003 return 0; 5004 return 1; 5005 } 5006 MODULE_ALIAS("ext3"); 5007 #else 5008 static inline void register_as_ext3(void) { } 5009 static inline void unregister_as_ext3(void) { } 5010 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } 5011 #endif 5012 5013 static struct file_system_type ext4_fs_type = { 5014 .owner = THIS_MODULE, 5015 .name = "ext4", 5016 .mount = ext4_mount, 5017 .kill_sb = kill_block_super, 5018 .fs_flags = FS_REQUIRES_DEV, 5019 }; 5020 5021 static int __init ext4_init_feat_adverts(void) 5022 { 5023 struct ext4_features *ef; 5024 int ret = -ENOMEM; 5025 5026 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); 5027 if (!ef) 5028 goto out; 5029 5030 ef->f_kobj.kset = ext4_kset; 5031 init_completion(&ef->f_kobj_unregister); 5032 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, 5033 "features"); 5034 if (ret) { 5035 kfree(ef); 5036 goto out; 5037 } 5038 5039 ext4_feat = ef; 5040 ret = 0; 5041 out: 5042 return ret; 5043 } 5044 5045 static void ext4_exit_feat_adverts(void) 5046 { 5047 kobject_put(&ext4_feat->f_kobj); 5048 wait_for_completion(&ext4_feat->f_kobj_unregister); 5049 kfree(ext4_feat); 5050 } 5051 5052 /* Shared across all ext4 file systems */ 5053 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5054 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 5055 5056 static int __init ext4_init_fs(void) 5057 { 5058 int i, err; 5059 5060 ext4_check_flag_values(); 5061 5062 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { 5063 mutex_init(&ext4__aio_mutex[i]); 5064 init_waitqueue_head(&ext4__ioend_wq[i]); 5065 } 5066 5067 err = ext4_init_pageio(); 5068 if (err) 5069 return err; 5070 err = ext4_init_system_zone(); 5071 if (err) 5072 goto out6; 5073 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 5074 if (!ext4_kset) 5075 goto out5; 5076 ext4_proc_root = proc_mkdir("fs/ext4", NULL); 5077 5078 err = ext4_init_feat_adverts(); 5079 if (err) 5080 goto out4; 5081 5082 err = ext4_init_mballoc(); 5083 if (err) 5084 goto out3; 5085 5086 err = ext4_init_xattr(); 5087 if (err) 5088 goto out2; 5089 err = init_inodecache(); 5090 if (err) 5091 goto out1; 5092 register_as_ext3(); 5093 register_as_ext2(); 5094 err = register_filesystem(&ext4_fs_type); 5095 if (err) 5096 goto out; 5097 5098 ext4_li_info = NULL; 5099 mutex_init(&ext4_li_mtx); 5100 return 0; 5101 out: 5102 unregister_as_ext2(); 5103 unregister_as_ext3(); 5104 destroy_inodecache(); 5105 out1: 5106 ext4_exit_xattr(); 5107 out2: 5108 ext4_exit_mballoc(); 5109 out3: 5110 ext4_exit_feat_adverts(); 5111 out4: 5112 if (ext4_proc_root) 5113 remove_proc_entry("fs/ext4", NULL); 5114 kset_unregister(ext4_kset); 5115 out5: 5116 ext4_exit_system_zone(); 5117 out6: 5118 ext4_exit_pageio(); 5119 return err; 5120 } 5121 5122 static void __exit ext4_exit_fs(void) 5123 { 5124 ext4_destroy_lazyinit_thread(); 5125 unregister_as_ext2(); 5126 unregister_as_ext3(); 5127 unregister_filesystem(&ext4_fs_type); 5128 destroy_inodecache(); 5129 ext4_exit_xattr(); 5130 ext4_exit_mballoc(); 5131 ext4_exit_feat_adverts(); 5132 remove_proc_entry("fs/ext4", NULL); 5133 kset_unregister(ext4_kset); 5134 ext4_exit_system_zone(); 5135 ext4_exit_pageio(); 5136 } 5137 5138 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 5139 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5140 MODULE_LICENSE("GPL"); 5141 module_init(ext4_init_fs) 5142 module_exit(ext4_exit_fs) 5143