1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * linux/fs/ext4/super.c 4 * 5 * Copyright (C) 1992, 1993, 1994, 1995 6 * Remy Card (card@masi.ibp.fr) 7 * Laboratoire MASI - Institut Blaise Pascal 8 * Universite Pierre et Marie Curie (Paris VI) 9 * 10 * from 11 * 12 * linux/fs/minix/inode.c 13 * 14 * Copyright (C) 1991, 1992 Linus Torvalds 15 * 16 * Big-endian to little-endian byte-swapping/bitmaps by 17 * David S. Miller (davem@caip.rutgers.edu), 1995 18 */ 19 20 #include <linux/module.h> 21 #include <linux/string.h> 22 #include <linux/fs.h> 23 #include <linux/time.h> 24 #include <linux/vmalloc.h> 25 #include <linux/slab.h> 26 #include <linux/init.h> 27 #include <linux/blkdev.h> 28 #include <linux/backing-dev.h> 29 #include <linux/parser.h> 30 #include <linux/buffer_head.h> 31 #include <linux/exportfs.h> 32 #include <linux/vfs.h> 33 #include <linux/random.h> 34 #include <linux/mount.h> 35 #include <linux/namei.h> 36 #include <linux/quotaops.h> 37 #include <linux/seq_file.h> 38 #include <linux/ctype.h> 39 #include <linux/log2.h> 40 #include <linux/crc16.h> 41 #include <linux/dax.h> 42 #include <linux/cleancache.h> 43 #include <linux/uaccess.h> 44 #include <linux/iversion.h> 45 #include <linux/unicode.h> 46 47 #include <linux/kthread.h> 48 #include <linux/freezer.h> 49 50 #include "ext4.h" 51 #include "ext4_extents.h" /* Needed for trace points definition */ 52 #include "ext4_jbd2.h" 53 #include "xattr.h" 54 #include "acl.h" 55 #include "mballoc.h" 56 #include "fsmap.h" 57 58 #define CREATE_TRACE_POINTS 59 #include <trace/events/ext4.h> 60 61 static struct ext4_lazy_init *ext4_li_info; 62 static struct mutex ext4_li_mtx; 63 static struct ratelimit_state ext4_mount_msg_ratelimit; 64 65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 66 unsigned long journal_devnum); 67 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 68 static int ext4_commit_super(struct super_block *sb, int sync); 69 static void ext4_mark_recovery_complete(struct super_block *sb, 70 struct ext4_super_block *es); 71 static void ext4_clear_journal_err(struct super_block *sb, 72 struct ext4_super_block *es); 73 static int ext4_sync_fs(struct super_block *sb, int wait); 74 static int ext4_remount(struct super_block *sb, int *flags, char *data); 75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76 static int ext4_unfreeze(struct super_block *sb); 77 static int ext4_freeze(struct super_block *sb); 78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79 const char *dev_name, void *data); 80 static inline int ext2_feature_set_ok(struct super_block *sb); 81 static inline int ext3_feature_set_ok(struct super_block *sb); 82 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83 static void ext4_destroy_lazyinit_thread(void); 84 static void ext4_unregister_li_request(struct super_block *sb); 85 static void ext4_clear_request_list(void); 86 static struct inode *ext4_get_journal_inode(struct super_block *sb, 87 unsigned int journal_inum); 88 89 /* 90 * Lock ordering 91 * 92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 93 * i_mmap_rwsem (inode->i_mmap_rwsem)! 94 * 95 * page fault path: 96 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 97 * page lock -> i_data_sem (rw) 98 * 99 * buffered write path: 100 * sb_start_write -> i_mutex -> mmap_sem 101 * sb_start_write -> i_mutex -> transaction start -> page lock -> 102 * i_data_sem (rw) 103 * 104 * truncate: 105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 107 * i_data_sem (rw) 108 * 109 * direct IO: 110 * sb_start_write -> i_mutex -> mmap_sem 111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 112 * 113 * writepages: 114 * transaction start -> page lock(s) -> i_data_sem (rw) 115 */ 116 117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 118 static struct file_system_type ext2_fs_type = { 119 .owner = THIS_MODULE, 120 .name = "ext2", 121 .mount = ext4_mount, 122 .kill_sb = kill_block_super, 123 .fs_flags = FS_REQUIRES_DEV, 124 }; 125 MODULE_ALIAS_FS("ext2"); 126 MODULE_ALIAS("ext2"); 127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 128 #else 129 #define IS_EXT2_SB(sb) (0) 130 #endif 131 132 133 static struct file_system_type ext3_fs_type = { 134 .owner = THIS_MODULE, 135 .name = "ext3", 136 .mount = ext4_mount, 137 .kill_sb = kill_block_super, 138 .fs_flags = FS_REQUIRES_DEV, 139 }; 140 MODULE_ALIAS_FS("ext3"); 141 MODULE_ALIAS("ext3"); 142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 143 144 /* 145 * This works like sb_bread() except it uses ERR_PTR for error 146 * returns. Currently with sb_bread it's impossible to distinguish 147 * between ENOMEM and EIO situations (since both result in a NULL 148 * return. 149 */ 150 struct buffer_head * 151 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags) 152 { 153 struct buffer_head *bh = sb_getblk(sb, block); 154 155 if (bh == NULL) 156 return ERR_PTR(-ENOMEM); 157 if (buffer_uptodate(bh)) 158 return bh; 159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh); 160 wait_on_buffer(bh); 161 if (buffer_uptodate(bh)) 162 return bh; 163 put_bh(bh); 164 return ERR_PTR(-EIO); 165 } 166 167 static int ext4_verify_csum_type(struct super_block *sb, 168 struct ext4_super_block *es) 169 { 170 if (!ext4_has_feature_metadata_csum(sb)) 171 return 1; 172 173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 174 } 175 176 static __le32 ext4_superblock_csum(struct super_block *sb, 177 struct ext4_super_block *es) 178 { 179 struct ext4_sb_info *sbi = EXT4_SB(sb); 180 int offset = offsetof(struct ext4_super_block, s_checksum); 181 __u32 csum; 182 183 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 184 185 return cpu_to_le32(csum); 186 } 187 188 static int ext4_superblock_csum_verify(struct super_block *sb, 189 struct ext4_super_block *es) 190 { 191 if (!ext4_has_metadata_csum(sb)) 192 return 1; 193 194 return es->s_checksum == ext4_superblock_csum(sb, es); 195 } 196 197 void ext4_superblock_csum_set(struct super_block *sb) 198 { 199 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 200 201 if (!ext4_has_metadata_csum(sb)) 202 return; 203 204 es->s_checksum = ext4_superblock_csum(sb, es); 205 } 206 207 void *ext4_kvmalloc(size_t size, gfp_t flags) 208 { 209 void *ret; 210 211 ret = kmalloc(size, flags | __GFP_NOWARN); 212 if (!ret) 213 ret = __vmalloc(size, flags, PAGE_KERNEL); 214 return ret; 215 } 216 217 void *ext4_kvzalloc(size_t size, gfp_t flags) 218 { 219 void *ret; 220 221 ret = kzalloc(size, flags | __GFP_NOWARN); 222 if (!ret) 223 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 224 return ret; 225 } 226 227 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 228 struct ext4_group_desc *bg) 229 { 230 return le32_to_cpu(bg->bg_block_bitmap_lo) | 231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 232 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 233 } 234 235 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 236 struct ext4_group_desc *bg) 237 { 238 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 240 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 241 } 242 243 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 244 struct ext4_group_desc *bg) 245 { 246 return le32_to_cpu(bg->bg_inode_table_lo) | 247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 248 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 249 } 250 251 __u32 ext4_free_group_clusters(struct super_block *sb, 252 struct ext4_group_desc *bg) 253 { 254 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 256 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 257 } 258 259 __u32 ext4_free_inodes_count(struct super_block *sb, 260 struct ext4_group_desc *bg) 261 { 262 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 263 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 264 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 265 } 266 267 __u32 ext4_used_dirs_count(struct super_block *sb, 268 struct ext4_group_desc *bg) 269 { 270 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 271 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 272 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 273 } 274 275 __u32 ext4_itable_unused_count(struct super_block *sb, 276 struct ext4_group_desc *bg) 277 { 278 return le16_to_cpu(bg->bg_itable_unused_lo) | 279 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 280 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 281 } 282 283 void ext4_block_bitmap_set(struct super_block *sb, 284 struct ext4_group_desc *bg, ext4_fsblk_t blk) 285 { 286 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 288 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 289 } 290 291 void ext4_inode_bitmap_set(struct super_block *sb, 292 struct ext4_group_desc *bg, ext4_fsblk_t blk) 293 { 294 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 296 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 297 } 298 299 void ext4_inode_table_set(struct super_block *sb, 300 struct ext4_group_desc *bg, ext4_fsblk_t blk) 301 { 302 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 304 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 305 } 306 307 void ext4_free_group_clusters_set(struct super_block *sb, 308 struct ext4_group_desc *bg, __u32 count) 309 { 310 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 312 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 313 } 314 315 void ext4_free_inodes_set(struct super_block *sb, 316 struct ext4_group_desc *bg, __u32 count) 317 { 318 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 319 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 320 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 321 } 322 323 void ext4_used_dirs_set(struct super_block *sb, 324 struct ext4_group_desc *bg, __u32 count) 325 { 326 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 327 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 328 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 329 } 330 331 void ext4_itable_unused_set(struct super_block *sb, 332 struct ext4_group_desc *bg, __u32 count) 333 { 334 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 335 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 336 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 337 } 338 339 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi) 340 { 341 time64_t now = ktime_get_real_seconds(); 342 343 now = clamp_val(now, 0, (1ull << 40) - 1); 344 345 *lo = cpu_to_le32(lower_32_bits(now)); 346 *hi = upper_32_bits(now); 347 } 348 349 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 350 { 351 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 352 } 353 #define ext4_update_tstamp(es, tstamp) \ 354 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 355 #define ext4_get_tstamp(es, tstamp) \ 356 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 357 358 static void __save_error_info(struct super_block *sb, const char *func, 359 unsigned int line) 360 { 361 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 362 363 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 364 if (bdev_read_only(sb->s_bdev)) 365 return; 366 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 367 ext4_update_tstamp(es, s_last_error_time); 368 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 369 es->s_last_error_line = cpu_to_le32(line); 370 if (!es->s_first_error_time) { 371 es->s_first_error_time = es->s_last_error_time; 372 es->s_first_error_time_hi = es->s_last_error_time_hi; 373 strncpy(es->s_first_error_func, func, 374 sizeof(es->s_first_error_func)); 375 es->s_first_error_line = cpu_to_le32(line); 376 es->s_first_error_ino = es->s_last_error_ino; 377 es->s_first_error_block = es->s_last_error_block; 378 } 379 /* 380 * Start the daily error reporting function if it hasn't been 381 * started already 382 */ 383 if (!es->s_error_count) 384 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 385 le32_add_cpu(&es->s_error_count, 1); 386 } 387 388 static void save_error_info(struct super_block *sb, const char *func, 389 unsigned int line) 390 { 391 __save_error_info(sb, func, line); 392 ext4_commit_super(sb, 1); 393 } 394 395 /* 396 * The del_gendisk() function uninitializes the disk-specific data 397 * structures, including the bdi structure, without telling anyone 398 * else. Once this happens, any attempt to call mark_buffer_dirty() 399 * (for example, by ext4_commit_super), will cause a kernel OOPS. 400 * This is a kludge to prevent these oops until we can put in a proper 401 * hook in del_gendisk() to inform the VFS and file system layers. 402 */ 403 static int block_device_ejected(struct super_block *sb) 404 { 405 struct inode *bd_inode = sb->s_bdev->bd_inode; 406 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 407 408 return bdi->dev == NULL; 409 } 410 411 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 412 { 413 struct super_block *sb = journal->j_private; 414 struct ext4_sb_info *sbi = EXT4_SB(sb); 415 int error = is_journal_aborted(journal); 416 struct ext4_journal_cb_entry *jce; 417 418 BUG_ON(txn->t_state == T_FINISHED); 419 420 ext4_process_freed_data(sb, txn->t_tid); 421 422 spin_lock(&sbi->s_md_lock); 423 while (!list_empty(&txn->t_private_list)) { 424 jce = list_entry(txn->t_private_list.next, 425 struct ext4_journal_cb_entry, jce_list); 426 list_del_init(&jce->jce_list); 427 spin_unlock(&sbi->s_md_lock); 428 jce->jce_func(sb, jce, error); 429 spin_lock(&sbi->s_md_lock); 430 } 431 spin_unlock(&sbi->s_md_lock); 432 } 433 434 static bool system_going_down(void) 435 { 436 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 437 || system_state == SYSTEM_RESTART; 438 } 439 440 /* Deal with the reporting of failure conditions on a filesystem such as 441 * inconsistencies detected or read IO failures. 442 * 443 * On ext2, we can store the error state of the filesystem in the 444 * superblock. That is not possible on ext4, because we may have other 445 * write ordering constraints on the superblock which prevent us from 446 * writing it out straight away; and given that the journal is about to 447 * be aborted, we can't rely on the current, or future, transactions to 448 * write out the superblock safely. 449 * 450 * We'll just use the jbd2_journal_abort() error code to record an error in 451 * the journal instead. On recovery, the journal will complain about 452 * that error until we've noted it down and cleared it. 453 */ 454 455 static void ext4_handle_error(struct super_block *sb) 456 { 457 if (test_opt(sb, WARN_ON_ERROR)) 458 WARN_ON_ONCE(1); 459 460 if (sb_rdonly(sb)) 461 return; 462 463 if (!test_opt(sb, ERRORS_CONT)) { 464 journal_t *journal = EXT4_SB(sb)->s_journal; 465 466 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 467 if (journal) 468 jbd2_journal_abort(journal, -EIO); 469 } 470 /* 471 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 472 * could panic during 'reboot -f' as the underlying device got already 473 * disabled. 474 */ 475 if (test_opt(sb, ERRORS_RO) || system_going_down()) { 476 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 477 /* 478 * Make sure updated value of ->s_mount_flags will be visible 479 * before ->s_flags update 480 */ 481 smp_wmb(); 482 sb->s_flags |= SB_RDONLY; 483 } else if (test_opt(sb, ERRORS_PANIC)) { 484 if (EXT4_SB(sb)->s_journal && 485 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 486 return; 487 panic("EXT4-fs (device %s): panic forced after error\n", 488 sb->s_id); 489 } 490 } 491 492 #define ext4_error_ratelimit(sb) \ 493 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 494 "EXT4-fs error") 495 496 void __ext4_error(struct super_block *sb, const char *function, 497 unsigned int line, const char *fmt, ...) 498 { 499 struct va_format vaf; 500 va_list args; 501 502 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 503 return; 504 505 trace_ext4_error(sb, function, line); 506 if (ext4_error_ratelimit(sb)) { 507 va_start(args, fmt); 508 vaf.fmt = fmt; 509 vaf.va = &args; 510 printk(KERN_CRIT 511 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 512 sb->s_id, function, line, current->comm, &vaf); 513 va_end(args); 514 } 515 save_error_info(sb, function, line); 516 ext4_handle_error(sb); 517 } 518 519 void __ext4_error_inode(struct inode *inode, const char *function, 520 unsigned int line, ext4_fsblk_t block, 521 const char *fmt, ...) 522 { 523 va_list args; 524 struct va_format vaf; 525 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 526 527 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 528 return; 529 530 trace_ext4_error(inode->i_sb, function, line); 531 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 532 es->s_last_error_block = cpu_to_le64(block); 533 if (ext4_error_ratelimit(inode->i_sb)) { 534 va_start(args, fmt); 535 vaf.fmt = fmt; 536 vaf.va = &args; 537 if (block) 538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 539 "inode #%lu: block %llu: comm %s: %pV\n", 540 inode->i_sb->s_id, function, line, inode->i_ino, 541 block, current->comm, &vaf); 542 else 543 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 544 "inode #%lu: comm %s: %pV\n", 545 inode->i_sb->s_id, function, line, inode->i_ino, 546 current->comm, &vaf); 547 va_end(args); 548 } 549 save_error_info(inode->i_sb, function, line); 550 ext4_handle_error(inode->i_sb); 551 } 552 553 void __ext4_error_file(struct file *file, const char *function, 554 unsigned int line, ext4_fsblk_t block, 555 const char *fmt, ...) 556 { 557 va_list args; 558 struct va_format vaf; 559 struct ext4_super_block *es; 560 struct inode *inode = file_inode(file); 561 char pathname[80], *path; 562 563 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 564 return; 565 566 trace_ext4_error(inode->i_sb, function, line); 567 es = EXT4_SB(inode->i_sb)->s_es; 568 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 569 if (ext4_error_ratelimit(inode->i_sb)) { 570 path = file_path(file, pathname, sizeof(pathname)); 571 if (IS_ERR(path)) 572 path = "(unknown)"; 573 va_start(args, fmt); 574 vaf.fmt = fmt; 575 vaf.va = &args; 576 if (block) 577 printk(KERN_CRIT 578 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 579 "block %llu: comm %s: path %s: %pV\n", 580 inode->i_sb->s_id, function, line, inode->i_ino, 581 block, current->comm, path, &vaf); 582 else 583 printk(KERN_CRIT 584 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 585 "comm %s: path %s: %pV\n", 586 inode->i_sb->s_id, function, line, inode->i_ino, 587 current->comm, path, &vaf); 588 va_end(args); 589 } 590 save_error_info(inode->i_sb, function, line); 591 ext4_handle_error(inode->i_sb); 592 } 593 594 const char *ext4_decode_error(struct super_block *sb, int errno, 595 char nbuf[16]) 596 { 597 char *errstr = NULL; 598 599 switch (errno) { 600 case -EFSCORRUPTED: 601 errstr = "Corrupt filesystem"; 602 break; 603 case -EFSBADCRC: 604 errstr = "Filesystem failed CRC"; 605 break; 606 case -EIO: 607 errstr = "IO failure"; 608 break; 609 case -ENOMEM: 610 errstr = "Out of memory"; 611 break; 612 case -EROFS: 613 if (!sb || (EXT4_SB(sb)->s_journal && 614 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 615 errstr = "Journal has aborted"; 616 else 617 errstr = "Readonly filesystem"; 618 break; 619 default: 620 /* If the caller passed in an extra buffer for unknown 621 * errors, textualise them now. Else we just return 622 * NULL. */ 623 if (nbuf) { 624 /* Check for truncated error codes... */ 625 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 626 errstr = nbuf; 627 } 628 break; 629 } 630 631 return errstr; 632 } 633 634 /* __ext4_std_error decodes expected errors from journaling functions 635 * automatically and invokes the appropriate error response. */ 636 637 void __ext4_std_error(struct super_block *sb, const char *function, 638 unsigned int line, int errno) 639 { 640 char nbuf[16]; 641 const char *errstr; 642 643 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 644 return; 645 646 /* Special case: if the error is EROFS, and we're not already 647 * inside a transaction, then there's really no point in logging 648 * an error. */ 649 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 650 return; 651 652 if (ext4_error_ratelimit(sb)) { 653 errstr = ext4_decode_error(sb, errno, nbuf); 654 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 655 sb->s_id, function, line, errstr); 656 } 657 658 save_error_info(sb, function, line); 659 ext4_handle_error(sb); 660 } 661 662 /* 663 * ext4_abort is a much stronger failure handler than ext4_error. The 664 * abort function may be used to deal with unrecoverable failures such 665 * as journal IO errors or ENOMEM at a critical moment in log management. 666 * 667 * We unconditionally force the filesystem into an ABORT|READONLY state, 668 * unless the error response on the fs has been set to panic in which 669 * case we take the easy way out and panic immediately. 670 */ 671 672 void __ext4_abort(struct super_block *sb, const char *function, 673 unsigned int line, const char *fmt, ...) 674 { 675 struct va_format vaf; 676 va_list args; 677 678 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 679 return; 680 681 save_error_info(sb, function, line); 682 va_start(args, fmt); 683 vaf.fmt = fmt; 684 vaf.va = &args; 685 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 686 sb->s_id, function, line, &vaf); 687 va_end(args); 688 689 if (sb_rdonly(sb) == 0) { 690 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 691 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 692 /* 693 * Make sure updated value of ->s_mount_flags will be visible 694 * before ->s_flags update 695 */ 696 smp_wmb(); 697 sb->s_flags |= SB_RDONLY; 698 if (EXT4_SB(sb)->s_journal) 699 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 700 save_error_info(sb, function, line); 701 } 702 if (test_opt(sb, ERRORS_PANIC)) { 703 if (EXT4_SB(sb)->s_journal && 704 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 705 return; 706 panic("EXT4-fs panic from previous error\n"); 707 } 708 } 709 710 void __ext4_msg(struct super_block *sb, 711 const char *prefix, const char *fmt, ...) 712 { 713 struct va_format vaf; 714 va_list args; 715 716 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 717 return; 718 719 va_start(args, fmt); 720 vaf.fmt = fmt; 721 vaf.va = &args; 722 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 723 va_end(args); 724 } 725 726 #define ext4_warning_ratelimit(sb) \ 727 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ 728 "EXT4-fs warning") 729 730 void __ext4_warning(struct super_block *sb, const char *function, 731 unsigned int line, const char *fmt, ...) 732 { 733 struct va_format vaf; 734 va_list args; 735 736 if (!ext4_warning_ratelimit(sb)) 737 return; 738 739 va_start(args, fmt); 740 vaf.fmt = fmt; 741 vaf.va = &args; 742 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 743 sb->s_id, function, line, &vaf); 744 va_end(args); 745 } 746 747 void __ext4_warning_inode(const struct inode *inode, const char *function, 748 unsigned int line, const char *fmt, ...) 749 { 750 struct va_format vaf; 751 va_list args; 752 753 if (!ext4_warning_ratelimit(inode->i_sb)) 754 return; 755 756 va_start(args, fmt); 757 vaf.fmt = fmt; 758 vaf.va = &args; 759 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 760 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 761 function, line, inode->i_ino, current->comm, &vaf); 762 va_end(args); 763 } 764 765 void __ext4_grp_locked_error(const char *function, unsigned int line, 766 struct super_block *sb, ext4_group_t grp, 767 unsigned long ino, ext4_fsblk_t block, 768 const char *fmt, ...) 769 __releases(bitlock) 770 __acquires(bitlock) 771 { 772 struct va_format vaf; 773 va_list args; 774 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 775 776 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 777 return; 778 779 trace_ext4_error(sb, function, line); 780 es->s_last_error_ino = cpu_to_le32(ino); 781 es->s_last_error_block = cpu_to_le64(block); 782 __save_error_info(sb, function, line); 783 784 if (ext4_error_ratelimit(sb)) { 785 va_start(args, fmt); 786 vaf.fmt = fmt; 787 vaf.va = &args; 788 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 789 sb->s_id, function, line, grp); 790 if (ino) 791 printk(KERN_CONT "inode %lu: ", ino); 792 if (block) 793 printk(KERN_CONT "block %llu:", 794 (unsigned long long) block); 795 printk(KERN_CONT "%pV\n", &vaf); 796 va_end(args); 797 } 798 799 if (test_opt(sb, WARN_ON_ERROR)) 800 WARN_ON_ONCE(1); 801 802 if (test_opt(sb, ERRORS_CONT)) { 803 ext4_commit_super(sb, 0); 804 return; 805 } 806 807 ext4_unlock_group(sb, grp); 808 ext4_commit_super(sb, 1); 809 ext4_handle_error(sb); 810 /* 811 * We only get here in the ERRORS_RO case; relocking the group 812 * may be dangerous, but nothing bad will happen since the 813 * filesystem will have already been marked read/only and the 814 * journal has been aborted. We return 1 as a hint to callers 815 * who might what to use the return value from 816 * ext4_grp_locked_error() to distinguish between the 817 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 818 * aggressively from the ext4 function in question, with a 819 * more appropriate error code. 820 */ 821 ext4_lock_group(sb, grp); 822 return; 823 } 824 825 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 826 ext4_group_t group, 827 unsigned int flags) 828 { 829 struct ext4_sb_info *sbi = EXT4_SB(sb); 830 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 831 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 832 int ret; 833 834 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 835 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 836 &grp->bb_state); 837 if (!ret) 838 percpu_counter_sub(&sbi->s_freeclusters_counter, 839 grp->bb_free); 840 } 841 842 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 843 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 844 &grp->bb_state); 845 if (!ret && gdp) { 846 int count; 847 848 count = ext4_free_inodes_count(sb, gdp); 849 percpu_counter_sub(&sbi->s_freeinodes_counter, 850 count); 851 } 852 } 853 } 854 855 void ext4_update_dynamic_rev(struct super_block *sb) 856 { 857 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 858 859 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 860 return; 861 862 ext4_warning(sb, 863 "updating to rev %d because of new feature flag, " 864 "running e2fsck is recommended", 865 EXT4_DYNAMIC_REV); 866 867 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 868 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 869 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 870 /* leave es->s_feature_*compat flags alone */ 871 /* es->s_uuid will be set by e2fsck if empty */ 872 873 /* 874 * The rest of the superblock fields should be zero, and if not it 875 * means they are likely already in use, so leave them alone. We 876 * can leave it up to e2fsck to clean up any inconsistencies there. 877 */ 878 } 879 880 /* 881 * Open the external journal device 882 */ 883 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 884 { 885 struct block_device *bdev; 886 char b[BDEVNAME_SIZE]; 887 888 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 889 if (IS_ERR(bdev)) 890 goto fail; 891 return bdev; 892 893 fail: 894 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 895 __bdevname(dev, b), PTR_ERR(bdev)); 896 return NULL; 897 } 898 899 /* 900 * Release the journal device 901 */ 902 static void ext4_blkdev_put(struct block_device *bdev) 903 { 904 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 905 } 906 907 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 908 { 909 struct block_device *bdev; 910 bdev = sbi->journal_bdev; 911 if (bdev) { 912 ext4_blkdev_put(bdev); 913 sbi->journal_bdev = NULL; 914 } 915 } 916 917 static inline struct inode *orphan_list_entry(struct list_head *l) 918 { 919 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 920 } 921 922 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 923 { 924 struct list_head *l; 925 926 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 927 le32_to_cpu(sbi->s_es->s_last_orphan)); 928 929 printk(KERN_ERR "sb_info orphan list:\n"); 930 list_for_each(l, &sbi->s_orphan) { 931 struct inode *inode = orphan_list_entry(l); 932 printk(KERN_ERR " " 933 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 934 inode->i_sb->s_id, inode->i_ino, inode, 935 inode->i_mode, inode->i_nlink, 936 NEXT_ORPHAN(inode)); 937 } 938 } 939 940 #ifdef CONFIG_QUOTA 941 static int ext4_quota_off(struct super_block *sb, int type); 942 943 static inline void ext4_quota_off_umount(struct super_block *sb) 944 { 945 int type; 946 947 /* Use our quota_off function to clear inode flags etc. */ 948 for (type = 0; type < EXT4_MAXQUOTAS; type++) 949 ext4_quota_off(sb, type); 950 } 951 952 /* 953 * This is a helper function which is used in the mount/remount 954 * codepaths (which holds s_umount) to fetch the quota file name. 955 */ 956 static inline char *get_qf_name(struct super_block *sb, 957 struct ext4_sb_info *sbi, 958 int type) 959 { 960 return rcu_dereference_protected(sbi->s_qf_names[type], 961 lockdep_is_held(&sb->s_umount)); 962 } 963 #else 964 static inline void ext4_quota_off_umount(struct super_block *sb) 965 { 966 } 967 #endif 968 969 static void ext4_put_super(struct super_block *sb) 970 { 971 struct ext4_sb_info *sbi = EXT4_SB(sb); 972 struct ext4_super_block *es = sbi->s_es; 973 int aborted = 0; 974 int i, err; 975 976 ext4_unregister_li_request(sb); 977 ext4_quota_off_umount(sb); 978 979 destroy_workqueue(sbi->rsv_conversion_wq); 980 981 if (sbi->s_journal) { 982 aborted = is_journal_aborted(sbi->s_journal); 983 err = jbd2_journal_destroy(sbi->s_journal); 984 sbi->s_journal = NULL; 985 if ((err < 0) && !aborted) 986 ext4_abort(sb, "Couldn't clean up the journal"); 987 } 988 989 ext4_unregister_sysfs(sb); 990 ext4_es_unregister_shrinker(sbi); 991 del_timer_sync(&sbi->s_err_report); 992 ext4_release_system_zone(sb); 993 ext4_mb_release(sb); 994 ext4_ext_release(sb); 995 996 if (!sb_rdonly(sb) && !aborted) { 997 ext4_clear_feature_journal_needs_recovery(sb); 998 es->s_state = cpu_to_le16(sbi->s_mount_state); 999 } 1000 if (!sb_rdonly(sb)) 1001 ext4_commit_super(sb, 1); 1002 1003 for (i = 0; i < sbi->s_gdb_count; i++) 1004 brelse(sbi->s_group_desc[i]); 1005 kvfree(sbi->s_group_desc); 1006 kvfree(sbi->s_flex_groups); 1007 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1008 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1009 percpu_counter_destroy(&sbi->s_dirs_counter); 1010 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1011 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 1012 #ifdef CONFIG_QUOTA 1013 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1014 kfree(get_qf_name(sb, sbi, i)); 1015 #endif 1016 1017 /* Debugging code just in case the in-memory inode orphan list 1018 * isn't empty. The on-disk one can be non-empty if we've 1019 * detected an error and taken the fs readonly, but the 1020 * in-memory list had better be clean by this point. */ 1021 if (!list_empty(&sbi->s_orphan)) 1022 dump_orphan_list(sb, sbi); 1023 J_ASSERT(list_empty(&sbi->s_orphan)); 1024 1025 sync_blockdev(sb->s_bdev); 1026 invalidate_bdev(sb->s_bdev); 1027 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 1028 /* 1029 * Invalidate the journal device's buffers. We don't want them 1030 * floating about in memory - the physical journal device may 1031 * hotswapped, and it breaks the `ro-after' testing code. 1032 */ 1033 sync_blockdev(sbi->journal_bdev); 1034 invalidate_bdev(sbi->journal_bdev); 1035 ext4_blkdev_remove(sbi); 1036 } 1037 1038 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1039 sbi->s_ea_inode_cache = NULL; 1040 1041 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1042 sbi->s_ea_block_cache = NULL; 1043 1044 if (sbi->s_mmp_tsk) 1045 kthread_stop(sbi->s_mmp_tsk); 1046 brelse(sbi->s_sbh); 1047 sb->s_fs_info = NULL; 1048 /* 1049 * Now that we are completely done shutting down the 1050 * superblock, we need to actually destroy the kobject. 1051 */ 1052 kobject_put(&sbi->s_kobj); 1053 wait_for_completion(&sbi->s_kobj_unregister); 1054 if (sbi->s_chksum_driver) 1055 crypto_free_shash(sbi->s_chksum_driver); 1056 kfree(sbi->s_blockgroup_lock); 1057 fs_put_dax(sbi->s_daxdev); 1058 #ifdef CONFIG_UNICODE 1059 utf8_unload(sbi->s_encoding); 1060 #endif 1061 kfree(sbi); 1062 } 1063 1064 static struct kmem_cache *ext4_inode_cachep; 1065 1066 /* 1067 * Called inside transaction, so use GFP_NOFS 1068 */ 1069 static struct inode *ext4_alloc_inode(struct super_block *sb) 1070 { 1071 struct ext4_inode_info *ei; 1072 1073 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1074 if (!ei) 1075 return NULL; 1076 1077 inode_set_iversion(&ei->vfs_inode, 1); 1078 spin_lock_init(&ei->i_raw_lock); 1079 INIT_LIST_HEAD(&ei->i_prealloc_list); 1080 spin_lock_init(&ei->i_prealloc_lock); 1081 ext4_es_init_tree(&ei->i_es_tree); 1082 rwlock_init(&ei->i_es_lock); 1083 INIT_LIST_HEAD(&ei->i_es_list); 1084 ei->i_es_all_nr = 0; 1085 ei->i_es_shk_nr = 0; 1086 ei->i_es_shrink_lblk = 0; 1087 ei->i_reserved_data_blocks = 0; 1088 ei->i_da_metadata_calc_len = 0; 1089 ei->i_da_metadata_calc_last_lblock = 0; 1090 spin_lock_init(&(ei->i_block_reservation_lock)); 1091 ext4_init_pending_tree(&ei->i_pending_tree); 1092 #ifdef CONFIG_QUOTA 1093 ei->i_reserved_quota = 0; 1094 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1095 #endif 1096 ei->jinode = NULL; 1097 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1098 spin_lock_init(&ei->i_completed_io_lock); 1099 ei->i_sync_tid = 0; 1100 ei->i_datasync_tid = 0; 1101 atomic_set(&ei->i_unwritten, 0); 1102 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1103 return &ei->vfs_inode; 1104 } 1105 1106 static int ext4_drop_inode(struct inode *inode) 1107 { 1108 int drop = generic_drop_inode(inode); 1109 1110 trace_ext4_drop_inode(inode, drop); 1111 return drop; 1112 } 1113 1114 static void ext4_free_in_core_inode(struct inode *inode) 1115 { 1116 fscrypt_free_inode(inode); 1117 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1118 } 1119 1120 static void ext4_destroy_inode(struct inode *inode) 1121 { 1122 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1123 ext4_msg(inode->i_sb, KERN_ERR, 1124 "Inode %lu (%p): orphan list check failed!", 1125 inode->i_ino, EXT4_I(inode)); 1126 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1127 EXT4_I(inode), sizeof(struct ext4_inode_info), 1128 true); 1129 dump_stack(); 1130 } 1131 } 1132 1133 static void init_once(void *foo) 1134 { 1135 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1136 1137 INIT_LIST_HEAD(&ei->i_orphan); 1138 init_rwsem(&ei->xattr_sem); 1139 init_rwsem(&ei->i_data_sem); 1140 init_rwsem(&ei->i_mmap_sem); 1141 inode_init_once(&ei->vfs_inode); 1142 } 1143 1144 static int __init init_inodecache(void) 1145 { 1146 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1147 sizeof(struct ext4_inode_info), 0, 1148 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1149 SLAB_ACCOUNT), 1150 offsetof(struct ext4_inode_info, i_data), 1151 sizeof_field(struct ext4_inode_info, i_data), 1152 init_once); 1153 if (ext4_inode_cachep == NULL) 1154 return -ENOMEM; 1155 return 0; 1156 } 1157 1158 static void destroy_inodecache(void) 1159 { 1160 /* 1161 * Make sure all delayed rcu free inodes are flushed before we 1162 * destroy cache. 1163 */ 1164 rcu_barrier(); 1165 kmem_cache_destroy(ext4_inode_cachep); 1166 } 1167 1168 void ext4_clear_inode(struct inode *inode) 1169 { 1170 invalidate_inode_buffers(inode); 1171 clear_inode(inode); 1172 dquot_drop(inode); 1173 ext4_discard_preallocations(inode); 1174 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1175 if (EXT4_I(inode)->jinode) { 1176 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1177 EXT4_I(inode)->jinode); 1178 jbd2_free_inode(EXT4_I(inode)->jinode); 1179 EXT4_I(inode)->jinode = NULL; 1180 } 1181 fscrypt_put_encryption_info(inode); 1182 } 1183 1184 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1185 u64 ino, u32 generation) 1186 { 1187 struct inode *inode; 1188 1189 /* 1190 * Currently we don't know the generation for parent directory, so 1191 * a generation of 0 means "accept any" 1192 */ 1193 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1194 if (IS_ERR(inode)) 1195 return ERR_CAST(inode); 1196 if (generation && inode->i_generation != generation) { 1197 iput(inode); 1198 return ERR_PTR(-ESTALE); 1199 } 1200 1201 return inode; 1202 } 1203 1204 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1205 int fh_len, int fh_type) 1206 { 1207 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1208 ext4_nfs_get_inode); 1209 } 1210 1211 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1212 int fh_len, int fh_type) 1213 { 1214 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1215 ext4_nfs_get_inode); 1216 } 1217 1218 static int ext4_nfs_commit_metadata(struct inode *inode) 1219 { 1220 struct writeback_control wbc = { 1221 .sync_mode = WB_SYNC_ALL 1222 }; 1223 1224 trace_ext4_nfs_commit_metadata(inode); 1225 return ext4_write_inode(inode, &wbc); 1226 } 1227 1228 /* 1229 * Try to release metadata pages (indirect blocks, directories) which are 1230 * mapped via the block device. Since these pages could have journal heads 1231 * which would prevent try_to_free_buffers() from freeing them, we must use 1232 * jbd2 layer's try_to_free_buffers() function to release them. 1233 */ 1234 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1235 gfp_t wait) 1236 { 1237 journal_t *journal = EXT4_SB(sb)->s_journal; 1238 1239 WARN_ON(PageChecked(page)); 1240 if (!page_has_buffers(page)) 1241 return 0; 1242 if (journal) 1243 return jbd2_journal_try_to_free_buffers(journal, page, 1244 wait & ~__GFP_DIRECT_RECLAIM); 1245 return try_to_free_buffers(page); 1246 } 1247 1248 #ifdef CONFIG_FS_ENCRYPTION 1249 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1250 { 1251 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1252 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1253 } 1254 1255 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1256 void *fs_data) 1257 { 1258 handle_t *handle = fs_data; 1259 int res, res2, credits, retries = 0; 1260 1261 /* 1262 * Encrypting the root directory is not allowed because e2fsck expects 1263 * lost+found to exist and be unencrypted, and encrypting the root 1264 * directory would imply encrypting the lost+found directory as well as 1265 * the filename "lost+found" itself. 1266 */ 1267 if (inode->i_ino == EXT4_ROOT_INO) 1268 return -EPERM; 1269 1270 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 1271 return -EINVAL; 1272 1273 res = ext4_convert_inline_data(inode); 1274 if (res) 1275 return res; 1276 1277 /* 1278 * If a journal handle was specified, then the encryption context is 1279 * being set on a new inode via inheritance and is part of a larger 1280 * transaction to create the inode. Otherwise the encryption context is 1281 * being set on an existing inode in its own transaction. Only in the 1282 * latter case should the "retry on ENOSPC" logic be used. 1283 */ 1284 1285 if (handle) { 1286 res = ext4_xattr_set_handle(handle, inode, 1287 EXT4_XATTR_INDEX_ENCRYPTION, 1288 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1289 ctx, len, 0); 1290 if (!res) { 1291 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1292 ext4_clear_inode_state(inode, 1293 EXT4_STATE_MAY_INLINE_DATA); 1294 /* 1295 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1296 * S_DAX may be disabled 1297 */ 1298 ext4_set_inode_flags(inode); 1299 } 1300 return res; 1301 } 1302 1303 res = dquot_initialize(inode); 1304 if (res) 1305 return res; 1306 retry: 1307 res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1308 &credits); 1309 if (res) 1310 return res; 1311 1312 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1313 if (IS_ERR(handle)) 1314 return PTR_ERR(handle); 1315 1316 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 1317 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1318 ctx, len, 0); 1319 if (!res) { 1320 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1321 /* 1322 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1323 * S_DAX may be disabled 1324 */ 1325 ext4_set_inode_flags(inode); 1326 res = ext4_mark_inode_dirty(handle, inode); 1327 if (res) 1328 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1329 } 1330 res2 = ext4_journal_stop(handle); 1331 1332 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 1333 goto retry; 1334 if (!res) 1335 res = res2; 1336 return res; 1337 } 1338 1339 static bool ext4_dummy_context(struct inode *inode) 1340 { 1341 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); 1342 } 1343 1344 static const struct fscrypt_operations ext4_cryptops = { 1345 .key_prefix = "ext4:", 1346 .get_context = ext4_get_context, 1347 .set_context = ext4_set_context, 1348 .dummy_context = ext4_dummy_context, 1349 .empty_dir = ext4_empty_dir, 1350 .max_namelen = EXT4_NAME_LEN, 1351 }; 1352 #endif 1353 1354 #ifdef CONFIG_QUOTA 1355 static const char * const quotatypes[] = INITQFNAMES; 1356 #define QTYPE2NAME(t) (quotatypes[t]) 1357 1358 static int ext4_write_dquot(struct dquot *dquot); 1359 static int ext4_acquire_dquot(struct dquot *dquot); 1360 static int ext4_release_dquot(struct dquot *dquot); 1361 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1362 static int ext4_write_info(struct super_block *sb, int type); 1363 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1364 const struct path *path); 1365 static int ext4_quota_on_mount(struct super_block *sb, int type); 1366 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1367 size_t len, loff_t off); 1368 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1369 const char *data, size_t len, loff_t off); 1370 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1371 unsigned int flags); 1372 static int ext4_enable_quotas(struct super_block *sb); 1373 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); 1374 1375 static struct dquot **ext4_get_dquots(struct inode *inode) 1376 { 1377 return EXT4_I(inode)->i_dquot; 1378 } 1379 1380 static const struct dquot_operations ext4_quota_operations = { 1381 .get_reserved_space = ext4_get_reserved_space, 1382 .write_dquot = ext4_write_dquot, 1383 .acquire_dquot = ext4_acquire_dquot, 1384 .release_dquot = ext4_release_dquot, 1385 .mark_dirty = ext4_mark_dquot_dirty, 1386 .write_info = ext4_write_info, 1387 .alloc_dquot = dquot_alloc, 1388 .destroy_dquot = dquot_destroy, 1389 .get_projid = ext4_get_projid, 1390 .get_inode_usage = ext4_get_inode_usage, 1391 .get_next_id = ext4_get_next_id, 1392 }; 1393 1394 static const struct quotactl_ops ext4_qctl_operations = { 1395 .quota_on = ext4_quota_on, 1396 .quota_off = ext4_quota_off, 1397 .quota_sync = dquot_quota_sync, 1398 .get_state = dquot_get_state, 1399 .set_info = dquot_set_dqinfo, 1400 .get_dqblk = dquot_get_dqblk, 1401 .set_dqblk = dquot_set_dqblk, 1402 .get_nextdqblk = dquot_get_next_dqblk, 1403 }; 1404 #endif 1405 1406 static const struct super_operations ext4_sops = { 1407 .alloc_inode = ext4_alloc_inode, 1408 .free_inode = ext4_free_in_core_inode, 1409 .destroy_inode = ext4_destroy_inode, 1410 .write_inode = ext4_write_inode, 1411 .dirty_inode = ext4_dirty_inode, 1412 .drop_inode = ext4_drop_inode, 1413 .evict_inode = ext4_evict_inode, 1414 .put_super = ext4_put_super, 1415 .sync_fs = ext4_sync_fs, 1416 .freeze_fs = ext4_freeze, 1417 .unfreeze_fs = ext4_unfreeze, 1418 .statfs = ext4_statfs, 1419 .remount_fs = ext4_remount, 1420 .show_options = ext4_show_options, 1421 #ifdef CONFIG_QUOTA 1422 .quota_read = ext4_quota_read, 1423 .quota_write = ext4_quota_write, 1424 .get_dquots = ext4_get_dquots, 1425 #endif 1426 .bdev_try_to_free_page = bdev_try_to_free_page, 1427 }; 1428 1429 static const struct export_operations ext4_export_ops = { 1430 .fh_to_dentry = ext4_fh_to_dentry, 1431 .fh_to_parent = ext4_fh_to_parent, 1432 .get_parent = ext4_get_parent, 1433 .commit_metadata = ext4_nfs_commit_metadata, 1434 }; 1435 1436 enum { 1437 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1438 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1439 Opt_nouid32, Opt_debug, Opt_removed, 1440 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1441 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1442 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1443 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1444 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1445 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1446 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1447 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1448 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1449 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, 1450 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1451 Opt_nowarn_on_error, Opt_mblk_io_submit, 1452 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 1453 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1454 Opt_inode_readahead_blks, Opt_journal_ioprio, 1455 Opt_dioread_nolock, Opt_dioread_lock, 1456 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1457 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1458 }; 1459 1460 static const match_table_t tokens = { 1461 {Opt_bsd_df, "bsddf"}, 1462 {Opt_minix_df, "minixdf"}, 1463 {Opt_grpid, "grpid"}, 1464 {Opt_grpid, "bsdgroups"}, 1465 {Opt_nogrpid, "nogrpid"}, 1466 {Opt_nogrpid, "sysvgroups"}, 1467 {Opt_resgid, "resgid=%u"}, 1468 {Opt_resuid, "resuid=%u"}, 1469 {Opt_sb, "sb=%u"}, 1470 {Opt_err_cont, "errors=continue"}, 1471 {Opt_err_panic, "errors=panic"}, 1472 {Opt_err_ro, "errors=remount-ro"}, 1473 {Opt_nouid32, "nouid32"}, 1474 {Opt_debug, "debug"}, 1475 {Opt_removed, "oldalloc"}, 1476 {Opt_removed, "orlov"}, 1477 {Opt_user_xattr, "user_xattr"}, 1478 {Opt_nouser_xattr, "nouser_xattr"}, 1479 {Opt_acl, "acl"}, 1480 {Opt_noacl, "noacl"}, 1481 {Opt_noload, "norecovery"}, 1482 {Opt_noload, "noload"}, 1483 {Opt_removed, "nobh"}, 1484 {Opt_removed, "bh"}, 1485 {Opt_commit, "commit=%u"}, 1486 {Opt_min_batch_time, "min_batch_time=%u"}, 1487 {Opt_max_batch_time, "max_batch_time=%u"}, 1488 {Opt_journal_dev, "journal_dev=%u"}, 1489 {Opt_journal_path, "journal_path=%s"}, 1490 {Opt_journal_checksum, "journal_checksum"}, 1491 {Opt_nojournal_checksum, "nojournal_checksum"}, 1492 {Opt_journal_async_commit, "journal_async_commit"}, 1493 {Opt_abort, "abort"}, 1494 {Opt_data_journal, "data=journal"}, 1495 {Opt_data_ordered, "data=ordered"}, 1496 {Opt_data_writeback, "data=writeback"}, 1497 {Opt_data_err_abort, "data_err=abort"}, 1498 {Opt_data_err_ignore, "data_err=ignore"}, 1499 {Opt_offusrjquota, "usrjquota="}, 1500 {Opt_usrjquota, "usrjquota=%s"}, 1501 {Opt_offgrpjquota, "grpjquota="}, 1502 {Opt_grpjquota, "grpjquota=%s"}, 1503 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1504 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1505 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1506 {Opt_grpquota, "grpquota"}, 1507 {Opt_noquota, "noquota"}, 1508 {Opt_quota, "quota"}, 1509 {Opt_usrquota, "usrquota"}, 1510 {Opt_prjquota, "prjquota"}, 1511 {Opt_barrier, "barrier=%u"}, 1512 {Opt_barrier, "barrier"}, 1513 {Opt_nobarrier, "nobarrier"}, 1514 {Opt_i_version, "i_version"}, 1515 {Opt_dax, "dax"}, 1516 {Opt_stripe, "stripe=%u"}, 1517 {Opt_delalloc, "delalloc"}, 1518 {Opt_warn_on_error, "warn_on_error"}, 1519 {Opt_nowarn_on_error, "nowarn_on_error"}, 1520 {Opt_lazytime, "lazytime"}, 1521 {Opt_nolazytime, "nolazytime"}, 1522 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1523 {Opt_nodelalloc, "nodelalloc"}, 1524 {Opt_removed, "mblk_io_submit"}, 1525 {Opt_removed, "nomblk_io_submit"}, 1526 {Opt_block_validity, "block_validity"}, 1527 {Opt_noblock_validity, "noblock_validity"}, 1528 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1529 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1530 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1531 {Opt_auto_da_alloc, "auto_da_alloc"}, 1532 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1533 {Opt_dioread_nolock, "dioread_nolock"}, 1534 {Opt_dioread_lock, "dioread_lock"}, 1535 {Opt_discard, "discard"}, 1536 {Opt_nodiscard, "nodiscard"}, 1537 {Opt_init_itable, "init_itable=%u"}, 1538 {Opt_init_itable, "init_itable"}, 1539 {Opt_noinit_itable, "noinit_itable"}, 1540 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1541 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1542 {Opt_nombcache, "nombcache"}, 1543 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1544 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1545 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1546 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1547 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1548 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1549 {Opt_err, NULL}, 1550 }; 1551 1552 static ext4_fsblk_t get_sb_block(void **data) 1553 { 1554 ext4_fsblk_t sb_block; 1555 char *options = (char *) *data; 1556 1557 if (!options || strncmp(options, "sb=", 3) != 0) 1558 return 1; /* Default location */ 1559 1560 options += 3; 1561 /* TODO: use simple_strtoll with >32bit ext4 */ 1562 sb_block = simple_strtoul(options, &options, 0); 1563 if (*options && *options != ',') { 1564 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1565 (char *) *data); 1566 return 1; 1567 } 1568 if (*options == ',') 1569 options++; 1570 *data = (void *) options; 1571 1572 return sb_block; 1573 } 1574 1575 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1576 static const char deprecated_msg[] = 1577 "Mount option \"%s\" will be removed by %s\n" 1578 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1579 1580 #ifdef CONFIG_QUOTA 1581 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1582 { 1583 struct ext4_sb_info *sbi = EXT4_SB(sb); 1584 char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 1585 int ret = -1; 1586 1587 if (sb_any_quota_loaded(sb) && !old_qname) { 1588 ext4_msg(sb, KERN_ERR, 1589 "Cannot change journaled " 1590 "quota options when quota turned on"); 1591 return -1; 1592 } 1593 if (ext4_has_feature_quota(sb)) { 1594 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1595 "ignored when QUOTA feature is enabled"); 1596 return 1; 1597 } 1598 qname = match_strdup(args); 1599 if (!qname) { 1600 ext4_msg(sb, KERN_ERR, 1601 "Not enough memory for storing quotafile name"); 1602 return -1; 1603 } 1604 if (old_qname) { 1605 if (strcmp(old_qname, qname) == 0) 1606 ret = 1; 1607 else 1608 ext4_msg(sb, KERN_ERR, 1609 "%s quota file already specified", 1610 QTYPE2NAME(qtype)); 1611 goto errout; 1612 } 1613 if (strchr(qname, '/')) { 1614 ext4_msg(sb, KERN_ERR, 1615 "quotafile must be on filesystem root"); 1616 goto errout; 1617 } 1618 rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1619 set_opt(sb, QUOTA); 1620 return 1; 1621 errout: 1622 kfree(qname); 1623 return ret; 1624 } 1625 1626 static int clear_qf_name(struct super_block *sb, int qtype) 1627 { 1628 1629 struct ext4_sb_info *sbi = EXT4_SB(sb); 1630 char *old_qname = get_qf_name(sb, sbi, qtype); 1631 1632 if (sb_any_quota_loaded(sb) && old_qname) { 1633 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1634 " when quota turned on"); 1635 return -1; 1636 } 1637 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 1638 synchronize_rcu(); 1639 kfree(old_qname); 1640 return 1; 1641 } 1642 #endif 1643 1644 #define MOPT_SET 0x0001 1645 #define MOPT_CLEAR 0x0002 1646 #define MOPT_NOSUPPORT 0x0004 1647 #define MOPT_EXPLICIT 0x0008 1648 #define MOPT_CLEAR_ERR 0x0010 1649 #define MOPT_GTE0 0x0020 1650 #ifdef CONFIG_QUOTA 1651 #define MOPT_Q 0 1652 #define MOPT_QFMT 0x0040 1653 #else 1654 #define MOPT_Q MOPT_NOSUPPORT 1655 #define MOPT_QFMT MOPT_NOSUPPORT 1656 #endif 1657 #define MOPT_DATAJ 0x0080 1658 #define MOPT_NO_EXT2 0x0100 1659 #define MOPT_NO_EXT3 0x0200 1660 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1661 #define MOPT_STRING 0x0400 1662 1663 static const struct mount_opts { 1664 int token; 1665 int mount_opt; 1666 int flags; 1667 } ext4_mount_opts[] = { 1668 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1669 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1670 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1671 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1672 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1673 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1674 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1675 MOPT_EXT4_ONLY | MOPT_SET}, 1676 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1677 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1678 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1679 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1680 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1681 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1682 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1683 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1684 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1685 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1686 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1687 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1688 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1689 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1690 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1691 EXT4_MOUNT_JOURNAL_CHECKSUM), 1692 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1693 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1694 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1695 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1696 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1697 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1698 MOPT_NO_EXT2}, 1699 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1700 MOPT_NO_EXT2}, 1701 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1702 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1703 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1704 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1705 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1706 {Opt_commit, 0, MOPT_GTE0}, 1707 {Opt_max_batch_time, 0, MOPT_GTE0}, 1708 {Opt_min_batch_time, 0, MOPT_GTE0}, 1709 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1710 {Opt_init_itable, 0, MOPT_GTE0}, 1711 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 1712 {Opt_stripe, 0, MOPT_GTE0}, 1713 {Opt_resuid, 0, MOPT_GTE0}, 1714 {Opt_resgid, 0, MOPT_GTE0}, 1715 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1716 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1717 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1718 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1719 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1720 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1721 MOPT_NO_EXT2 | MOPT_DATAJ}, 1722 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1723 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1724 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1725 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1726 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1727 #else 1728 {Opt_acl, 0, MOPT_NOSUPPORT}, 1729 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1730 #endif 1731 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1732 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1733 {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 1734 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1735 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1736 MOPT_SET | MOPT_Q}, 1737 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1738 MOPT_SET | MOPT_Q}, 1739 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1740 MOPT_SET | MOPT_Q}, 1741 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1742 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1743 MOPT_CLEAR | MOPT_Q}, 1744 {Opt_usrjquota, 0, MOPT_Q}, 1745 {Opt_grpjquota, 0, MOPT_Q}, 1746 {Opt_offusrjquota, 0, MOPT_Q}, 1747 {Opt_offgrpjquota, 0, MOPT_Q}, 1748 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1749 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1750 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1751 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1752 {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1753 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1754 {Opt_err, 0, 0} 1755 }; 1756 1757 #ifdef CONFIG_UNICODE 1758 static const struct ext4_sb_encodings { 1759 __u16 magic; 1760 char *name; 1761 char *version; 1762 } ext4_sb_encoding_map[] = { 1763 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1764 }; 1765 1766 static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1767 const struct ext4_sb_encodings **encoding, 1768 __u16 *flags) 1769 { 1770 __u16 magic = le16_to_cpu(es->s_encoding); 1771 int i; 1772 1773 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1774 if (magic == ext4_sb_encoding_map[i].magic) 1775 break; 1776 1777 if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1778 return -EINVAL; 1779 1780 *encoding = &ext4_sb_encoding_map[i]; 1781 *flags = le16_to_cpu(es->s_encoding_flags); 1782 1783 return 0; 1784 } 1785 #endif 1786 1787 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1788 substring_t *args, unsigned long *journal_devnum, 1789 unsigned int *journal_ioprio, int is_remount) 1790 { 1791 struct ext4_sb_info *sbi = EXT4_SB(sb); 1792 const struct mount_opts *m; 1793 kuid_t uid; 1794 kgid_t gid; 1795 int arg = 0; 1796 1797 #ifdef CONFIG_QUOTA 1798 if (token == Opt_usrjquota) 1799 return set_qf_name(sb, USRQUOTA, &args[0]); 1800 else if (token == Opt_grpjquota) 1801 return set_qf_name(sb, GRPQUOTA, &args[0]); 1802 else if (token == Opt_offusrjquota) 1803 return clear_qf_name(sb, USRQUOTA); 1804 else if (token == Opt_offgrpjquota) 1805 return clear_qf_name(sb, GRPQUOTA); 1806 #endif 1807 switch (token) { 1808 case Opt_noacl: 1809 case Opt_nouser_xattr: 1810 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1811 break; 1812 case Opt_sb: 1813 return 1; /* handled by get_sb_block() */ 1814 case Opt_removed: 1815 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1816 return 1; 1817 case Opt_abort: 1818 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1819 return 1; 1820 case Opt_i_version: 1821 sb->s_flags |= SB_I_VERSION; 1822 return 1; 1823 case Opt_lazytime: 1824 sb->s_flags |= SB_LAZYTIME; 1825 return 1; 1826 case Opt_nolazytime: 1827 sb->s_flags &= ~SB_LAZYTIME; 1828 return 1; 1829 } 1830 1831 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1832 if (token == m->token) 1833 break; 1834 1835 if (m->token == Opt_err) { 1836 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1837 "or missing value", opt); 1838 return -1; 1839 } 1840 1841 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1842 ext4_msg(sb, KERN_ERR, 1843 "Mount option \"%s\" incompatible with ext2", opt); 1844 return -1; 1845 } 1846 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1847 ext4_msg(sb, KERN_ERR, 1848 "Mount option \"%s\" incompatible with ext3", opt); 1849 return -1; 1850 } 1851 1852 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1853 return -1; 1854 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1855 return -1; 1856 if (m->flags & MOPT_EXPLICIT) { 1857 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 1858 set_opt2(sb, EXPLICIT_DELALLOC); 1859 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 1860 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1861 } else 1862 return -1; 1863 } 1864 if (m->flags & MOPT_CLEAR_ERR) 1865 clear_opt(sb, ERRORS_MASK); 1866 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1867 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1868 "options when quota turned on"); 1869 return -1; 1870 } 1871 1872 if (m->flags & MOPT_NOSUPPORT) { 1873 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1874 } else if (token == Opt_commit) { 1875 if (arg == 0) 1876 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1877 sbi->s_commit_interval = HZ * arg; 1878 } else if (token == Opt_debug_want_extra_isize) { 1879 sbi->s_want_extra_isize = arg; 1880 } else if (token == Opt_max_batch_time) { 1881 sbi->s_max_batch_time = arg; 1882 } else if (token == Opt_min_batch_time) { 1883 sbi->s_min_batch_time = arg; 1884 } else if (token == Opt_inode_readahead_blks) { 1885 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1886 ext4_msg(sb, KERN_ERR, 1887 "EXT4-fs: inode_readahead_blks must be " 1888 "0 or a power of 2 smaller than 2^31"); 1889 return -1; 1890 } 1891 sbi->s_inode_readahead_blks = arg; 1892 } else if (token == Opt_init_itable) { 1893 set_opt(sb, INIT_INODE_TABLE); 1894 if (!args->from) 1895 arg = EXT4_DEF_LI_WAIT_MULT; 1896 sbi->s_li_wait_mult = arg; 1897 } else if (token == Opt_max_dir_size_kb) { 1898 sbi->s_max_dir_size_kb = arg; 1899 } else if (token == Opt_stripe) { 1900 sbi->s_stripe = arg; 1901 } else if (token == Opt_resuid) { 1902 uid = make_kuid(current_user_ns(), arg); 1903 if (!uid_valid(uid)) { 1904 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1905 return -1; 1906 } 1907 sbi->s_resuid = uid; 1908 } else if (token == Opt_resgid) { 1909 gid = make_kgid(current_user_ns(), arg); 1910 if (!gid_valid(gid)) { 1911 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1912 return -1; 1913 } 1914 sbi->s_resgid = gid; 1915 } else if (token == Opt_journal_dev) { 1916 if (is_remount) { 1917 ext4_msg(sb, KERN_ERR, 1918 "Cannot specify journal on remount"); 1919 return -1; 1920 } 1921 *journal_devnum = arg; 1922 } else if (token == Opt_journal_path) { 1923 char *journal_path; 1924 struct inode *journal_inode; 1925 struct path path; 1926 int error; 1927 1928 if (is_remount) { 1929 ext4_msg(sb, KERN_ERR, 1930 "Cannot specify journal on remount"); 1931 return -1; 1932 } 1933 journal_path = match_strdup(&args[0]); 1934 if (!journal_path) { 1935 ext4_msg(sb, KERN_ERR, "error: could not dup " 1936 "journal device string"); 1937 return -1; 1938 } 1939 1940 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1941 if (error) { 1942 ext4_msg(sb, KERN_ERR, "error: could not find " 1943 "journal device path: error %d", error); 1944 kfree(journal_path); 1945 return -1; 1946 } 1947 1948 journal_inode = d_inode(path.dentry); 1949 if (!S_ISBLK(journal_inode->i_mode)) { 1950 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1951 "is not a block device", journal_path); 1952 path_put(&path); 1953 kfree(journal_path); 1954 return -1; 1955 } 1956 1957 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1958 path_put(&path); 1959 kfree(journal_path); 1960 } else if (token == Opt_journal_ioprio) { 1961 if (arg > 7) { 1962 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1963 " (must be 0-7)"); 1964 return -1; 1965 } 1966 *journal_ioprio = 1967 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1968 } else if (token == Opt_test_dummy_encryption) { 1969 #ifdef CONFIG_FS_ENCRYPTION 1970 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 1971 ext4_msg(sb, KERN_WARNING, 1972 "Test dummy encryption mode enabled"); 1973 #else 1974 ext4_msg(sb, KERN_WARNING, 1975 "Test dummy encryption mount option ignored"); 1976 #endif 1977 } else if (m->flags & MOPT_DATAJ) { 1978 if (is_remount) { 1979 if (!sbi->s_journal) 1980 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1981 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 1982 ext4_msg(sb, KERN_ERR, 1983 "Cannot change data mode on remount"); 1984 return -1; 1985 } 1986 } else { 1987 clear_opt(sb, DATA_FLAGS); 1988 sbi->s_mount_opt |= m->mount_opt; 1989 } 1990 #ifdef CONFIG_QUOTA 1991 } else if (m->flags & MOPT_QFMT) { 1992 if (sb_any_quota_loaded(sb) && 1993 sbi->s_jquota_fmt != m->mount_opt) { 1994 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 1995 "quota options when quota turned on"); 1996 return -1; 1997 } 1998 if (ext4_has_feature_quota(sb)) { 1999 ext4_msg(sb, KERN_INFO, 2000 "Quota format mount options ignored " 2001 "when QUOTA feature is enabled"); 2002 return 1; 2003 } 2004 sbi->s_jquota_fmt = m->mount_opt; 2005 #endif 2006 } else if (token == Opt_dax) { 2007 #ifdef CONFIG_FS_DAX 2008 ext4_msg(sb, KERN_WARNING, 2009 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 2010 sbi->s_mount_opt |= m->mount_opt; 2011 #else 2012 ext4_msg(sb, KERN_INFO, "dax option not supported"); 2013 return -1; 2014 #endif 2015 } else if (token == Opt_data_err_abort) { 2016 sbi->s_mount_opt |= m->mount_opt; 2017 } else if (token == Opt_data_err_ignore) { 2018 sbi->s_mount_opt &= ~m->mount_opt; 2019 } else { 2020 if (!args->from) 2021 arg = 1; 2022 if (m->flags & MOPT_CLEAR) 2023 arg = !arg; 2024 else if (unlikely(!(m->flags & MOPT_SET))) { 2025 ext4_msg(sb, KERN_WARNING, 2026 "buggy handling of option %s", opt); 2027 WARN_ON(1); 2028 return -1; 2029 } 2030 if (arg != 0) 2031 sbi->s_mount_opt |= m->mount_opt; 2032 else 2033 sbi->s_mount_opt &= ~m->mount_opt; 2034 } 2035 return 1; 2036 } 2037 2038 static int parse_options(char *options, struct super_block *sb, 2039 unsigned long *journal_devnum, 2040 unsigned int *journal_ioprio, 2041 int is_remount) 2042 { 2043 struct ext4_sb_info *sbi = EXT4_SB(sb); 2044 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2045 substring_t args[MAX_OPT_ARGS]; 2046 int token; 2047 2048 if (!options) 2049 return 1; 2050 2051 while ((p = strsep(&options, ",")) != NULL) { 2052 if (!*p) 2053 continue; 2054 /* 2055 * Initialize args struct so we know whether arg was 2056 * found; some options take optional arguments. 2057 */ 2058 args[0].to = args[0].from = NULL; 2059 token = match_token(p, tokens, args); 2060 if (handle_mount_opt(sb, p, token, args, journal_devnum, 2061 journal_ioprio, is_remount) < 0) 2062 return 0; 2063 } 2064 #ifdef CONFIG_QUOTA 2065 /* 2066 * We do the test below only for project quotas. 'usrquota' and 2067 * 'grpquota' mount options are allowed even without quota feature 2068 * to support legacy quotas in quota files. 2069 */ 2070 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 2071 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 2072 "Cannot enable project quota enforcement."); 2073 return 0; 2074 } 2075 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 2076 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 2077 if (usr_qf_name || grp_qf_name) { 2078 if (test_opt(sb, USRQUOTA) && usr_qf_name) 2079 clear_opt(sb, USRQUOTA); 2080 2081 if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2082 clear_opt(sb, GRPQUOTA); 2083 2084 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2085 ext4_msg(sb, KERN_ERR, "old and new quota " 2086 "format mixing"); 2087 return 0; 2088 } 2089 2090 if (!sbi->s_jquota_fmt) { 2091 ext4_msg(sb, KERN_ERR, "journaled quota format " 2092 "not specified"); 2093 return 0; 2094 } 2095 } 2096 #endif 2097 if (test_opt(sb, DIOREAD_NOLOCK)) { 2098 int blocksize = 2099 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2100 2101 if (blocksize < PAGE_SIZE) { 2102 ext4_msg(sb, KERN_ERR, "can't mount with " 2103 "dioread_nolock if block size != PAGE_SIZE"); 2104 return 0; 2105 } 2106 } 2107 return 1; 2108 } 2109 2110 static inline void ext4_show_quota_options(struct seq_file *seq, 2111 struct super_block *sb) 2112 { 2113 #if defined(CONFIG_QUOTA) 2114 struct ext4_sb_info *sbi = EXT4_SB(sb); 2115 char *usr_qf_name, *grp_qf_name; 2116 2117 if (sbi->s_jquota_fmt) { 2118 char *fmtname = ""; 2119 2120 switch (sbi->s_jquota_fmt) { 2121 case QFMT_VFS_OLD: 2122 fmtname = "vfsold"; 2123 break; 2124 case QFMT_VFS_V0: 2125 fmtname = "vfsv0"; 2126 break; 2127 case QFMT_VFS_V1: 2128 fmtname = "vfsv1"; 2129 break; 2130 } 2131 seq_printf(seq, ",jqfmt=%s", fmtname); 2132 } 2133 2134 rcu_read_lock(); 2135 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2136 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2137 if (usr_qf_name) 2138 seq_show_option(seq, "usrjquota", usr_qf_name); 2139 if (grp_qf_name) 2140 seq_show_option(seq, "grpjquota", grp_qf_name); 2141 rcu_read_unlock(); 2142 #endif 2143 } 2144 2145 static const char *token2str(int token) 2146 { 2147 const struct match_token *t; 2148 2149 for (t = tokens; t->token != Opt_err; t++) 2150 if (t->token == token && !strchr(t->pattern, '=')) 2151 break; 2152 return t->pattern; 2153 } 2154 2155 /* 2156 * Show an option if 2157 * - it's set to a non-default value OR 2158 * - if the per-sb default is different from the global default 2159 */ 2160 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2161 int nodefs) 2162 { 2163 struct ext4_sb_info *sbi = EXT4_SB(sb); 2164 struct ext4_super_block *es = sbi->s_es; 2165 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2166 const struct mount_opts *m; 2167 char sep = nodefs ? '\n' : ','; 2168 2169 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2170 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2171 2172 if (sbi->s_sb_block != 1) 2173 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2174 2175 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2176 int want_set = m->flags & MOPT_SET; 2177 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2178 (m->flags & MOPT_CLEAR_ERR)) 2179 continue; 2180 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2181 continue; /* skip if same as the default */ 2182 if ((want_set && 2183 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2184 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2185 continue; /* select Opt_noFoo vs Opt_Foo */ 2186 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2187 } 2188 2189 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2190 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2191 SEQ_OPTS_PRINT("resuid=%u", 2192 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2193 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2194 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2195 SEQ_OPTS_PRINT("resgid=%u", 2196 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2197 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2198 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2199 SEQ_OPTS_PUTS("errors=remount-ro"); 2200 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2201 SEQ_OPTS_PUTS("errors=continue"); 2202 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2203 SEQ_OPTS_PUTS("errors=panic"); 2204 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2205 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2206 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2207 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2208 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2209 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2210 if (sb->s_flags & SB_I_VERSION) 2211 SEQ_OPTS_PUTS("i_version"); 2212 if (nodefs || sbi->s_stripe) 2213 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2214 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2215 (sbi->s_mount_opt ^ def_mount_opt)) { 2216 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2217 SEQ_OPTS_PUTS("data=journal"); 2218 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2219 SEQ_OPTS_PUTS("data=ordered"); 2220 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2221 SEQ_OPTS_PUTS("data=writeback"); 2222 } 2223 if (nodefs || 2224 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2225 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2226 sbi->s_inode_readahead_blks); 2227 2228 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2229 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2230 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2231 if (nodefs || sbi->s_max_dir_size_kb) 2232 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2233 if (test_opt(sb, DATA_ERR_ABORT)) 2234 SEQ_OPTS_PUTS("data_err=abort"); 2235 if (DUMMY_ENCRYPTION_ENABLED(sbi)) 2236 SEQ_OPTS_PUTS("test_dummy_encryption"); 2237 2238 ext4_show_quota_options(seq, sb); 2239 return 0; 2240 } 2241 2242 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2243 { 2244 return _ext4_show_options(seq, root->d_sb, 0); 2245 } 2246 2247 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2248 { 2249 struct super_block *sb = seq->private; 2250 int rc; 2251 2252 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 2253 rc = _ext4_show_options(seq, sb, 1); 2254 seq_puts(seq, "\n"); 2255 return rc; 2256 } 2257 2258 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2259 int read_only) 2260 { 2261 struct ext4_sb_info *sbi = EXT4_SB(sb); 2262 int err = 0; 2263 2264 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2265 ext4_msg(sb, KERN_ERR, "revision level too high, " 2266 "forcing read-only mode"); 2267 err = -EROFS; 2268 } 2269 if (read_only) 2270 goto done; 2271 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2272 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2273 "running e2fsck is recommended"); 2274 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2275 ext4_msg(sb, KERN_WARNING, 2276 "warning: mounting fs with errors, " 2277 "running e2fsck is recommended"); 2278 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2279 le16_to_cpu(es->s_mnt_count) >= 2280 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2281 ext4_msg(sb, KERN_WARNING, 2282 "warning: maximal mount count reached, " 2283 "running e2fsck is recommended"); 2284 else if (le32_to_cpu(es->s_checkinterval) && 2285 (ext4_get_tstamp(es, s_lastcheck) + 2286 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2287 ext4_msg(sb, KERN_WARNING, 2288 "warning: checktime reached, " 2289 "running e2fsck is recommended"); 2290 if (!sbi->s_journal) 2291 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2292 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2293 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2294 le16_add_cpu(&es->s_mnt_count, 1); 2295 ext4_update_tstamp(es, s_mtime); 2296 if (sbi->s_journal) 2297 ext4_set_feature_journal_needs_recovery(sb); 2298 2299 err = ext4_commit_super(sb, 1); 2300 done: 2301 if (test_opt(sb, DEBUG)) 2302 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2303 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2304 sb->s_blocksize, 2305 sbi->s_groups_count, 2306 EXT4_BLOCKS_PER_GROUP(sb), 2307 EXT4_INODES_PER_GROUP(sb), 2308 sbi->s_mount_opt, sbi->s_mount_opt2); 2309 2310 cleancache_init_fs(sb); 2311 return err; 2312 } 2313 2314 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2315 { 2316 struct ext4_sb_info *sbi = EXT4_SB(sb); 2317 struct flex_groups *new_groups; 2318 int size; 2319 2320 if (!sbi->s_log_groups_per_flex) 2321 return 0; 2322 2323 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2324 if (size <= sbi->s_flex_groups_allocated) 2325 return 0; 2326 2327 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2328 new_groups = kvzalloc(size, GFP_KERNEL); 2329 if (!new_groups) { 2330 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2331 size / (int) sizeof(struct flex_groups)); 2332 return -ENOMEM; 2333 } 2334 2335 if (sbi->s_flex_groups) { 2336 memcpy(new_groups, sbi->s_flex_groups, 2337 (sbi->s_flex_groups_allocated * 2338 sizeof(struct flex_groups))); 2339 kvfree(sbi->s_flex_groups); 2340 } 2341 sbi->s_flex_groups = new_groups; 2342 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2343 return 0; 2344 } 2345 2346 static int ext4_fill_flex_info(struct super_block *sb) 2347 { 2348 struct ext4_sb_info *sbi = EXT4_SB(sb); 2349 struct ext4_group_desc *gdp = NULL; 2350 ext4_group_t flex_group; 2351 int i, err; 2352 2353 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2354 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2355 sbi->s_log_groups_per_flex = 0; 2356 return 1; 2357 } 2358 2359 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2360 if (err) 2361 goto failed; 2362 2363 for (i = 0; i < sbi->s_groups_count; i++) { 2364 gdp = ext4_get_group_desc(sb, i, NULL); 2365 2366 flex_group = ext4_flex_group(sbi, i); 2367 atomic_add(ext4_free_inodes_count(sb, gdp), 2368 &sbi->s_flex_groups[flex_group].free_inodes); 2369 atomic64_add(ext4_free_group_clusters(sb, gdp), 2370 &sbi->s_flex_groups[flex_group].free_clusters); 2371 atomic_add(ext4_used_dirs_count(sb, gdp), 2372 &sbi->s_flex_groups[flex_group].used_dirs); 2373 } 2374 2375 return 1; 2376 failed: 2377 return 0; 2378 } 2379 2380 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2381 struct ext4_group_desc *gdp) 2382 { 2383 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2384 __u16 crc = 0; 2385 __le32 le_group = cpu_to_le32(block_group); 2386 struct ext4_sb_info *sbi = EXT4_SB(sb); 2387 2388 if (ext4_has_metadata_csum(sbi->s_sb)) { 2389 /* Use new metadata_csum algorithm */ 2390 __u32 csum32; 2391 __u16 dummy_csum = 0; 2392 2393 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2394 sizeof(le_group)); 2395 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2396 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2397 sizeof(dummy_csum)); 2398 offset += sizeof(dummy_csum); 2399 if (offset < sbi->s_desc_size) 2400 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2401 sbi->s_desc_size - offset); 2402 2403 crc = csum32 & 0xFFFF; 2404 goto out; 2405 } 2406 2407 /* old crc16 code */ 2408 if (!ext4_has_feature_gdt_csum(sb)) 2409 return 0; 2410 2411 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2412 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2413 crc = crc16(crc, (__u8 *)gdp, offset); 2414 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2415 /* for checksum of struct ext4_group_desc do the rest...*/ 2416 if (ext4_has_feature_64bit(sb) && 2417 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2418 crc = crc16(crc, (__u8 *)gdp + offset, 2419 le16_to_cpu(sbi->s_es->s_desc_size) - 2420 offset); 2421 2422 out: 2423 return cpu_to_le16(crc); 2424 } 2425 2426 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2427 struct ext4_group_desc *gdp) 2428 { 2429 if (ext4_has_group_desc_csum(sb) && 2430 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2431 return 0; 2432 2433 return 1; 2434 } 2435 2436 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2437 struct ext4_group_desc *gdp) 2438 { 2439 if (!ext4_has_group_desc_csum(sb)) 2440 return; 2441 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2442 } 2443 2444 /* Called at mount-time, super-block is locked */ 2445 static int ext4_check_descriptors(struct super_block *sb, 2446 ext4_fsblk_t sb_block, 2447 ext4_group_t *first_not_zeroed) 2448 { 2449 struct ext4_sb_info *sbi = EXT4_SB(sb); 2450 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2451 ext4_fsblk_t last_block; 2452 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2453 ext4_fsblk_t block_bitmap; 2454 ext4_fsblk_t inode_bitmap; 2455 ext4_fsblk_t inode_table; 2456 int flexbg_flag = 0; 2457 ext4_group_t i, grp = sbi->s_groups_count; 2458 2459 if (ext4_has_feature_flex_bg(sb)) 2460 flexbg_flag = 1; 2461 2462 ext4_debug("Checking group descriptors"); 2463 2464 for (i = 0; i < sbi->s_groups_count; i++) { 2465 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2466 2467 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2468 last_block = ext4_blocks_count(sbi->s_es) - 1; 2469 else 2470 last_block = first_block + 2471 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2472 2473 if ((grp == sbi->s_groups_count) && 2474 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2475 grp = i; 2476 2477 block_bitmap = ext4_block_bitmap(sb, gdp); 2478 if (block_bitmap == sb_block) { 2479 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2480 "Block bitmap for group %u overlaps " 2481 "superblock", i); 2482 if (!sb_rdonly(sb)) 2483 return 0; 2484 } 2485 if (block_bitmap >= sb_block + 1 && 2486 block_bitmap <= last_bg_block) { 2487 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2488 "Block bitmap for group %u overlaps " 2489 "block group descriptors", i); 2490 if (!sb_rdonly(sb)) 2491 return 0; 2492 } 2493 if (block_bitmap < first_block || block_bitmap > last_block) { 2494 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2495 "Block bitmap for group %u not in group " 2496 "(block %llu)!", i, block_bitmap); 2497 return 0; 2498 } 2499 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2500 if (inode_bitmap == sb_block) { 2501 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2502 "Inode bitmap for group %u overlaps " 2503 "superblock", i); 2504 if (!sb_rdonly(sb)) 2505 return 0; 2506 } 2507 if (inode_bitmap >= sb_block + 1 && 2508 inode_bitmap <= last_bg_block) { 2509 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2510 "Inode bitmap for group %u overlaps " 2511 "block group descriptors", i); 2512 if (!sb_rdonly(sb)) 2513 return 0; 2514 } 2515 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2516 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2517 "Inode bitmap for group %u not in group " 2518 "(block %llu)!", i, inode_bitmap); 2519 return 0; 2520 } 2521 inode_table = ext4_inode_table(sb, gdp); 2522 if (inode_table == sb_block) { 2523 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2524 "Inode table for group %u overlaps " 2525 "superblock", i); 2526 if (!sb_rdonly(sb)) 2527 return 0; 2528 } 2529 if (inode_table >= sb_block + 1 && 2530 inode_table <= last_bg_block) { 2531 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2532 "Inode table for group %u overlaps " 2533 "block group descriptors", i); 2534 if (!sb_rdonly(sb)) 2535 return 0; 2536 } 2537 if (inode_table < first_block || 2538 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2539 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2540 "Inode table for group %u not in group " 2541 "(block %llu)!", i, inode_table); 2542 return 0; 2543 } 2544 ext4_lock_group(sb, i); 2545 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2546 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2547 "Checksum for group %u failed (%u!=%u)", 2548 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2549 gdp)), le16_to_cpu(gdp->bg_checksum)); 2550 if (!sb_rdonly(sb)) { 2551 ext4_unlock_group(sb, i); 2552 return 0; 2553 } 2554 } 2555 ext4_unlock_group(sb, i); 2556 if (!flexbg_flag) 2557 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2558 } 2559 if (NULL != first_not_zeroed) 2560 *first_not_zeroed = grp; 2561 return 1; 2562 } 2563 2564 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2565 * the superblock) which were deleted from all directories, but held open by 2566 * a process at the time of a crash. We walk the list and try to delete these 2567 * inodes at recovery time (only with a read-write filesystem). 2568 * 2569 * In order to keep the orphan inode chain consistent during traversal (in 2570 * case of crash during recovery), we link each inode into the superblock 2571 * orphan list_head and handle it the same way as an inode deletion during 2572 * normal operation (which journals the operations for us). 2573 * 2574 * We only do an iget() and an iput() on each inode, which is very safe if we 2575 * accidentally point at an in-use or already deleted inode. The worst that 2576 * can happen in this case is that we get a "bit already cleared" message from 2577 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2578 * e2fsck was run on this filesystem, and it must have already done the orphan 2579 * inode cleanup for us, so we can safely abort without any further action. 2580 */ 2581 static void ext4_orphan_cleanup(struct super_block *sb, 2582 struct ext4_super_block *es) 2583 { 2584 unsigned int s_flags = sb->s_flags; 2585 int ret, nr_orphans = 0, nr_truncates = 0; 2586 #ifdef CONFIG_QUOTA 2587 int quota_update = 0; 2588 int i; 2589 #endif 2590 if (!es->s_last_orphan) { 2591 jbd_debug(4, "no orphan inodes to clean up\n"); 2592 return; 2593 } 2594 2595 if (bdev_read_only(sb->s_bdev)) { 2596 ext4_msg(sb, KERN_ERR, "write access " 2597 "unavailable, skipping orphan cleanup"); 2598 return; 2599 } 2600 2601 /* Check if feature set would not allow a r/w mount */ 2602 if (!ext4_feature_set_ok(sb, 0)) { 2603 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2604 "unknown ROCOMPAT features"); 2605 return; 2606 } 2607 2608 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2609 /* don't clear list on RO mount w/ errors */ 2610 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 2611 ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2612 "clearing orphan list.\n"); 2613 es->s_last_orphan = 0; 2614 } 2615 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2616 return; 2617 } 2618 2619 if (s_flags & SB_RDONLY) { 2620 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2621 sb->s_flags &= ~SB_RDONLY; 2622 } 2623 #ifdef CONFIG_QUOTA 2624 /* Needed for iput() to work correctly and not trash data */ 2625 sb->s_flags |= SB_ACTIVE; 2626 2627 /* 2628 * Turn on quotas which were not enabled for read-only mounts if 2629 * filesystem has quota feature, so that they are updated correctly. 2630 */ 2631 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 2632 int ret = ext4_enable_quotas(sb); 2633 2634 if (!ret) 2635 quota_update = 1; 2636 else 2637 ext4_msg(sb, KERN_ERR, 2638 "Cannot turn on quotas: error %d", ret); 2639 } 2640 2641 /* Turn on journaled quotas used for old sytle */ 2642 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2643 if (EXT4_SB(sb)->s_qf_names[i]) { 2644 int ret = ext4_quota_on_mount(sb, i); 2645 2646 if (!ret) 2647 quota_update = 1; 2648 else 2649 ext4_msg(sb, KERN_ERR, 2650 "Cannot turn on journaled " 2651 "quota: type %d: error %d", i, ret); 2652 } 2653 } 2654 #endif 2655 2656 while (es->s_last_orphan) { 2657 struct inode *inode; 2658 2659 /* 2660 * We may have encountered an error during cleanup; if 2661 * so, skip the rest. 2662 */ 2663 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2664 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2665 es->s_last_orphan = 0; 2666 break; 2667 } 2668 2669 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2670 if (IS_ERR(inode)) { 2671 es->s_last_orphan = 0; 2672 break; 2673 } 2674 2675 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2676 dquot_initialize(inode); 2677 if (inode->i_nlink) { 2678 if (test_opt(sb, DEBUG)) 2679 ext4_msg(sb, KERN_DEBUG, 2680 "%s: truncating inode %lu to %lld bytes", 2681 __func__, inode->i_ino, inode->i_size); 2682 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2683 inode->i_ino, inode->i_size); 2684 inode_lock(inode); 2685 truncate_inode_pages(inode->i_mapping, inode->i_size); 2686 ret = ext4_truncate(inode); 2687 if (ret) 2688 ext4_std_error(inode->i_sb, ret); 2689 inode_unlock(inode); 2690 nr_truncates++; 2691 } else { 2692 if (test_opt(sb, DEBUG)) 2693 ext4_msg(sb, KERN_DEBUG, 2694 "%s: deleting unreferenced inode %lu", 2695 __func__, inode->i_ino); 2696 jbd_debug(2, "deleting unreferenced inode %lu\n", 2697 inode->i_ino); 2698 nr_orphans++; 2699 } 2700 iput(inode); /* The delete magic happens here! */ 2701 } 2702 2703 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2704 2705 if (nr_orphans) 2706 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2707 PLURAL(nr_orphans)); 2708 if (nr_truncates) 2709 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2710 PLURAL(nr_truncates)); 2711 #ifdef CONFIG_QUOTA 2712 /* Turn off quotas if they were enabled for orphan cleanup */ 2713 if (quota_update) { 2714 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2715 if (sb_dqopt(sb)->files[i]) 2716 dquot_quota_off(sb, i); 2717 } 2718 } 2719 #endif 2720 sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2721 } 2722 2723 /* 2724 * Maximal extent format file size. 2725 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2726 * extent format containers, within a sector_t, and within i_blocks 2727 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2728 * so that won't be a limiting factor. 2729 * 2730 * However there is other limiting factor. We do store extents in the form 2731 * of starting block and length, hence the resulting length of the extent 2732 * covering maximum file size must fit into on-disk format containers as 2733 * well. Given that length is always by 1 unit bigger than max unit (because 2734 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2735 * 2736 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2737 */ 2738 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2739 { 2740 loff_t res; 2741 loff_t upper_limit = MAX_LFS_FILESIZE; 2742 2743 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 2744 2745 if (!has_huge_files) { 2746 upper_limit = (1LL << 32) - 1; 2747 2748 /* total blocks in file system block size */ 2749 upper_limit >>= (blkbits - 9); 2750 upper_limit <<= blkbits; 2751 } 2752 2753 /* 2754 * 32-bit extent-start container, ee_block. We lower the maxbytes 2755 * by one fs block, so ee_len can cover the extent of maximum file 2756 * size 2757 */ 2758 res = (1LL << 32) - 1; 2759 res <<= blkbits; 2760 2761 /* Sanity check against vm- & vfs- imposed limits */ 2762 if (res > upper_limit) 2763 res = upper_limit; 2764 2765 return res; 2766 } 2767 2768 /* 2769 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2770 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2771 * We need to be 1 filesystem block less than the 2^48 sector limit. 2772 */ 2773 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2774 { 2775 loff_t res = EXT4_NDIR_BLOCKS; 2776 int meta_blocks; 2777 loff_t upper_limit; 2778 /* This is calculated to be the largest file size for a dense, block 2779 * mapped file such that the file's total number of 512-byte sectors, 2780 * including data and all indirect blocks, does not exceed (2^48 - 1). 2781 * 2782 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2783 * number of 512-byte sectors of the file. 2784 */ 2785 2786 if (!has_huge_files) { 2787 /* 2788 * !has_huge_files or implies that the inode i_block field 2789 * represents total file blocks in 2^32 512-byte sectors == 2790 * size of vfs inode i_blocks * 8 2791 */ 2792 upper_limit = (1LL << 32) - 1; 2793 2794 /* total blocks in file system block size */ 2795 upper_limit >>= (bits - 9); 2796 2797 } else { 2798 /* 2799 * We use 48 bit ext4_inode i_blocks 2800 * With EXT4_HUGE_FILE_FL set the i_blocks 2801 * represent total number of blocks in 2802 * file system block size 2803 */ 2804 upper_limit = (1LL << 48) - 1; 2805 2806 } 2807 2808 /* indirect blocks */ 2809 meta_blocks = 1; 2810 /* double indirect blocks */ 2811 meta_blocks += 1 + (1LL << (bits-2)); 2812 /* tripple indirect blocks */ 2813 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2814 2815 upper_limit -= meta_blocks; 2816 upper_limit <<= bits; 2817 2818 res += 1LL << (bits-2); 2819 res += 1LL << (2*(bits-2)); 2820 res += 1LL << (3*(bits-2)); 2821 res <<= bits; 2822 if (res > upper_limit) 2823 res = upper_limit; 2824 2825 if (res > MAX_LFS_FILESIZE) 2826 res = MAX_LFS_FILESIZE; 2827 2828 return res; 2829 } 2830 2831 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2832 ext4_fsblk_t logical_sb_block, int nr) 2833 { 2834 struct ext4_sb_info *sbi = EXT4_SB(sb); 2835 ext4_group_t bg, first_meta_bg; 2836 int has_super = 0; 2837 2838 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2839 2840 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 2841 return logical_sb_block + nr + 1; 2842 bg = sbi->s_desc_per_block * nr; 2843 if (ext4_bg_has_super(sb, bg)) 2844 has_super = 1; 2845 2846 /* 2847 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2848 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2849 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2850 * compensate. 2851 */ 2852 if (sb->s_blocksize == 1024 && nr == 0 && 2853 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 2854 has_super++; 2855 2856 return (has_super + ext4_group_first_block_no(sb, bg)); 2857 } 2858 2859 /** 2860 * ext4_get_stripe_size: Get the stripe size. 2861 * @sbi: In memory super block info 2862 * 2863 * If we have specified it via mount option, then 2864 * use the mount option value. If the value specified at mount time is 2865 * greater than the blocks per group use the super block value. 2866 * If the super block value is greater than blocks per group return 0. 2867 * Allocator needs it be less than blocks per group. 2868 * 2869 */ 2870 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2871 { 2872 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2873 unsigned long stripe_width = 2874 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2875 int ret; 2876 2877 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2878 ret = sbi->s_stripe; 2879 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 2880 ret = stripe_width; 2881 else if (stride && stride <= sbi->s_blocks_per_group) 2882 ret = stride; 2883 else 2884 ret = 0; 2885 2886 /* 2887 * If the stripe width is 1, this makes no sense and 2888 * we set it to 0 to turn off stripe handling code. 2889 */ 2890 if (ret <= 1) 2891 ret = 0; 2892 2893 return ret; 2894 } 2895 2896 /* 2897 * Check whether this filesystem can be mounted based on 2898 * the features present and the RDONLY/RDWR mount requested. 2899 * Returns 1 if this filesystem can be mounted as requested, 2900 * 0 if it cannot be. 2901 */ 2902 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2903 { 2904 if (ext4_has_unknown_ext4_incompat_features(sb)) { 2905 ext4_msg(sb, KERN_ERR, 2906 "Couldn't mount because of " 2907 "unsupported optional features (%x)", 2908 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2909 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2910 return 0; 2911 } 2912 2913 #ifndef CONFIG_UNICODE 2914 if (ext4_has_feature_casefold(sb)) { 2915 ext4_msg(sb, KERN_ERR, 2916 "Filesystem with casefold feature cannot be " 2917 "mounted without CONFIG_UNICODE"); 2918 return 0; 2919 } 2920 #endif 2921 2922 if (readonly) 2923 return 1; 2924 2925 if (ext4_has_feature_readonly(sb)) { 2926 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 2927 sb->s_flags |= SB_RDONLY; 2928 return 1; 2929 } 2930 2931 /* Check that feature set is OK for a read-write mount */ 2932 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2933 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2934 "unsupported optional features (%x)", 2935 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2936 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2937 return 0; 2938 } 2939 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2940 ext4_msg(sb, KERN_ERR, 2941 "Can't support bigalloc feature without " 2942 "extents feature\n"); 2943 return 0; 2944 } 2945 2946 #ifndef CONFIG_QUOTA 2947 if (ext4_has_feature_quota(sb) && !readonly) { 2948 ext4_msg(sb, KERN_ERR, 2949 "Filesystem with quota feature cannot be mounted RDWR " 2950 "without CONFIG_QUOTA"); 2951 return 0; 2952 } 2953 if (ext4_has_feature_project(sb) && !readonly) { 2954 ext4_msg(sb, KERN_ERR, 2955 "Filesystem with project quota feature cannot be mounted RDWR " 2956 "without CONFIG_QUOTA"); 2957 return 0; 2958 } 2959 #endif /* CONFIG_QUOTA */ 2960 return 1; 2961 } 2962 2963 /* 2964 * This function is called once a day if we have errors logged 2965 * on the file system 2966 */ 2967 static void print_daily_error_info(struct timer_list *t) 2968 { 2969 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 2970 struct super_block *sb = sbi->s_sb; 2971 struct ext4_super_block *es = sbi->s_es; 2972 2973 if (es->s_error_count) 2974 /* fsck newer than v1.41.13 is needed to clean this condition. */ 2975 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 2976 le32_to_cpu(es->s_error_count)); 2977 if (es->s_first_error_time) { 2978 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 2979 sb->s_id, 2980 ext4_get_tstamp(es, s_first_error_time), 2981 (int) sizeof(es->s_first_error_func), 2982 es->s_first_error_func, 2983 le32_to_cpu(es->s_first_error_line)); 2984 if (es->s_first_error_ino) 2985 printk(KERN_CONT ": inode %u", 2986 le32_to_cpu(es->s_first_error_ino)); 2987 if (es->s_first_error_block) 2988 printk(KERN_CONT ": block %llu", (unsigned long long) 2989 le64_to_cpu(es->s_first_error_block)); 2990 printk(KERN_CONT "\n"); 2991 } 2992 if (es->s_last_error_time) { 2993 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 2994 sb->s_id, 2995 ext4_get_tstamp(es, s_last_error_time), 2996 (int) sizeof(es->s_last_error_func), 2997 es->s_last_error_func, 2998 le32_to_cpu(es->s_last_error_line)); 2999 if (es->s_last_error_ino) 3000 printk(KERN_CONT ": inode %u", 3001 le32_to_cpu(es->s_last_error_ino)); 3002 if (es->s_last_error_block) 3003 printk(KERN_CONT ": block %llu", (unsigned long long) 3004 le64_to_cpu(es->s_last_error_block)); 3005 printk(KERN_CONT "\n"); 3006 } 3007 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3008 } 3009 3010 /* Find next suitable group and run ext4_init_inode_table */ 3011 static int ext4_run_li_request(struct ext4_li_request *elr) 3012 { 3013 struct ext4_group_desc *gdp = NULL; 3014 ext4_group_t group, ngroups; 3015 struct super_block *sb; 3016 unsigned long timeout = 0; 3017 int ret = 0; 3018 3019 sb = elr->lr_super; 3020 ngroups = EXT4_SB(sb)->s_groups_count; 3021 3022 for (group = elr->lr_next_group; group < ngroups; group++) { 3023 gdp = ext4_get_group_desc(sb, group, NULL); 3024 if (!gdp) { 3025 ret = 1; 3026 break; 3027 } 3028 3029 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3030 break; 3031 } 3032 3033 if (group >= ngroups) 3034 ret = 1; 3035 3036 if (!ret) { 3037 timeout = jiffies; 3038 ret = ext4_init_inode_table(sb, group, 3039 elr->lr_timeout ? 0 : 1); 3040 if (elr->lr_timeout == 0) { 3041 timeout = (jiffies - timeout) * 3042 elr->lr_sbi->s_li_wait_mult; 3043 elr->lr_timeout = timeout; 3044 } 3045 elr->lr_next_sched = jiffies + elr->lr_timeout; 3046 elr->lr_next_group = group + 1; 3047 } 3048 return ret; 3049 } 3050 3051 /* 3052 * Remove lr_request from the list_request and free the 3053 * request structure. Should be called with li_list_mtx held 3054 */ 3055 static void ext4_remove_li_request(struct ext4_li_request *elr) 3056 { 3057 struct ext4_sb_info *sbi; 3058 3059 if (!elr) 3060 return; 3061 3062 sbi = elr->lr_sbi; 3063 3064 list_del(&elr->lr_request); 3065 sbi->s_li_request = NULL; 3066 kfree(elr); 3067 } 3068 3069 static void ext4_unregister_li_request(struct super_block *sb) 3070 { 3071 mutex_lock(&ext4_li_mtx); 3072 if (!ext4_li_info) { 3073 mutex_unlock(&ext4_li_mtx); 3074 return; 3075 } 3076 3077 mutex_lock(&ext4_li_info->li_list_mtx); 3078 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3079 mutex_unlock(&ext4_li_info->li_list_mtx); 3080 mutex_unlock(&ext4_li_mtx); 3081 } 3082 3083 static struct task_struct *ext4_lazyinit_task; 3084 3085 /* 3086 * This is the function where ext4lazyinit thread lives. It walks 3087 * through the request list searching for next scheduled filesystem. 3088 * When such a fs is found, run the lazy initialization request 3089 * (ext4_rn_li_request) and keep track of the time spend in this 3090 * function. Based on that time we compute next schedule time of 3091 * the request. When walking through the list is complete, compute 3092 * next waking time and put itself into sleep. 3093 */ 3094 static int ext4_lazyinit_thread(void *arg) 3095 { 3096 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3097 struct list_head *pos, *n; 3098 struct ext4_li_request *elr; 3099 unsigned long next_wakeup, cur; 3100 3101 BUG_ON(NULL == eli); 3102 3103 cont_thread: 3104 while (true) { 3105 next_wakeup = MAX_JIFFY_OFFSET; 3106 3107 mutex_lock(&eli->li_list_mtx); 3108 if (list_empty(&eli->li_request_list)) { 3109 mutex_unlock(&eli->li_list_mtx); 3110 goto exit_thread; 3111 } 3112 list_for_each_safe(pos, n, &eli->li_request_list) { 3113 int err = 0; 3114 int progress = 0; 3115 elr = list_entry(pos, struct ext4_li_request, 3116 lr_request); 3117 3118 if (time_before(jiffies, elr->lr_next_sched)) { 3119 if (time_before(elr->lr_next_sched, next_wakeup)) 3120 next_wakeup = elr->lr_next_sched; 3121 continue; 3122 } 3123 if (down_read_trylock(&elr->lr_super->s_umount)) { 3124 if (sb_start_write_trylock(elr->lr_super)) { 3125 progress = 1; 3126 /* 3127 * We hold sb->s_umount, sb can not 3128 * be removed from the list, it is 3129 * now safe to drop li_list_mtx 3130 */ 3131 mutex_unlock(&eli->li_list_mtx); 3132 err = ext4_run_li_request(elr); 3133 sb_end_write(elr->lr_super); 3134 mutex_lock(&eli->li_list_mtx); 3135 n = pos->next; 3136 } 3137 up_read((&elr->lr_super->s_umount)); 3138 } 3139 /* error, remove the lazy_init job */ 3140 if (err) { 3141 ext4_remove_li_request(elr); 3142 continue; 3143 } 3144 if (!progress) { 3145 elr->lr_next_sched = jiffies + 3146 (prandom_u32() 3147 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3148 } 3149 if (time_before(elr->lr_next_sched, next_wakeup)) 3150 next_wakeup = elr->lr_next_sched; 3151 } 3152 mutex_unlock(&eli->li_list_mtx); 3153 3154 try_to_freeze(); 3155 3156 cur = jiffies; 3157 if ((time_after_eq(cur, next_wakeup)) || 3158 (MAX_JIFFY_OFFSET == next_wakeup)) { 3159 cond_resched(); 3160 continue; 3161 } 3162 3163 schedule_timeout_interruptible(next_wakeup - cur); 3164 3165 if (kthread_should_stop()) { 3166 ext4_clear_request_list(); 3167 goto exit_thread; 3168 } 3169 } 3170 3171 exit_thread: 3172 /* 3173 * It looks like the request list is empty, but we need 3174 * to check it under the li_list_mtx lock, to prevent any 3175 * additions into it, and of course we should lock ext4_li_mtx 3176 * to atomically free the list and ext4_li_info, because at 3177 * this point another ext4 filesystem could be registering 3178 * new one. 3179 */ 3180 mutex_lock(&ext4_li_mtx); 3181 mutex_lock(&eli->li_list_mtx); 3182 if (!list_empty(&eli->li_request_list)) { 3183 mutex_unlock(&eli->li_list_mtx); 3184 mutex_unlock(&ext4_li_mtx); 3185 goto cont_thread; 3186 } 3187 mutex_unlock(&eli->li_list_mtx); 3188 kfree(ext4_li_info); 3189 ext4_li_info = NULL; 3190 mutex_unlock(&ext4_li_mtx); 3191 3192 return 0; 3193 } 3194 3195 static void ext4_clear_request_list(void) 3196 { 3197 struct list_head *pos, *n; 3198 struct ext4_li_request *elr; 3199 3200 mutex_lock(&ext4_li_info->li_list_mtx); 3201 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3202 elr = list_entry(pos, struct ext4_li_request, 3203 lr_request); 3204 ext4_remove_li_request(elr); 3205 } 3206 mutex_unlock(&ext4_li_info->li_list_mtx); 3207 } 3208 3209 static int ext4_run_lazyinit_thread(void) 3210 { 3211 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3212 ext4_li_info, "ext4lazyinit"); 3213 if (IS_ERR(ext4_lazyinit_task)) { 3214 int err = PTR_ERR(ext4_lazyinit_task); 3215 ext4_clear_request_list(); 3216 kfree(ext4_li_info); 3217 ext4_li_info = NULL; 3218 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3219 "initialization thread\n", 3220 err); 3221 return err; 3222 } 3223 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3224 return 0; 3225 } 3226 3227 /* 3228 * Check whether it make sense to run itable init. thread or not. 3229 * If there is at least one uninitialized inode table, return 3230 * corresponding group number, else the loop goes through all 3231 * groups and return total number of groups. 3232 */ 3233 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3234 { 3235 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3236 struct ext4_group_desc *gdp = NULL; 3237 3238 if (!ext4_has_group_desc_csum(sb)) 3239 return ngroups; 3240 3241 for (group = 0; group < ngroups; group++) { 3242 gdp = ext4_get_group_desc(sb, group, NULL); 3243 if (!gdp) 3244 continue; 3245 3246 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3247 break; 3248 } 3249 3250 return group; 3251 } 3252 3253 static int ext4_li_info_new(void) 3254 { 3255 struct ext4_lazy_init *eli = NULL; 3256 3257 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3258 if (!eli) 3259 return -ENOMEM; 3260 3261 INIT_LIST_HEAD(&eli->li_request_list); 3262 mutex_init(&eli->li_list_mtx); 3263 3264 eli->li_state |= EXT4_LAZYINIT_QUIT; 3265 3266 ext4_li_info = eli; 3267 3268 return 0; 3269 } 3270 3271 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3272 ext4_group_t start) 3273 { 3274 struct ext4_sb_info *sbi = EXT4_SB(sb); 3275 struct ext4_li_request *elr; 3276 3277 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3278 if (!elr) 3279 return NULL; 3280 3281 elr->lr_super = sb; 3282 elr->lr_sbi = sbi; 3283 elr->lr_next_group = start; 3284 3285 /* 3286 * Randomize first schedule time of the request to 3287 * spread the inode table initialization requests 3288 * better. 3289 */ 3290 elr->lr_next_sched = jiffies + (prandom_u32() % 3291 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3292 return elr; 3293 } 3294 3295 int ext4_register_li_request(struct super_block *sb, 3296 ext4_group_t first_not_zeroed) 3297 { 3298 struct ext4_sb_info *sbi = EXT4_SB(sb); 3299 struct ext4_li_request *elr = NULL; 3300 ext4_group_t ngroups = sbi->s_groups_count; 3301 int ret = 0; 3302 3303 mutex_lock(&ext4_li_mtx); 3304 if (sbi->s_li_request != NULL) { 3305 /* 3306 * Reset timeout so it can be computed again, because 3307 * s_li_wait_mult might have changed. 3308 */ 3309 sbi->s_li_request->lr_timeout = 0; 3310 goto out; 3311 } 3312 3313 if (first_not_zeroed == ngroups || sb_rdonly(sb) || 3314 !test_opt(sb, INIT_INODE_TABLE)) 3315 goto out; 3316 3317 elr = ext4_li_request_new(sb, first_not_zeroed); 3318 if (!elr) { 3319 ret = -ENOMEM; 3320 goto out; 3321 } 3322 3323 if (NULL == ext4_li_info) { 3324 ret = ext4_li_info_new(); 3325 if (ret) 3326 goto out; 3327 } 3328 3329 mutex_lock(&ext4_li_info->li_list_mtx); 3330 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3331 mutex_unlock(&ext4_li_info->li_list_mtx); 3332 3333 sbi->s_li_request = elr; 3334 /* 3335 * set elr to NULL here since it has been inserted to 3336 * the request_list and the removal and free of it is 3337 * handled by ext4_clear_request_list from now on. 3338 */ 3339 elr = NULL; 3340 3341 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3342 ret = ext4_run_lazyinit_thread(); 3343 if (ret) 3344 goto out; 3345 } 3346 out: 3347 mutex_unlock(&ext4_li_mtx); 3348 if (ret) 3349 kfree(elr); 3350 return ret; 3351 } 3352 3353 /* 3354 * We do not need to lock anything since this is called on 3355 * module unload. 3356 */ 3357 static void ext4_destroy_lazyinit_thread(void) 3358 { 3359 /* 3360 * If thread exited earlier 3361 * there's nothing to be done. 3362 */ 3363 if (!ext4_li_info || !ext4_lazyinit_task) 3364 return; 3365 3366 kthread_stop(ext4_lazyinit_task); 3367 } 3368 3369 static int set_journal_csum_feature_set(struct super_block *sb) 3370 { 3371 int ret = 1; 3372 int compat, incompat; 3373 struct ext4_sb_info *sbi = EXT4_SB(sb); 3374 3375 if (ext4_has_metadata_csum(sb)) { 3376 /* journal checksum v3 */ 3377 compat = 0; 3378 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3379 } else { 3380 /* journal checksum v1 */ 3381 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3382 incompat = 0; 3383 } 3384 3385 jbd2_journal_clear_features(sbi->s_journal, 3386 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3387 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3388 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3389 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3390 ret = jbd2_journal_set_features(sbi->s_journal, 3391 compat, 0, 3392 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3393 incompat); 3394 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3395 ret = jbd2_journal_set_features(sbi->s_journal, 3396 compat, 0, 3397 incompat); 3398 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3399 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3400 } else { 3401 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3402 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3403 } 3404 3405 return ret; 3406 } 3407 3408 /* 3409 * Note: calculating the overhead so we can be compatible with 3410 * historical BSD practice is quite difficult in the face of 3411 * clusters/bigalloc. This is because multiple metadata blocks from 3412 * different block group can end up in the same allocation cluster. 3413 * Calculating the exact overhead in the face of clustered allocation 3414 * requires either O(all block bitmaps) in memory or O(number of block 3415 * groups**2) in time. We will still calculate the superblock for 3416 * older file systems --- and if we come across with a bigalloc file 3417 * system with zero in s_overhead_clusters the estimate will be close to 3418 * correct especially for very large cluster sizes --- but for newer 3419 * file systems, it's better to calculate this figure once at mkfs 3420 * time, and store it in the superblock. If the superblock value is 3421 * present (even for non-bigalloc file systems), we will use it. 3422 */ 3423 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3424 char *buf) 3425 { 3426 struct ext4_sb_info *sbi = EXT4_SB(sb); 3427 struct ext4_group_desc *gdp; 3428 ext4_fsblk_t first_block, last_block, b; 3429 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3430 int s, j, count = 0; 3431 3432 if (!ext4_has_feature_bigalloc(sb)) 3433 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3434 sbi->s_itb_per_group + 2); 3435 3436 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3437 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3438 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3439 for (i = 0; i < ngroups; i++) { 3440 gdp = ext4_get_group_desc(sb, i, NULL); 3441 b = ext4_block_bitmap(sb, gdp); 3442 if (b >= first_block && b <= last_block) { 3443 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3444 count++; 3445 } 3446 b = ext4_inode_bitmap(sb, gdp); 3447 if (b >= first_block && b <= last_block) { 3448 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3449 count++; 3450 } 3451 b = ext4_inode_table(sb, gdp); 3452 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3453 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3454 int c = EXT4_B2C(sbi, b - first_block); 3455 ext4_set_bit(c, buf); 3456 count++; 3457 } 3458 if (i != grp) 3459 continue; 3460 s = 0; 3461 if (ext4_bg_has_super(sb, grp)) { 3462 ext4_set_bit(s++, buf); 3463 count++; 3464 } 3465 j = ext4_bg_num_gdb(sb, grp); 3466 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3467 ext4_error(sb, "Invalid number of block group " 3468 "descriptor blocks: %d", j); 3469 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3470 } 3471 count += j; 3472 for (; j > 0; j--) 3473 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3474 } 3475 if (!count) 3476 return 0; 3477 return EXT4_CLUSTERS_PER_GROUP(sb) - 3478 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3479 } 3480 3481 /* 3482 * Compute the overhead and stash it in sbi->s_overhead 3483 */ 3484 int ext4_calculate_overhead(struct super_block *sb) 3485 { 3486 struct ext4_sb_info *sbi = EXT4_SB(sb); 3487 struct ext4_super_block *es = sbi->s_es; 3488 struct inode *j_inode; 3489 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3490 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3491 ext4_fsblk_t overhead = 0; 3492 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3493 3494 if (!buf) 3495 return -ENOMEM; 3496 3497 /* 3498 * Compute the overhead (FS structures). This is constant 3499 * for a given filesystem unless the number of block groups 3500 * changes so we cache the previous value until it does. 3501 */ 3502 3503 /* 3504 * All of the blocks before first_data_block are overhead 3505 */ 3506 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3507 3508 /* 3509 * Add the overhead found in each block group 3510 */ 3511 for (i = 0; i < ngroups; i++) { 3512 int blks; 3513 3514 blks = count_overhead(sb, i, buf); 3515 overhead += blks; 3516 if (blks) 3517 memset(buf, 0, PAGE_SIZE); 3518 cond_resched(); 3519 } 3520 3521 /* 3522 * Add the internal journal blocks whether the journal has been 3523 * loaded or not 3524 */ 3525 if (sbi->s_journal && !sbi->journal_bdev) 3526 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3527 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { 3528 j_inode = ext4_get_journal_inode(sb, j_inum); 3529 if (j_inode) { 3530 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3531 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3532 iput(j_inode); 3533 } else { 3534 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3535 } 3536 } 3537 sbi->s_overhead = overhead; 3538 smp_wmb(); 3539 free_page((unsigned long) buf); 3540 return 0; 3541 } 3542 3543 static void ext4_clamp_want_extra_isize(struct super_block *sb) 3544 { 3545 struct ext4_sb_info *sbi = EXT4_SB(sb); 3546 struct ext4_super_block *es = sbi->s_es; 3547 3548 /* determine the minimum size of new large inodes, if present */ 3549 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE && 3550 sbi->s_want_extra_isize == 0) { 3551 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3552 EXT4_GOOD_OLD_INODE_SIZE; 3553 if (ext4_has_feature_extra_isize(sb)) { 3554 if (sbi->s_want_extra_isize < 3555 le16_to_cpu(es->s_want_extra_isize)) 3556 sbi->s_want_extra_isize = 3557 le16_to_cpu(es->s_want_extra_isize); 3558 if (sbi->s_want_extra_isize < 3559 le16_to_cpu(es->s_min_extra_isize)) 3560 sbi->s_want_extra_isize = 3561 le16_to_cpu(es->s_min_extra_isize); 3562 } 3563 } 3564 /* Check if enough inode space is available */ 3565 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 3566 sbi->s_inode_size) { 3567 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3568 EXT4_GOOD_OLD_INODE_SIZE; 3569 ext4_msg(sb, KERN_INFO, 3570 "required extra inode space not available"); 3571 } 3572 } 3573 3574 static void ext4_set_resv_clusters(struct super_block *sb) 3575 { 3576 ext4_fsblk_t resv_clusters; 3577 struct ext4_sb_info *sbi = EXT4_SB(sb); 3578 3579 /* 3580 * There's no need to reserve anything when we aren't using extents. 3581 * The space estimates are exact, there are no unwritten extents, 3582 * hole punching doesn't need new metadata... This is needed especially 3583 * to keep ext2/3 backward compatibility. 3584 */ 3585 if (!ext4_has_feature_extents(sb)) 3586 return; 3587 /* 3588 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3589 * This should cover the situations where we can not afford to run 3590 * out of space like for example punch hole, or converting 3591 * unwritten extents in delalloc path. In most cases such 3592 * allocation would require 1, or 2 blocks, higher numbers are 3593 * very rare. 3594 */ 3595 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3596 sbi->s_cluster_bits); 3597 3598 do_div(resv_clusters, 50); 3599 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3600 3601 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3602 } 3603 3604 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3605 { 3606 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3607 char *orig_data = kstrdup(data, GFP_KERNEL); 3608 struct buffer_head *bh; 3609 struct ext4_super_block *es = NULL; 3610 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3611 ext4_fsblk_t block; 3612 ext4_fsblk_t sb_block = get_sb_block(&data); 3613 ext4_fsblk_t logical_sb_block; 3614 unsigned long offset = 0; 3615 unsigned long journal_devnum = 0; 3616 unsigned long def_mount_opts; 3617 struct inode *root; 3618 const char *descr; 3619 int ret = -ENOMEM; 3620 int blocksize, clustersize; 3621 unsigned int db_count; 3622 unsigned int i; 3623 int needs_recovery, has_huge_files, has_bigalloc; 3624 __u64 blocks_count; 3625 int err = 0; 3626 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3627 ext4_group_t first_not_zeroed; 3628 3629 if ((data && !orig_data) || !sbi) 3630 goto out_free_base; 3631 3632 sbi->s_daxdev = dax_dev; 3633 sbi->s_blockgroup_lock = 3634 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3635 if (!sbi->s_blockgroup_lock) 3636 goto out_free_base; 3637 3638 sb->s_fs_info = sbi; 3639 sbi->s_sb = sb; 3640 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3641 sbi->s_sb_block = sb_block; 3642 if (sb->s_bdev->bd_part) 3643 sbi->s_sectors_written_start = 3644 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3645 3646 /* Cleanup superblock name */ 3647 strreplace(sb->s_id, '/', '!'); 3648 3649 /* -EINVAL is default */ 3650 ret = -EINVAL; 3651 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3652 if (!blocksize) { 3653 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3654 goto out_fail; 3655 } 3656 3657 /* 3658 * The ext4 superblock will not be buffer aligned for other than 1kB 3659 * block sizes. We need to calculate the offset from buffer start. 3660 */ 3661 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3662 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3663 offset = do_div(logical_sb_block, blocksize); 3664 } else { 3665 logical_sb_block = sb_block; 3666 } 3667 3668 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3669 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3670 goto out_fail; 3671 } 3672 /* 3673 * Note: s_es must be initialized as soon as possible because 3674 * some ext4 macro-instructions depend on its value 3675 */ 3676 es = (struct ext4_super_block *) (bh->b_data + offset); 3677 sbi->s_es = es; 3678 sb->s_magic = le16_to_cpu(es->s_magic); 3679 if (sb->s_magic != EXT4_SUPER_MAGIC) 3680 goto cantfind_ext4; 3681 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3682 3683 /* Warn if metadata_csum and gdt_csum are both set. */ 3684 if (ext4_has_feature_metadata_csum(sb) && 3685 ext4_has_feature_gdt_csum(sb)) 3686 ext4_warning(sb, "metadata_csum and uninit_bg are " 3687 "redundant flags; please run fsck."); 3688 3689 /* Check for a known checksum algorithm */ 3690 if (!ext4_verify_csum_type(sb, es)) { 3691 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3692 "unknown checksum algorithm."); 3693 silent = 1; 3694 goto cantfind_ext4; 3695 } 3696 3697 /* Load the checksum driver */ 3698 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3699 if (IS_ERR(sbi->s_chksum_driver)) { 3700 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3701 ret = PTR_ERR(sbi->s_chksum_driver); 3702 sbi->s_chksum_driver = NULL; 3703 goto failed_mount; 3704 } 3705 3706 /* Check superblock checksum */ 3707 if (!ext4_superblock_csum_verify(sb, es)) { 3708 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3709 "invalid superblock checksum. Run e2fsck?"); 3710 silent = 1; 3711 ret = -EFSBADCRC; 3712 goto cantfind_ext4; 3713 } 3714 3715 /* Precompute checksum seed for all metadata */ 3716 if (ext4_has_feature_csum_seed(sb)) 3717 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3718 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3719 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3720 sizeof(es->s_uuid)); 3721 3722 /* Set defaults before we parse the mount options */ 3723 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3724 set_opt(sb, INIT_INODE_TABLE); 3725 if (def_mount_opts & EXT4_DEFM_DEBUG) 3726 set_opt(sb, DEBUG); 3727 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3728 set_opt(sb, GRPID); 3729 if (def_mount_opts & EXT4_DEFM_UID16) 3730 set_opt(sb, NO_UID32); 3731 /* xattr user namespace & acls are now defaulted on */ 3732 set_opt(sb, XATTR_USER); 3733 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3734 set_opt(sb, POSIX_ACL); 3735 #endif 3736 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 3737 if (ext4_has_metadata_csum(sb)) 3738 set_opt(sb, JOURNAL_CHECKSUM); 3739 3740 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3741 set_opt(sb, JOURNAL_DATA); 3742 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3743 set_opt(sb, ORDERED_DATA); 3744 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3745 set_opt(sb, WRITEBACK_DATA); 3746 3747 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3748 set_opt(sb, ERRORS_PANIC); 3749 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3750 set_opt(sb, ERRORS_CONT); 3751 else 3752 set_opt(sb, ERRORS_RO); 3753 /* block_validity enabled by default; disable with noblock_validity */ 3754 set_opt(sb, BLOCK_VALIDITY); 3755 if (def_mount_opts & EXT4_DEFM_DISCARD) 3756 set_opt(sb, DISCARD); 3757 3758 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3759 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3760 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3761 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3762 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3763 3764 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3765 set_opt(sb, BARRIER); 3766 3767 /* 3768 * enable delayed allocation by default 3769 * Use -o nodelalloc to turn it off 3770 */ 3771 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3772 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3773 set_opt(sb, DELALLOC); 3774 3775 /* 3776 * set default s_li_wait_mult for lazyinit, for the case there is 3777 * no mount option specified. 3778 */ 3779 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3780 3781 if (sbi->s_es->s_mount_opts[0]) { 3782 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 3783 sizeof(sbi->s_es->s_mount_opts), 3784 GFP_KERNEL); 3785 if (!s_mount_opts) 3786 goto failed_mount; 3787 if (!parse_options(s_mount_opts, sb, &journal_devnum, 3788 &journal_ioprio, 0)) { 3789 ext4_msg(sb, KERN_WARNING, 3790 "failed to parse options in superblock: %s", 3791 s_mount_opts); 3792 } 3793 kfree(s_mount_opts); 3794 } 3795 sbi->s_def_mount_opt = sbi->s_mount_opt; 3796 if (!parse_options((char *) data, sb, &journal_devnum, 3797 &journal_ioprio, 0)) 3798 goto failed_mount; 3799 3800 #ifdef CONFIG_UNICODE 3801 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 3802 const struct ext4_sb_encodings *encoding_info; 3803 struct unicode_map *encoding; 3804 __u16 encoding_flags; 3805 3806 if (ext4_has_feature_encrypt(sb)) { 3807 ext4_msg(sb, KERN_ERR, 3808 "Can't mount with encoding and encryption"); 3809 goto failed_mount; 3810 } 3811 3812 if (ext4_sb_read_encoding(es, &encoding_info, 3813 &encoding_flags)) { 3814 ext4_msg(sb, KERN_ERR, 3815 "Encoding requested by superblock is unknown"); 3816 goto failed_mount; 3817 } 3818 3819 encoding = utf8_load(encoding_info->version); 3820 if (IS_ERR(encoding)) { 3821 ext4_msg(sb, KERN_ERR, 3822 "can't mount with superblock charset: %s-%s " 3823 "not supported by the kernel. flags: 0x%x.", 3824 encoding_info->name, encoding_info->version, 3825 encoding_flags); 3826 goto failed_mount; 3827 } 3828 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 3829 "%s-%s with flags 0x%hx", encoding_info->name, 3830 encoding_info->version?:"\b", encoding_flags); 3831 3832 sbi->s_encoding = encoding; 3833 sbi->s_encoding_flags = encoding_flags; 3834 } 3835 #endif 3836 3837 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3838 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3839 "with data=journal disables delayed " 3840 "allocation and O_DIRECT support!\n"); 3841 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3842 ext4_msg(sb, KERN_ERR, "can't mount with " 3843 "both data=journal and delalloc"); 3844 goto failed_mount; 3845 } 3846 if (test_opt(sb, DIOREAD_NOLOCK)) { 3847 ext4_msg(sb, KERN_ERR, "can't mount with " 3848 "both data=journal and dioread_nolock"); 3849 goto failed_mount; 3850 } 3851 if (test_opt(sb, DAX)) { 3852 ext4_msg(sb, KERN_ERR, "can't mount with " 3853 "both data=journal and dax"); 3854 goto failed_mount; 3855 } 3856 if (ext4_has_feature_encrypt(sb)) { 3857 ext4_msg(sb, KERN_WARNING, 3858 "encrypted files will use data=ordered " 3859 "instead of data journaling mode"); 3860 } 3861 if (test_opt(sb, DELALLOC)) 3862 clear_opt(sb, DELALLOC); 3863 } else { 3864 sb->s_iflags |= SB_I_CGROUPWB; 3865 } 3866 3867 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3868 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 3869 3870 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3871 (ext4_has_compat_features(sb) || 3872 ext4_has_ro_compat_features(sb) || 3873 ext4_has_incompat_features(sb))) 3874 ext4_msg(sb, KERN_WARNING, 3875 "feature flags set on rev 0 fs, " 3876 "running e2fsck is recommended"); 3877 3878 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3879 set_opt2(sb, HURD_COMPAT); 3880 if (ext4_has_feature_64bit(sb)) { 3881 ext4_msg(sb, KERN_ERR, 3882 "The Hurd can't support 64-bit file systems"); 3883 goto failed_mount; 3884 } 3885 3886 /* 3887 * ea_inode feature uses l_i_version field which is not 3888 * available in HURD_COMPAT mode. 3889 */ 3890 if (ext4_has_feature_ea_inode(sb)) { 3891 ext4_msg(sb, KERN_ERR, 3892 "ea_inode feature is not supported for Hurd"); 3893 goto failed_mount; 3894 } 3895 } 3896 3897 if (IS_EXT2_SB(sb)) { 3898 if (ext2_feature_set_ok(sb)) 3899 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3900 "using the ext4 subsystem"); 3901 else { 3902 /* 3903 * If we're probing be silent, if this looks like 3904 * it's actually an ext[34] filesystem. 3905 */ 3906 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3907 goto failed_mount; 3908 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3909 "to feature incompatibilities"); 3910 goto failed_mount; 3911 } 3912 } 3913 3914 if (IS_EXT3_SB(sb)) { 3915 if (ext3_feature_set_ok(sb)) 3916 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3917 "using the ext4 subsystem"); 3918 else { 3919 /* 3920 * If we're probing be silent, if this looks like 3921 * it's actually an ext4 filesystem. 3922 */ 3923 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3924 goto failed_mount; 3925 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3926 "to feature incompatibilities"); 3927 goto failed_mount; 3928 } 3929 } 3930 3931 /* 3932 * Check feature flags regardless of the revision level, since we 3933 * previously didn't change the revision level when setting the flags, 3934 * so there is a chance incompat flags are set on a rev 0 filesystem. 3935 */ 3936 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 3937 goto failed_mount; 3938 3939 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3940 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3941 blocksize > EXT4_MAX_BLOCK_SIZE) { 3942 ext4_msg(sb, KERN_ERR, 3943 "Unsupported filesystem blocksize %d (%d log_block_size)", 3944 blocksize, le32_to_cpu(es->s_log_block_size)); 3945 goto failed_mount; 3946 } 3947 if (le32_to_cpu(es->s_log_block_size) > 3948 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3949 ext4_msg(sb, KERN_ERR, 3950 "Invalid log block size: %u", 3951 le32_to_cpu(es->s_log_block_size)); 3952 goto failed_mount; 3953 } 3954 if (le32_to_cpu(es->s_log_cluster_size) > 3955 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3956 ext4_msg(sb, KERN_ERR, 3957 "Invalid log cluster size: %u", 3958 le32_to_cpu(es->s_log_cluster_size)); 3959 goto failed_mount; 3960 } 3961 3962 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 3963 ext4_msg(sb, KERN_ERR, 3964 "Number of reserved GDT blocks insanely large: %d", 3965 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 3966 goto failed_mount; 3967 } 3968 3969 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 3970 if (ext4_has_feature_inline_data(sb)) { 3971 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 3972 " that may contain inline data"); 3973 goto failed_mount; 3974 } 3975 if (!bdev_dax_supported(sb->s_bdev, blocksize)) { 3976 ext4_msg(sb, KERN_ERR, 3977 "DAX unsupported by block device."); 3978 goto failed_mount; 3979 } 3980 } 3981 3982 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 3983 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 3984 es->s_encryption_level); 3985 goto failed_mount; 3986 } 3987 3988 if (sb->s_blocksize != blocksize) { 3989 /* Validate the filesystem blocksize */ 3990 if (!sb_set_blocksize(sb, blocksize)) { 3991 ext4_msg(sb, KERN_ERR, "bad block size %d", 3992 blocksize); 3993 goto failed_mount; 3994 } 3995 3996 brelse(bh); 3997 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3998 offset = do_div(logical_sb_block, blocksize); 3999 bh = sb_bread_unmovable(sb, logical_sb_block); 4000 if (!bh) { 4001 ext4_msg(sb, KERN_ERR, 4002 "Can't read superblock on 2nd try"); 4003 goto failed_mount; 4004 } 4005 es = (struct ext4_super_block *)(bh->b_data + offset); 4006 sbi->s_es = es; 4007 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4008 ext4_msg(sb, KERN_ERR, 4009 "Magic mismatch, very weird!"); 4010 goto failed_mount; 4011 } 4012 } 4013 4014 has_huge_files = ext4_has_feature_huge_file(sb); 4015 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4016 has_huge_files); 4017 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4018 4019 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4020 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4021 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4022 } else { 4023 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4024 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4025 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4026 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4027 sbi->s_first_ino); 4028 goto failed_mount; 4029 } 4030 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4031 (!is_power_of_2(sbi->s_inode_size)) || 4032 (sbi->s_inode_size > blocksize)) { 4033 ext4_msg(sb, KERN_ERR, 4034 "unsupported inode size: %d", 4035 sbi->s_inode_size); 4036 goto failed_mount; 4037 } 4038 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 4039 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 4040 } 4041 4042 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4043 if (ext4_has_feature_64bit(sb)) { 4044 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4045 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4046 !is_power_of_2(sbi->s_desc_size)) { 4047 ext4_msg(sb, KERN_ERR, 4048 "unsupported descriptor size %lu", 4049 sbi->s_desc_size); 4050 goto failed_mount; 4051 } 4052 } else 4053 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4054 4055 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4056 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4057 4058 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4059 if (sbi->s_inodes_per_block == 0) 4060 goto cantfind_ext4; 4061 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4062 sbi->s_inodes_per_group > blocksize * 8) { 4063 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4064 sbi->s_blocks_per_group); 4065 goto failed_mount; 4066 } 4067 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4068 sbi->s_inodes_per_block; 4069 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4070 sbi->s_sbh = bh; 4071 sbi->s_mount_state = le16_to_cpu(es->s_state); 4072 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4073 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4074 4075 for (i = 0; i < 4; i++) 4076 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4077 sbi->s_def_hash_version = es->s_def_hash_version; 4078 if (ext4_has_feature_dir_index(sb)) { 4079 i = le32_to_cpu(es->s_flags); 4080 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4081 sbi->s_hash_unsigned = 3; 4082 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4083 #ifdef __CHAR_UNSIGNED__ 4084 if (!sb_rdonly(sb)) 4085 es->s_flags |= 4086 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4087 sbi->s_hash_unsigned = 3; 4088 #else 4089 if (!sb_rdonly(sb)) 4090 es->s_flags |= 4091 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4092 #endif 4093 } 4094 } 4095 4096 /* Handle clustersize */ 4097 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4098 has_bigalloc = ext4_has_feature_bigalloc(sb); 4099 if (has_bigalloc) { 4100 if (clustersize < blocksize) { 4101 ext4_msg(sb, KERN_ERR, 4102 "cluster size (%d) smaller than " 4103 "block size (%d)", clustersize, blocksize); 4104 goto failed_mount; 4105 } 4106 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4107 le32_to_cpu(es->s_log_block_size); 4108 sbi->s_clusters_per_group = 4109 le32_to_cpu(es->s_clusters_per_group); 4110 if (sbi->s_clusters_per_group > blocksize * 8) { 4111 ext4_msg(sb, KERN_ERR, 4112 "#clusters per group too big: %lu", 4113 sbi->s_clusters_per_group); 4114 goto failed_mount; 4115 } 4116 if (sbi->s_blocks_per_group != 4117 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4118 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4119 "clusters per group (%lu) inconsistent", 4120 sbi->s_blocks_per_group, 4121 sbi->s_clusters_per_group); 4122 goto failed_mount; 4123 } 4124 } else { 4125 if (clustersize != blocksize) { 4126 ext4_msg(sb, KERN_ERR, 4127 "fragment/cluster size (%d) != " 4128 "block size (%d)", clustersize, blocksize); 4129 goto failed_mount; 4130 } 4131 if (sbi->s_blocks_per_group > blocksize * 8) { 4132 ext4_msg(sb, KERN_ERR, 4133 "#blocks per group too big: %lu", 4134 sbi->s_blocks_per_group); 4135 goto failed_mount; 4136 } 4137 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4138 sbi->s_cluster_bits = 0; 4139 } 4140 sbi->s_cluster_ratio = clustersize / blocksize; 4141 4142 /* Do we have standard group size of clustersize * 8 blocks ? */ 4143 if (sbi->s_blocks_per_group == clustersize << 3) 4144 set_opt2(sb, STD_GROUP_SIZE); 4145 4146 /* 4147 * Test whether we have more sectors than will fit in sector_t, 4148 * and whether the max offset is addressable by the page cache. 4149 */ 4150 err = generic_check_addressable(sb->s_blocksize_bits, 4151 ext4_blocks_count(es)); 4152 if (err) { 4153 ext4_msg(sb, KERN_ERR, "filesystem" 4154 " too large to mount safely on this system"); 4155 goto failed_mount; 4156 } 4157 4158 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4159 goto cantfind_ext4; 4160 4161 /* check blocks count against device size */ 4162 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 4163 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4164 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4165 "exceeds size of device (%llu blocks)", 4166 ext4_blocks_count(es), blocks_count); 4167 goto failed_mount; 4168 } 4169 4170 /* 4171 * It makes no sense for the first data block to be beyond the end 4172 * of the filesystem. 4173 */ 4174 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4175 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4176 "block %u is beyond end of filesystem (%llu)", 4177 le32_to_cpu(es->s_first_data_block), 4178 ext4_blocks_count(es)); 4179 goto failed_mount; 4180 } 4181 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4182 (sbi->s_cluster_ratio == 1)) { 4183 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4184 "block is 0 with a 1k block and cluster size"); 4185 goto failed_mount; 4186 } 4187 4188 blocks_count = (ext4_blocks_count(es) - 4189 le32_to_cpu(es->s_first_data_block) + 4190 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4191 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4192 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4193 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 4194 "(block count %llu, first data block %u, " 4195 "blocks per group %lu)", sbi->s_groups_count, 4196 ext4_blocks_count(es), 4197 le32_to_cpu(es->s_first_data_block), 4198 EXT4_BLOCKS_PER_GROUP(sb)); 4199 goto failed_mount; 4200 } 4201 sbi->s_groups_count = blocks_count; 4202 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4203 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4204 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4205 le32_to_cpu(es->s_inodes_count)) { 4206 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4207 le32_to_cpu(es->s_inodes_count), 4208 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4209 ret = -EINVAL; 4210 goto failed_mount; 4211 } 4212 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4213 EXT4_DESC_PER_BLOCK(sb); 4214 if (ext4_has_feature_meta_bg(sb)) { 4215 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4216 ext4_msg(sb, KERN_WARNING, 4217 "first meta block group too large: %u " 4218 "(group descriptor block count %u)", 4219 le32_to_cpu(es->s_first_meta_bg), db_count); 4220 goto failed_mount; 4221 } 4222 } 4223 sbi->s_group_desc = kvmalloc_array(db_count, 4224 sizeof(struct buffer_head *), 4225 GFP_KERNEL); 4226 if (sbi->s_group_desc == NULL) { 4227 ext4_msg(sb, KERN_ERR, "not enough memory"); 4228 ret = -ENOMEM; 4229 goto failed_mount; 4230 } 4231 4232 bgl_lock_init(sbi->s_blockgroup_lock); 4233 4234 /* Pre-read the descriptors into the buffer cache */ 4235 for (i = 0; i < db_count; i++) { 4236 block = descriptor_loc(sb, logical_sb_block, i); 4237 sb_breadahead(sb, block); 4238 } 4239 4240 for (i = 0; i < db_count; i++) { 4241 block = descriptor_loc(sb, logical_sb_block, i); 4242 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 4243 if (!sbi->s_group_desc[i]) { 4244 ext4_msg(sb, KERN_ERR, 4245 "can't read group descriptor %d", i); 4246 db_count = i; 4247 goto failed_mount2; 4248 } 4249 } 4250 sbi->s_gdb_count = db_count; 4251 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4252 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4253 ret = -EFSCORRUPTED; 4254 goto failed_mount2; 4255 } 4256 4257 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4258 4259 /* Register extent status tree shrinker */ 4260 if (ext4_es_register_shrinker(sbi)) 4261 goto failed_mount3; 4262 4263 sbi->s_stripe = ext4_get_stripe_size(sbi); 4264 sbi->s_extent_max_zeroout_kb = 32; 4265 4266 /* 4267 * set up enough so that it can read an inode 4268 */ 4269 sb->s_op = &ext4_sops; 4270 sb->s_export_op = &ext4_export_ops; 4271 sb->s_xattr = ext4_xattr_handlers; 4272 #ifdef CONFIG_FS_ENCRYPTION 4273 sb->s_cop = &ext4_cryptops; 4274 #endif 4275 #ifdef CONFIG_QUOTA 4276 sb->dq_op = &ext4_quota_operations; 4277 if (ext4_has_feature_quota(sb)) 4278 sb->s_qcop = &dquot_quotactl_sysfile_ops; 4279 else 4280 sb->s_qcop = &ext4_qctl_operations; 4281 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4282 #endif 4283 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4284 4285 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 4286 mutex_init(&sbi->s_orphan_lock); 4287 4288 sb->s_root = NULL; 4289 4290 needs_recovery = (es->s_last_orphan != 0 || 4291 ext4_has_feature_journal_needs_recovery(sb)); 4292 4293 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4294 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 4295 goto failed_mount3a; 4296 4297 /* 4298 * The first inode we look at is the journal inode. Don't try 4299 * root first: it may be modified in the journal! 4300 */ 4301 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 4302 err = ext4_load_journal(sb, es, journal_devnum); 4303 if (err) 4304 goto failed_mount3a; 4305 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4306 ext4_has_feature_journal_needs_recovery(sb)) { 4307 ext4_msg(sb, KERN_ERR, "required journal recovery " 4308 "suppressed and not mounted read-only"); 4309 goto failed_mount_wq; 4310 } else { 4311 /* Nojournal mode, all journal mount options are illegal */ 4312 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 4313 ext4_msg(sb, KERN_ERR, "can't mount with " 4314 "journal_checksum, fs mounted w/o journal"); 4315 goto failed_mount_wq; 4316 } 4317 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4318 ext4_msg(sb, KERN_ERR, "can't mount with " 4319 "journal_async_commit, fs mounted w/o journal"); 4320 goto failed_mount_wq; 4321 } 4322 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 4323 ext4_msg(sb, KERN_ERR, "can't mount with " 4324 "commit=%lu, fs mounted w/o journal", 4325 sbi->s_commit_interval / HZ); 4326 goto failed_mount_wq; 4327 } 4328 if (EXT4_MOUNT_DATA_FLAGS & 4329 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 4330 ext4_msg(sb, KERN_ERR, "can't mount with " 4331 "data=, fs mounted w/o journal"); 4332 goto failed_mount_wq; 4333 } 4334 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 4335 clear_opt(sb, JOURNAL_CHECKSUM); 4336 clear_opt(sb, DATA_FLAGS); 4337 sbi->s_journal = NULL; 4338 needs_recovery = 0; 4339 goto no_journal; 4340 } 4341 4342 if (ext4_has_feature_64bit(sb) && 4343 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4344 JBD2_FEATURE_INCOMPAT_64BIT)) { 4345 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4346 goto failed_mount_wq; 4347 } 4348 4349 if (!set_journal_csum_feature_set(sb)) { 4350 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4351 "feature set"); 4352 goto failed_mount_wq; 4353 } 4354 4355 /* We have now updated the journal if required, so we can 4356 * validate the data journaling mode. */ 4357 switch (test_opt(sb, DATA_FLAGS)) { 4358 case 0: 4359 /* No mode set, assume a default based on the journal 4360 * capabilities: ORDERED_DATA if the journal can 4361 * cope, else JOURNAL_DATA 4362 */ 4363 if (jbd2_journal_check_available_features 4364 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4365 set_opt(sb, ORDERED_DATA); 4366 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4367 } else { 4368 set_opt(sb, JOURNAL_DATA); 4369 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4370 } 4371 break; 4372 4373 case EXT4_MOUNT_ORDERED_DATA: 4374 case EXT4_MOUNT_WRITEBACK_DATA: 4375 if (!jbd2_journal_check_available_features 4376 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4377 ext4_msg(sb, KERN_ERR, "Journal does not support " 4378 "requested data journaling mode"); 4379 goto failed_mount_wq; 4380 } 4381 default: 4382 break; 4383 } 4384 4385 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4386 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4387 ext4_msg(sb, KERN_ERR, "can't mount with " 4388 "journal_async_commit in data=ordered mode"); 4389 goto failed_mount_wq; 4390 } 4391 4392 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4393 4394 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4395 4396 no_journal: 4397 if (!test_opt(sb, NO_MBCACHE)) { 4398 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 4399 if (!sbi->s_ea_block_cache) { 4400 ext4_msg(sb, KERN_ERR, 4401 "Failed to create ea_block_cache"); 4402 goto failed_mount_wq; 4403 } 4404 4405 if (ext4_has_feature_ea_inode(sb)) { 4406 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4407 if (!sbi->s_ea_inode_cache) { 4408 ext4_msg(sb, KERN_ERR, 4409 "Failed to create ea_inode_cache"); 4410 goto failed_mount_wq; 4411 } 4412 } 4413 } 4414 4415 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && 4416 (blocksize != PAGE_SIZE)) { 4417 ext4_msg(sb, KERN_ERR, 4418 "Unsupported blocksize for fs encryption"); 4419 goto failed_mount_wq; 4420 } 4421 4422 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4423 !ext4_has_feature_encrypt(sb)) { 4424 ext4_set_feature_encrypt(sb); 4425 ext4_commit_super(sb, 1); 4426 } 4427 4428 /* 4429 * Get the # of file system overhead blocks from the 4430 * superblock if present. 4431 */ 4432 if (es->s_overhead_clusters) 4433 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4434 else { 4435 err = ext4_calculate_overhead(sb); 4436 if (err) 4437 goto failed_mount_wq; 4438 } 4439 4440 /* 4441 * The maximum number of concurrent works can be high and 4442 * concurrency isn't really necessary. Limit it to 1. 4443 */ 4444 EXT4_SB(sb)->rsv_conversion_wq = 4445 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4446 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4447 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4448 ret = -ENOMEM; 4449 goto failed_mount4; 4450 } 4451 4452 /* 4453 * The jbd2_journal_load will have done any necessary log recovery, 4454 * so we can safely mount the rest of the filesystem now. 4455 */ 4456 4457 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 4458 if (IS_ERR(root)) { 4459 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4460 ret = PTR_ERR(root); 4461 root = NULL; 4462 goto failed_mount4; 4463 } 4464 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4465 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4466 iput(root); 4467 goto failed_mount4; 4468 } 4469 4470 #ifdef CONFIG_UNICODE 4471 if (sbi->s_encoding) 4472 sb->s_d_op = &ext4_dentry_ops; 4473 #endif 4474 4475 sb->s_root = d_make_root(root); 4476 if (!sb->s_root) { 4477 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4478 ret = -ENOMEM; 4479 goto failed_mount4; 4480 } 4481 4482 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4483 if (ret == -EROFS) { 4484 sb->s_flags |= SB_RDONLY; 4485 ret = 0; 4486 } else if (ret) 4487 goto failed_mount4a; 4488 4489 ext4_clamp_want_extra_isize(sb); 4490 4491 ext4_set_resv_clusters(sb); 4492 4493 err = ext4_setup_system_zone(sb); 4494 if (err) { 4495 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4496 "zone (%d)", err); 4497 goto failed_mount4a; 4498 } 4499 4500 ext4_ext_init(sb); 4501 err = ext4_mb_init(sb); 4502 if (err) { 4503 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4504 err); 4505 goto failed_mount5; 4506 } 4507 4508 block = ext4_count_free_clusters(sb); 4509 ext4_free_blocks_count_set(sbi->s_es, 4510 EXT4_C2B(sbi, block)); 4511 ext4_superblock_csum_set(sb); 4512 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4513 GFP_KERNEL); 4514 if (!err) { 4515 unsigned long freei = ext4_count_free_inodes(sb); 4516 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4517 ext4_superblock_csum_set(sb); 4518 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4519 GFP_KERNEL); 4520 } 4521 if (!err) 4522 err = percpu_counter_init(&sbi->s_dirs_counter, 4523 ext4_count_dirs(sb), GFP_KERNEL); 4524 if (!err) 4525 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4526 GFP_KERNEL); 4527 if (!err) 4528 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 4529 4530 if (err) { 4531 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4532 goto failed_mount6; 4533 } 4534 4535 if (ext4_has_feature_flex_bg(sb)) 4536 if (!ext4_fill_flex_info(sb)) { 4537 ext4_msg(sb, KERN_ERR, 4538 "unable to initialize " 4539 "flex_bg meta info!"); 4540 goto failed_mount6; 4541 } 4542 4543 err = ext4_register_li_request(sb, first_not_zeroed); 4544 if (err) 4545 goto failed_mount6; 4546 4547 err = ext4_register_sysfs(sb); 4548 if (err) 4549 goto failed_mount7; 4550 4551 #ifdef CONFIG_QUOTA 4552 /* Enable quota usage during mount. */ 4553 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 4554 err = ext4_enable_quotas(sb); 4555 if (err) 4556 goto failed_mount8; 4557 } 4558 #endif /* CONFIG_QUOTA */ 4559 4560 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4561 ext4_orphan_cleanup(sb, es); 4562 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4563 if (needs_recovery) { 4564 ext4_msg(sb, KERN_INFO, "recovery complete"); 4565 ext4_mark_recovery_complete(sb, es); 4566 } 4567 if (EXT4_SB(sb)->s_journal) { 4568 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4569 descr = " journalled data mode"; 4570 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4571 descr = " ordered data mode"; 4572 else 4573 descr = " writeback data mode"; 4574 } else 4575 descr = "out journal"; 4576 4577 if (test_opt(sb, DISCARD)) { 4578 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4579 if (!blk_queue_discard(q)) 4580 ext4_msg(sb, KERN_WARNING, 4581 "mounting with \"discard\" option, but " 4582 "the device does not support discard"); 4583 } 4584 4585 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4586 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4587 "Opts: %.*s%s%s", descr, 4588 (int) sizeof(sbi->s_es->s_mount_opts), 4589 sbi->s_es->s_mount_opts, 4590 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4591 4592 if (es->s_error_count) 4593 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4594 4595 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4596 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4597 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4598 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4599 4600 kfree(orig_data); 4601 return 0; 4602 4603 cantfind_ext4: 4604 if (!silent) 4605 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4606 goto failed_mount; 4607 4608 #ifdef CONFIG_QUOTA 4609 failed_mount8: 4610 ext4_unregister_sysfs(sb); 4611 #endif 4612 failed_mount7: 4613 ext4_unregister_li_request(sb); 4614 failed_mount6: 4615 ext4_mb_release(sb); 4616 if (sbi->s_flex_groups) 4617 kvfree(sbi->s_flex_groups); 4618 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4619 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4620 percpu_counter_destroy(&sbi->s_dirs_counter); 4621 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4622 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 4623 failed_mount5: 4624 ext4_ext_release(sb); 4625 ext4_release_system_zone(sb); 4626 failed_mount4a: 4627 dput(sb->s_root); 4628 sb->s_root = NULL; 4629 failed_mount4: 4630 ext4_msg(sb, KERN_ERR, "mount failed"); 4631 if (EXT4_SB(sb)->rsv_conversion_wq) 4632 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4633 failed_mount_wq: 4634 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4635 sbi->s_ea_inode_cache = NULL; 4636 4637 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 4638 sbi->s_ea_block_cache = NULL; 4639 4640 if (sbi->s_journal) { 4641 jbd2_journal_destroy(sbi->s_journal); 4642 sbi->s_journal = NULL; 4643 } 4644 failed_mount3a: 4645 ext4_es_unregister_shrinker(sbi); 4646 failed_mount3: 4647 del_timer_sync(&sbi->s_err_report); 4648 if (sbi->s_mmp_tsk) 4649 kthread_stop(sbi->s_mmp_tsk); 4650 failed_mount2: 4651 for (i = 0; i < db_count; i++) 4652 brelse(sbi->s_group_desc[i]); 4653 kvfree(sbi->s_group_desc); 4654 failed_mount: 4655 if (sbi->s_chksum_driver) 4656 crypto_free_shash(sbi->s_chksum_driver); 4657 4658 #ifdef CONFIG_UNICODE 4659 utf8_unload(sbi->s_encoding); 4660 #endif 4661 4662 #ifdef CONFIG_QUOTA 4663 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4664 kfree(sbi->s_qf_names[i]); 4665 #endif 4666 ext4_blkdev_remove(sbi); 4667 brelse(bh); 4668 out_fail: 4669 sb->s_fs_info = NULL; 4670 kfree(sbi->s_blockgroup_lock); 4671 out_free_base: 4672 kfree(sbi); 4673 kfree(orig_data); 4674 fs_put_dax(dax_dev); 4675 return err ? err : ret; 4676 } 4677 4678 /* 4679 * Setup any per-fs journal parameters now. We'll do this both on 4680 * initial mount, once the journal has been initialised but before we've 4681 * done any recovery; and again on any subsequent remount. 4682 */ 4683 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4684 { 4685 struct ext4_sb_info *sbi = EXT4_SB(sb); 4686 4687 journal->j_commit_interval = sbi->s_commit_interval; 4688 journal->j_min_batch_time = sbi->s_min_batch_time; 4689 journal->j_max_batch_time = sbi->s_max_batch_time; 4690 4691 write_lock(&journal->j_state_lock); 4692 if (test_opt(sb, BARRIER)) 4693 journal->j_flags |= JBD2_BARRIER; 4694 else 4695 journal->j_flags &= ~JBD2_BARRIER; 4696 if (test_opt(sb, DATA_ERR_ABORT)) 4697 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4698 else 4699 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4700 write_unlock(&journal->j_state_lock); 4701 } 4702 4703 static struct inode *ext4_get_journal_inode(struct super_block *sb, 4704 unsigned int journal_inum) 4705 { 4706 struct inode *journal_inode; 4707 4708 /* 4709 * Test for the existence of a valid inode on disk. Bad things 4710 * happen if we iget() an unused inode, as the subsequent iput() 4711 * will try to delete it. 4712 */ 4713 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 4714 if (IS_ERR(journal_inode)) { 4715 ext4_msg(sb, KERN_ERR, "no journal found"); 4716 return NULL; 4717 } 4718 if (!journal_inode->i_nlink) { 4719 make_bad_inode(journal_inode); 4720 iput(journal_inode); 4721 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4722 return NULL; 4723 } 4724 4725 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4726 journal_inode, journal_inode->i_size); 4727 if (!S_ISREG(journal_inode->i_mode)) { 4728 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4729 iput(journal_inode); 4730 return NULL; 4731 } 4732 return journal_inode; 4733 } 4734 4735 static journal_t *ext4_get_journal(struct super_block *sb, 4736 unsigned int journal_inum) 4737 { 4738 struct inode *journal_inode; 4739 journal_t *journal; 4740 4741 BUG_ON(!ext4_has_feature_journal(sb)); 4742 4743 journal_inode = ext4_get_journal_inode(sb, journal_inum); 4744 if (!journal_inode) 4745 return NULL; 4746 4747 journal = jbd2_journal_init_inode(journal_inode); 4748 if (!journal) { 4749 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4750 iput(journal_inode); 4751 return NULL; 4752 } 4753 journal->j_private = sb; 4754 ext4_init_journal_params(sb, journal); 4755 return journal; 4756 } 4757 4758 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4759 dev_t j_dev) 4760 { 4761 struct buffer_head *bh; 4762 journal_t *journal; 4763 ext4_fsblk_t start; 4764 ext4_fsblk_t len; 4765 int hblock, blocksize; 4766 ext4_fsblk_t sb_block; 4767 unsigned long offset; 4768 struct ext4_super_block *es; 4769 struct block_device *bdev; 4770 4771 BUG_ON(!ext4_has_feature_journal(sb)); 4772 4773 bdev = ext4_blkdev_get(j_dev, sb); 4774 if (bdev == NULL) 4775 return NULL; 4776 4777 blocksize = sb->s_blocksize; 4778 hblock = bdev_logical_block_size(bdev); 4779 if (blocksize < hblock) { 4780 ext4_msg(sb, KERN_ERR, 4781 "blocksize too small for journal device"); 4782 goto out_bdev; 4783 } 4784 4785 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4786 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4787 set_blocksize(bdev, blocksize); 4788 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4789 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4790 "external journal"); 4791 goto out_bdev; 4792 } 4793 4794 es = (struct ext4_super_block *) (bh->b_data + offset); 4795 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4796 !(le32_to_cpu(es->s_feature_incompat) & 4797 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4798 ext4_msg(sb, KERN_ERR, "external journal has " 4799 "bad superblock"); 4800 brelse(bh); 4801 goto out_bdev; 4802 } 4803 4804 if ((le32_to_cpu(es->s_feature_ro_compat) & 4805 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4806 es->s_checksum != ext4_superblock_csum(sb, es)) { 4807 ext4_msg(sb, KERN_ERR, "external journal has " 4808 "corrupt superblock"); 4809 brelse(bh); 4810 goto out_bdev; 4811 } 4812 4813 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4814 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4815 brelse(bh); 4816 goto out_bdev; 4817 } 4818 4819 len = ext4_blocks_count(es); 4820 start = sb_block + 1; 4821 brelse(bh); /* we're done with the superblock */ 4822 4823 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4824 start, len, blocksize); 4825 if (!journal) { 4826 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4827 goto out_bdev; 4828 } 4829 journal->j_private = sb; 4830 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4831 wait_on_buffer(journal->j_sb_buffer); 4832 if (!buffer_uptodate(journal->j_sb_buffer)) { 4833 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4834 goto out_journal; 4835 } 4836 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4837 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4838 "user (unsupported) - %d", 4839 be32_to_cpu(journal->j_superblock->s_nr_users)); 4840 goto out_journal; 4841 } 4842 EXT4_SB(sb)->journal_bdev = bdev; 4843 ext4_init_journal_params(sb, journal); 4844 return journal; 4845 4846 out_journal: 4847 jbd2_journal_destroy(journal); 4848 out_bdev: 4849 ext4_blkdev_put(bdev); 4850 return NULL; 4851 } 4852 4853 static int ext4_load_journal(struct super_block *sb, 4854 struct ext4_super_block *es, 4855 unsigned long journal_devnum) 4856 { 4857 journal_t *journal; 4858 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4859 dev_t journal_dev; 4860 int err = 0; 4861 int really_read_only; 4862 4863 BUG_ON(!ext4_has_feature_journal(sb)); 4864 4865 if (journal_devnum && 4866 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4867 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4868 "numbers have changed"); 4869 journal_dev = new_decode_dev(journal_devnum); 4870 } else 4871 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4872 4873 really_read_only = bdev_read_only(sb->s_bdev); 4874 4875 /* 4876 * Are we loading a blank journal or performing recovery after a 4877 * crash? For recovery, we need to check in advance whether we 4878 * can get read-write access to the device. 4879 */ 4880 if (ext4_has_feature_journal_needs_recovery(sb)) { 4881 if (sb_rdonly(sb)) { 4882 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4883 "required on readonly filesystem"); 4884 if (really_read_only) { 4885 ext4_msg(sb, KERN_ERR, "write access " 4886 "unavailable, cannot proceed " 4887 "(try mounting with noload)"); 4888 return -EROFS; 4889 } 4890 ext4_msg(sb, KERN_INFO, "write access will " 4891 "be enabled during recovery"); 4892 } 4893 } 4894 4895 if (journal_inum && journal_dev) { 4896 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4897 "and inode journals!"); 4898 return -EINVAL; 4899 } 4900 4901 if (journal_inum) { 4902 if (!(journal = ext4_get_journal(sb, journal_inum))) 4903 return -EINVAL; 4904 } else { 4905 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4906 return -EINVAL; 4907 } 4908 4909 if (!(journal->j_flags & JBD2_BARRIER)) 4910 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4911 4912 if (!ext4_has_feature_journal_needs_recovery(sb)) 4913 err = jbd2_journal_wipe(journal, !really_read_only); 4914 if (!err) { 4915 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4916 if (save) 4917 memcpy(save, ((char *) es) + 4918 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4919 err = jbd2_journal_load(journal); 4920 if (save) 4921 memcpy(((char *) es) + EXT4_S_ERR_START, 4922 save, EXT4_S_ERR_LEN); 4923 kfree(save); 4924 } 4925 4926 if (err) { 4927 ext4_msg(sb, KERN_ERR, "error loading journal"); 4928 jbd2_journal_destroy(journal); 4929 return err; 4930 } 4931 4932 EXT4_SB(sb)->s_journal = journal; 4933 ext4_clear_journal_err(sb, es); 4934 4935 if (!really_read_only && journal_devnum && 4936 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4937 es->s_journal_dev = cpu_to_le32(journal_devnum); 4938 4939 /* Make sure we flush the recovery flag to disk. */ 4940 ext4_commit_super(sb, 1); 4941 } 4942 4943 return 0; 4944 } 4945 4946 static int ext4_commit_super(struct super_block *sb, int sync) 4947 { 4948 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4949 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4950 int error = 0; 4951 4952 if (!sbh || block_device_ejected(sb)) 4953 return error; 4954 4955 /* 4956 * The superblock bh should be mapped, but it might not be if the 4957 * device was hot-removed. Not much we can do but fail the I/O. 4958 */ 4959 if (!buffer_mapped(sbh)) 4960 return error; 4961 4962 /* 4963 * If the file system is mounted read-only, don't update the 4964 * superblock write time. This avoids updating the superblock 4965 * write time when we are mounting the root file system 4966 * read/only but we need to replay the journal; at that point, 4967 * for people who are east of GMT and who make their clock 4968 * tick in localtime for Windows bug-for-bug compatibility, 4969 * the clock is set in the future, and this will cause e2fsck 4970 * to complain and force a full file system check. 4971 */ 4972 if (!(sb->s_flags & SB_RDONLY)) 4973 ext4_update_tstamp(es, s_wtime); 4974 if (sb->s_bdev->bd_part) 4975 es->s_kbytes_written = 4976 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4977 ((part_stat_read(sb->s_bdev->bd_part, 4978 sectors[STAT_WRITE]) - 4979 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4980 else 4981 es->s_kbytes_written = 4982 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4983 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 4984 ext4_free_blocks_count_set(es, 4985 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4986 &EXT4_SB(sb)->s_freeclusters_counter))); 4987 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 4988 es->s_free_inodes_count = 4989 cpu_to_le32(percpu_counter_sum_positive( 4990 &EXT4_SB(sb)->s_freeinodes_counter)); 4991 BUFFER_TRACE(sbh, "marking dirty"); 4992 ext4_superblock_csum_set(sb); 4993 if (sync) 4994 lock_buffer(sbh); 4995 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 4996 /* 4997 * Oh, dear. A previous attempt to write the 4998 * superblock failed. This could happen because the 4999 * USB device was yanked out. Or it could happen to 5000 * be a transient write error and maybe the block will 5001 * be remapped. Nothing we can do but to retry the 5002 * write and hope for the best. 5003 */ 5004 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5005 "superblock detected"); 5006 clear_buffer_write_io_error(sbh); 5007 set_buffer_uptodate(sbh); 5008 } 5009 mark_buffer_dirty(sbh); 5010 if (sync) { 5011 unlock_buffer(sbh); 5012 error = __sync_dirty_buffer(sbh, 5013 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5014 if (buffer_write_io_error(sbh)) { 5015 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5016 "superblock"); 5017 clear_buffer_write_io_error(sbh); 5018 set_buffer_uptodate(sbh); 5019 } 5020 } 5021 return error; 5022 } 5023 5024 /* 5025 * Have we just finished recovery? If so, and if we are mounting (or 5026 * remounting) the filesystem readonly, then we will end up with a 5027 * consistent fs on disk. Record that fact. 5028 */ 5029 static void ext4_mark_recovery_complete(struct super_block *sb, 5030 struct ext4_super_block *es) 5031 { 5032 journal_t *journal = EXT4_SB(sb)->s_journal; 5033 5034 if (!ext4_has_feature_journal(sb)) { 5035 BUG_ON(journal != NULL); 5036 return; 5037 } 5038 jbd2_journal_lock_updates(journal); 5039 if (jbd2_journal_flush(journal) < 0) 5040 goto out; 5041 5042 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5043 ext4_clear_feature_journal_needs_recovery(sb); 5044 ext4_commit_super(sb, 1); 5045 } 5046 5047 out: 5048 jbd2_journal_unlock_updates(journal); 5049 } 5050 5051 /* 5052 * If we are mounting (or read-write remounting) a filesystem whose journal 5053 * has recorded an error from a previous lifetime, move that error to the 5054 * main filesystem now. 5055 */ 5056 static void ext4_clear_journal_err(struct super_block *sb, 5057 struct ext4_super_block *es) 5058 { 5059 journal_t *journal; 5060 int j_errno; 5061 const char *errstr; 5062 5063 BUG_ON(!ext4_has_feature_journal(sb)); 5064 5065 journal = EXT4_SB(sb)->s_journal; 5066 5067 /* 5068 * Now check for any error status which may have been recorded in the 5069 * journal by a prior ext4_error() or ext4_abort() 5070 */ 5071 5072 j_errno = jbd2_journal_errno(journal); 5073 if (j_errno) { 5074 char nbuf[16]; 5075 5076 errstr = ext4_decode_error(sb, j_errno, nbuf); 5077 ext4_warning(sb, "Filesystem error recorded " 5078 "from previous mount: %s", errstr); 5079 ext4_warning(sb, "Marking fs in need of filesystem check."); 5080 5081 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5082 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5083 ext4_commit_super(sb, 1); 5084 5085 jbd2_journal_clear_err(journal); 5086 jbd2_journal_update_sb_errno(journal); 5087 } 5088 } 5089 5090 /* 5091 * Force the running and committing transactions to commit, 5092 * and wait on the commit. 5093 */ 5094 int ext4_force_commit(struct super_block *sb) 5095 { 5096 journal_t *journal; 5097 5098 if (sb_rdonly(sb)) 5099 return 0; 5100 5101 journal = EXT4_SB(sb)->s_journal; 5102 return ext4_journal_force_commit(journal); 5103 } 5104 5105 static int ext4_sync_fs(struct super_block *sb, int wait) 5106 { 5107 int ret = 0; 5108 tid_t target; 5109 bool needs_barrier = false; 5110 struct ext4_sb_info *sbi = EXT4_SB(sb); 5111 5112 if (unlikely(ext4_forced_shutdown(sbi))) 5113 return 0; 5114 5115 trace_ext4_sync_fs(sb, wait); 5116 flush_workqueue(sbi->rsv_conversion_wq); 5117 /* 5118 * Writeback quota in non-journalled quota case - journalled quota has 5119 * no dirty dquots 5120 */ 5121 dquot_writeback_dquots(sb, -1); 5122 /* 5123 * Data writeback is possible w/o journal transaction, so barrier must 5124 * being sent at the end of the function. But we can skip it if 5125 * transaction_commit will do it for us. 5126 */ 5127 if (sbi->s_journal) { 5128 target = jbd2_get_latest_transaction(sbi->s_journal); 5129 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 5130 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 5131 needs_barrier = true; 5132 5133 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5134 if (wait) 5135 ret = jbd2_log_wait_commit(sbi->s_journal, 5136 target); 5137 } 5138 } else if (wait && test_opt(sb, BARRIER)) 5139 needs_barrier = true; 5140 if (needs_barrier) { 5141 int err; 5142 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 5143 if (!ret) 5144 ret = err; 5145 } 5146 5147 return ret; 5148 } 5149 5150 /* 5151 * LVM calls this function before a (read-only) snapshot is created. This 5152 * gives us a chance to flush the journal completely and mark the fs clean. 5153 * 5154 * Note that only this function cannot bring a filesystem to be in a clean 5155 * state independently. It relies on upper layer to stop all data & metadata 5156 * modifications. 5157 */ 5158 static int ext4_freeze(struct super_block *sb) 5159 { 5160 int error = 0; 5161 journal_t *journal; 5162 5163 if (sb_rdonly(sb)) 5164 return 0; 5165 5166 journal = EXT4_SB(sb)->s_journal; 5167 5168 if (journal) { 5169 /* Now we set up the journal barrier. */ 5170 jbd2_journal_lock_updates(journal); 5171 5172 /* 5173 * Don't clear the needs_recovery flag if we failed to 5174 * flush the journal. 5175 */ 5176 error = jbd2_journal_flush(journal); 5177 if (error < 0) 5178 goto out; 5179 5180 /* Journal blocked and flushed, clear needs_recovery flag. */ 5181 ext4_clear_feature_journal_needs_recovery(sb); 5182 } 5183 5184 error = ext4_commit_super(sb, 1); 5185 out: 5186 if (journal) 5187 /* we rely on upper layer to stop further updates */ 5188 jbd2_journal_unlock_updates(journal); 5189 return error; 5190 } 5191 5192 /* 5193 * Called by LVM after the snapshot is done. We need to reset the RECOVER 5194 * flag here, even though the filesystem is not technically dirty yet. 5195 */ 5196 static int ext4_unfreeze(struct super_block *sb) 5197 { 5198 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 5199 return 0; 5200 5201 if (EXT4_SB(sb)->s_journal) { 5202 /* Reset the needs_recovery flag before the fs is unlocked. */ 5203 ext4_set_feature_journal_needs_recovery(sb); 5204 } 5205 5206 ext4_commit_super(sb, 1); 5207 return 0; 5208 } 5209 5210 /* 5211 * Structure to save mount options for ext4_remount's benefit 5212 */ 5213 struct ext4_mount_options { 5214 unsigned long s_mount_opt; 5215 unsigned long s_mount_opt2; 5216 kuid_t s_resuid; 5217 kgid_t s_resgid; 5218 unsigned long s_commit_interval; 5219 u32 s_min_batch_time, s_max_batch_time; 5220 #ifdef CONFIG_QUOTA 5221 int s_jquota_fmt; 5222 char *s_qf_names[EXT4_MAXQUOTAS]; 5223 #endif 5224 }; 5225 5226 static int ext4_remount(struct super_block *sb, int *flags, char *data) 5227 { 5228 struct ext4_super_block *es; 5229 struct ext4_sb_info *sbi = EXT4_SB(sb); 5230 unsigned long old_sb_flags; 5231 struct ext4_mount_options old_opts; 5232 int enable_quota = 0; 5233 ext4_group_t g; 5234 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5235 int err = 0; 5236 #ifdef CONFIG_QUOTA 5237 int i, j; 5238 char *to_free[EXT4_MAXQUOTAS]; 5239 #endif 5240 char *orig_data = kstrdup(data, GFP_KERNEL); 5241 5242 if (data && !orig_data) 5243 return -ENOMEM; 5244 5245 /* Store the original options */ 5246 old_sb_flags = sb->s_flags; 5247 old_opts.s_mount_opt = sbi->s_mount_opt; 5248 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5249 old_opts.s_resuid = sbi->s_resuid; 5250 old_opts.s_resgid = sbi->s_resgid; 5251 old_opts.s_commit_interval = sbi->s_commit_interval; 5252 old_opts.s_min_batch_time = sbi->s_min_batch_time; 5253 old_opts.s_max_batch_time = sbi->s_max_batch_time; 5254 #ifdef CONFIG_QUOTA 5255 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5256 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5257 if (sbi->s_qf_names[i]) { 5258 char *qf_name = get_qf_name(sb, sbi, i); 5259 5260 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 5261 if (!old_opts.s_qf_names[i]) { 5262 for (j = 0; j < i; j++) 5263 kfree(old_opts.s_qf_names[j]); 5264 kfree(orig_data); 5265 return -ENOMEM; 5266 } 5267 } else 5268 old_opts.s_qf_names[i] = NULL; 5269 #endif 5270 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5271 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5272 5273 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5274 err = -EINVAL; 5275 goto restore_opts; 5276 } 5277 5278 ext4_clamp_want_extra_isize(sb); 5279 5280 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 5281 test_opt(sb, JOURNAL_CHECKSUM)) { 5282 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 5283 "during remount not supported; ignoring"); 5284 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5285 } 5286 5287 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5288 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5289 ext4_msg(sb, KERN_ERR, "can't mount with " 5290 "both data=journal and delalloc"); 5291 err = -EINVAL; 5292 goto restore_opts; 5293 } 5294 if (test_opt(sb, DIOREAD_NOLOCK)) { 5295 ext4_msg(sb, KERN_ERR, "can't mount with " 5296 "both data=journal and dioread_nolock"); 5297 err = -EINVAL; 5298 goto restore_opts; 5299 } 5300 if (test_opt(sb, DAX)) { 5301 ext4_msg(sb, KERN_ERR, "can't mount with " 5302 "both data=journal and dax"); 5303 err = -EINVAL; 5304 goto restore_opts; 5305 } 5306 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5307 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5308 ext4_msg(sb, KERN_ERR, "can't mount with " 5309 "journal_async_commit in data=ordered mode"); 5310 err = -EINVAL; 5311 goto restore_opts; 5312 } 5313 } 5314 5315 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5316 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5317 err = -EINVAL; 5318 goto restore_opts; 5319 } 5320 5321 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 5322 ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 5323 "dax flag with busy inodes while remounting"); 5324 sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 5325 } 5326 5327 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 5328 ext4_abort(sb, "Abort forced by user"); 5329 5330 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5331 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5332 5333 es = sbi->s_es; 5334 5335 if (sbi->s_journal) { 5336 ext4_init_journal_params(sb, sbi->s_journal); 5337 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5338 } 5339 5340 if (*flags & SB_LAZYTIME) 5341 sb->s_flags |= SB_LAZYTIME; 5342 5343 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 5344 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5345 err = -EROFS; 5346 goto restore_opts; 5347 } 5348 5349 if (*flags & SB_RDONLY) { 5350 err = sync_filesystem(sb); 5351 if (err < 0) 5352 goto restore_opts; 5353 err = dquot_suspend(sb, -1); 5354 if (err < 0) 5355 goto restore_opts; 5356 5357 /* 5358 * First of all, the unconditional stuff we have to do 5359 * to disable replay of the journal when we next remount 5360 */ 5361 sb->s_flags |= SB_RDONLY; 5362 5363 /* 5364 * OK, test if we are remounting a valid rw partition 5365 * readonly, and if so set the rdonly flag and then 5366 * mark the partition as valid again. 5367 */ 5368 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5369 (sbi->s_mount_state & EXT4_VALID_FS)) 5370 es->s_state = cpu_to_le16(sbi->s_mount_state); 5371 5372 if (sbi->s_journal) 5373 ext4_mark_recovery_complete(sb, es); 5374 if (sbi->s_mmp_tsk) 5375 kthread_stop(sbi->s_mmp_tsk); 5376 } else { 5377 /* Make sure we can mount this feature set readwrite */ 5378 if (ext4_has_feature_readonly(sb) || 5379 !ext4_feature_set_ok(sb, 0)) { 5380 err = -EROFS; 5381 goto restore_opts; 5382 } 5383 /* 5384 * Make sure the group descriptor checksums 5385 * are sane. If they aren't, refuse to remount r/w. 5386 */ 5387 for (g = 0; g < sbi->s_groups_count; g++) { 5388 struct ext4_group_desc *gdp = 5389 ext4_get_group_desc(sb, g, NULL); 5390 5391 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5392 ext4_msg(sb, KERN_ERR, 5393 "ext4_remount: Checksum for group %u failed (%u!=%u)", 5394 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 5395 le16_to_cpu(gdp->bg_checksum)); 5396 err = -EFSBADCRC; 5397 goto restore_opts; 5398 } 5399 } 5400 5401 /* 5402 * If we have an unprocessed orphan list hanging 5403 * around from a previously readonly bdev mount, 5404 * require a full umount/remount for now. 5405 */ 5406 if (es->s_last_orphan) { 5407 ext4_msg(sb, KERN_WARNING, "Couldn't " 5408 "remount RDWR because of unprocessed " 5409 "orphan inode list. Please " 5410 "umount/remount instead"); 5411 err = -EINVAL; 5412 goto restore_opts; 5413 } 5414 5415 /* 5416 * Mounting a RDONLY partition read-write, so reread 5417 * and store the current valid flag. (It may have 5418 * been changed by e2fsck since we originally mounted 5419 * the partition.) 5420 */ 5421 if (sbi->s_journal) 5422 ext4_clear_journal_err(sb, es); 5423 sbi->s_mount_state = le16_to_cpu(es->s_state); 5424 5425 err = ext4_setup_super(sb, es, 0); 5426 if (err) 5427 goto restore_opts; 5428 5429 sb->s_flags &= ~SB_RDONLY; 5430 if (ext4_has_feature_mmp(sb)) 5431 if (ext4_multi_mount_protect(sb, 5432 le64_to_cpu(es->s_mmp_block))) { 5433 err = -EROFS; 5434 goto restore_opts; 5435 } 5436 enable_quota = 1; 5437 } 5438 } 5439 5440 /* 5441 * Reinitialize lazy itable initialization thread based on 5442 * current settings 5443 */ 5444 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5445 ext4_unregister_li_request(sb); 5446 else { 5447 ext4_group_t first_not_zeroed; 5448 first_not_zeroed = ext4_has_uninit_itable(sb); 5449 ext4_register_li_request(sb, first_not_zeroed); 5450 } 5451 5452 ext4_setup_system_zone(sb); 5453 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5454 err = ext4_commit_super(sb, 1); 5455 if (err) 5456 goto restore_opts; 5457 } 5458 5459 #ifdef CONFIG_QUOTA 5460 /* Release old quota file names */ 5461 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5462 kfree(old_opts.s_qf_names[i]); 5463 if (enable_quota) { 5464 if (sb_any_quota_suspended(sb)) 5465 dquot_resume(sb, -1); 5466 else if (ext4_has_feature_quota(sb)) { 5467 err = ext4_enable_quotas(sb); 5468 if (err) 5469 goto restore_opts; 5470 } 5471 } 5472 #endif 5473 5474 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 5475 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5476 kfree(orig_data); 5477 return 0; 5478 5479 restore_opts: 5480 sb->s_flags = old_sb_flags; 5481 sbi->s_mount_opt = old_opts.s_mount_opt; 5482 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5483 sbi->s_resuid = old_opts.s_resuid; 5484 sbi->s_resgid = old_opts.s_resgid; 5485 sbi->s_commit_interval = old_opts.s_commit_interval; 5486 sbi->s_min_batch_time = old_opts.s_min_batch_time; 5487 sbi->s_max_batch_time = old_opts.s_max_batch_time; 5488 #ifdef CONFIG_QUOTA 5489 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5490 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 5491 to_free[i] = get_qf_name(sb, sbi, i); 5492 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5493 } 5494 synchronize_rcu(); 5495 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5496 kfree(to_free[i]); 5497 #endif 5498 kfree(orig_data); 5499 return err; 5500 } 5501 5502 #ifdef CONFIG_QUOTA 5503 static int ext4_statfs_project(struct super_block *sb, 5504 kprojid_t projid, struct kstatfs *buf) 5505 { 5506 struct kqid qid; 5507 struct dquot *dquot; 5508 u64 limit; 5509 u64 curblock; 5510 5511 qid = make_kqid_projid(projid); 5512 dquot = dqget(sb, qid); 5513 if (IS_ERR(dquot)) 5514 return PTR_ERR(dquot); 5515 spin_lock(&dquot->dq_dqb_lock); 5516 5517 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 5518 dquot->dq_dqb.dqb_bsoftlimit : 5519 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 5520 if (limit && buf->f_blocks > limit) { 5521 curblock = (dquot->dq_dqb.dqb_curspace + 5522 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5523 buf->f_blocks = limit; 5524 buf->f_bfree = buf->f_bavail = 5525 (buf->f_blocks > curblock) ? 5526 (buf->f_blocks - curblock) : 0; 5527 } 5528 5529 limit = dquot->dq_dqb.dqb_isoftlimit ? 5530 dquot->dq_dqb.dqb_isoftlimit : 5531 dquot->dq_dqb.dqb_ihardlimit; 5532 if (limit && buf->f_files > limit) { 5533 buf->f_files = limit; 5534 buf->f_ffree = 5535 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5536 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5537 } 5538 5539 spin_unlock(&dquot->dq_dqb_lock); 5540 dqput(dquot); 5541 return 0; 5542 } 5543 #endif 5544 5545 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5546 { 5547 struct super_block *sb = dentry->d_sb; 5548 struct ext4_sb_info *sbi = EXT4_SB(sb); 5549 struct ext4_super_block *es = sbi->s_es; 5550 ext4_fsblk_t overhead = 0, resv_blocks; 5551 u64 fsid; 5552 s64 bfree; 5553 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5554 5555 if (!test_opt(sb, MINIX_DF)) 5556 overhead = sbi->s_overhead; 5557 5558 buf->f_type = EXT4_SUPER_MAGIC; 5559 buf->f_bsize = sb->s_blocksize; 5560 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5561 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5562 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5563 /* prevent underflow in case that few free space is available */ 5564 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5565 buf->f_bavail = buf->f_bfree - 5566 (ext4_r_blocks_count(es) + resv_blocks); 5567 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5568 buf->f_bavail = 0; 5569 buf->f_files = le32_to_cpu(es->s_inodes_count); 5570 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5571 buf->f_namelen = EXT4_NAME_LEN; 5572 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5573 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5574 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5575 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5576 5577 #ifdef CONFIG_QUOTA 5578 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5579 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5580 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5581 #endif 5582 return 0; 5583 } 5584 5585 5586 #ifdef CONFIG_QUOTA 5587 5588 /* 5589 * Helper functions so that transaction is started before we acquire dqio_sem 5590 * to keep correct lock ordering of transaction > dqio_sem 5591 */ 5592 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5593 { 5594 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5595 } 5596 5597 static int ext4_write_dquot(struct dquot *dquot) 5598 { 5599 int ret, err; 5600 handle_t *handle; 5601 struct inode *inode; 5602 5603 inode = dquot_to_inode(dquot); 5604 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5605 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5606 if (IS_ERR(handle)) 5607 return PTR_ERR(handle); 5608 ret = dquot_commit(dquot); 5609 err = ext4_journal_stop(handle); 5610 if (!ret) 5611 ret = err; 5612 return ret; 5613 } 5614 5615 static int ext4_acquire_dquot(struct dquot *dquot) 5616 { 5617 int ret, err; 5618 handle_t *handle; 5619 5620 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5621 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5622 if (IS_ERR(handle)) 5623 return PTR_ERR(handle); 5624 ret = dquot_acquire(dquot); 5625 err = ext4_journal_stop(handle); 5626 if (!ret) 5627 ret = err; 5628 return ret; 5629 } 5630 5631 static int ext4_release_dquot(struct dquot *dquot) 5632 { 5633 int ret, err; 5634 handle_t *handle; 5635 5636 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5637 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5638 if (IS_ERR(handle)) { 5639 /* Release dquot anyway to avoid endless cycle in dqput() */ 5640 dquot_release(dquot); 5641 return PTR_ERR(handle); 5642 } 5643 ret = dquot_release(dquot); 5644 err = ext4_journal_stop(handle); 5645 if (!ret) 5646 ret = err; 5647 return ret; 5648 } 5649 5650 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5651 { 5652 struct super_block *sb = dquot->dq_sb; 5653 struct ext4_sb_info *sbi = EXT4_SB(sb); 5654 5655 /* Are we journaling quotas? */ 5656 if (ext4_has_feature_quota(sb) || 5657 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5658 dquot_mark_dquot_dirty(dquot); 5659 return ext4_write_dquot(dquot); 5660 } else { 5661 return dquot_mark_dquot_dirty(dquot); 5662 } 5663 } 5664 5665 static int ext4_write_info(struct super_block *sb, int type) 5666 { 5667 int ret, err; 5668 handle_t *handle; 5669 5670 /* Data block + inode block */ 5671 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5672 if (IS_ERR(handle)) 5673 return PTR_ERR(handle); 5674 ret = dquot_commit_info(sb, type); 5675 err = ext4_journal_stop(handle); 5676 if (!ret) 5677 ret = err; 5678 return ret; 5679 } 5680 5681 /* 5682 * Turn on quotas during mount time - we need to find 5683 * the quota file and such... 5684 */ 5685 static int ext4_quota_on_mount(struct super_block *sb, int type) 5686 { 5687 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 5688 EXT4_SB(sb)->s_jquota_fmt, type); 5689 } 5690 5691 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5692 { 5693 struct ext4_inode_info *ei = EXT4_I(inode); 5694 5695 /* The first argument of lockdep_set_subclass has to be 5696 * *exactly* the same as the argument to init_rwsem() --- in 5697 * this case, in init_once() --- or lockdep gets unhappy 5698 * because the name of the lock is set using the 5699 * stringification of the argument to init_rwsem(). 5700 */ 5701 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5702 lockdep_set_subclass(&ei->i_data_sem, subclass); 5703 } 5704 5705 /* 5706 * Standard function to be called on quota_on 5707 */ 5708 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5709 const struct path *path) 5710 { 5711 int err; 5712 5713 if (!test_opt(sb, QUOTA)) 5714 return -EINVAL; 5715 5716 /* Quotafile not on the same filesystem? */ 5717 if (path->dentry->d_sb != sb) 5718 return -EXDEV; 5719 /* Journaling quota? */ 5720 if (EXT4_SB(sb)->s_qf_names[type]) { 5721 /* Quotafile not in fs root? */ 5722 if (path->dentry->d_parent != sb->s_root) 5723 ext4_msg(sb, KERN_WARNING, 5724 "Quota file not on filesystem root. " 5725 "Journaled quota will not work"); 5726 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 5727 } else { 5728 /* 5729 * Clear the flag just in case mount options changed since 5730 * last time. 5731 */ 5732 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 5733 } 5734 5735 /* 5736 * When we journal data on quota file, we have to flush journal to see 5737 * all updates to the file when we bypass pagecache... 5738 */ 5739 if (EXT4_SB(sb)->s_journal && 5740 ext4_should_journal_data(d_inode(path->dentry))) { 5741 /* 5742 * We don't need to lock updates but journal_flush() could 5743 * otherwise be livelocked... 5744 */ 5745 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5746 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5747 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5748 if (err) 5749 return err; 5750 } 5751 5752 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5753 err = dquot_quota_on(sb, type, format_id, path); 5754 if (err) { 5755 lockdep_set_quota_inode(path->dentry->d_inode, 5756 I_DATA_SEM_NORMAL); 5757 } else { 5758 struct inode *inode = d_inode(path->dentry); 5759 handle_t *handle; 5760 5761 /* 5762 * Set inode flags to prevent userspace from messing with quota 5763 * files. If this fails, we return success anyway since quotas 5764 * are already enabled and this is not a hard failure. 5765 */ 5766 inode_lock(inode); 5767 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5768 if (IS_ERR(handle)) 5769 goto unlock_inode; 5770 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 5771 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 5772 S_NOATIME | S_IMMUTABLE); 5773 ext4_mark_inode_dirty(handle, inode); 5774 ext4_journal_stop(handle); 5775 unlock_inode: 5776 inode_unlock(inode); 5777 } 5778 return err; 5779 } 5780 5781 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5782 unsigned int flags) 5783 { 5784 int err; 5785 struct inode *qf_inode; 5786 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5787 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5788 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5789 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5790 }; 5791 5792 BUG_ON(!ext4_has_feature_quota(sb)); 5793 5794 if (!qf_inums[type]) 5795 return -EPERM; 5796 5797 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 5798 if (IS_ERR(qf_inode)) { 5799 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5800 return PTR_ERR(qf_inode); 5801 } 5802 5803 /* Don't account quota for quota files to avoid recursion */ 5804 qf_inode->i_flags |= S_NOQUOTA; 5805 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 5806 err = dquot_enable(qf_inode, type, format_id, flags); 5807 if (err) 5808 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 5809 iput(qf_inode); 5810 5811 return err; 5812 } 5813 5814 /* Enable usage tracking for all quota types. */ 5815 static int ext4_enable_quotas(struct super_block *sb) 5816 { 5817 int type, err = 0; 5818 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5819 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5820 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5821 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5822 }; 5823 bool quota_mopt[EXT4_MAXQUOTAS] = { 5824 test_opt(sb, USRQUOTA), 5825 test_opt(sb, GRPQUOTA), 5826 test_opt(sb, PRJQUOTA), 5827 }; 5828 5829 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 5830 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 5831 if (qf_inums[type]) { 5832 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5833 DQUOT_USAGE_ENABLED | 5834 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 5835 if (err) { 5836 ext4_warning(sb, 5837 "Failed to enable quota tracking " 5838 "(type=%d, err=%d). Please run " 5839 "e2fsck to fix.", type, err); 5840 for (type--; type >= 0; type--) 5841 dquot_quota_off(sb, type); 5842 5843 return err; 5844 } 5845 } 5846 } 5847 return 0; 5848 } 5849 5850 static int ext4_quota_off(struct super_block *sb, int type) 5851 { 5852 struct inode *inode = sb_dqopt(sb)->files[type]; 5853 handle_t *handle; 5854 int err; 5855 5856 /* Force all delayed allocation blocks to be allocated. 5857 * Caller already holds s_umount sem */ 5858 if (test_opt(sb, DELALLOC)) 5859 sync_filesystem(sb); 5860 5861 if (!inode || !igrab(inode)) 5862 goto out; 5863 5864 err = dquot_quota_off(sb, type); 5865 if (err || ext4_has_feature_quota(sb)) 5866 goto out_put; 5867 5868 inode_lock(inode); 5869 /* 5870 * Update modification times of quota files when userspace can 5871 * start looking at them. If we fail, we return success anyway since 5872 * this is not a hard failure and quotas are already disabled. 5873 */ 5874 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5875 if (IS_ERR(handle)) 5876 goto out_unlock; 5877 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 5878 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 5879 inode->i_mtime = inode->i_ctime = current_time(inode); 5880 ext4_mark_inode_dirty(handle, inode); 5881 ext4_journal_stop(handle); 5882 out_unlock: 5883 inode_unlock(inode); 5884 out_put: 5885 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 5886 iput(inode); 5887 return err; 5888 out: 5889 return dquot_quota_off(sb, type); 5890 } 5891 5892 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5893 * acquiring the locks... As quota files are never truncated and quota code 5894 * itself serializes the operations (and no one else should touch the files) 5895 * we don't have to be afraid of races */ 5896 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5897 size_t len, loff_t off) 5898 { 5899 struct inode *inode = sb_dqopt(sb)->files[type]; 5900 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5901 int offset = off & (sb->s_blocksize - 1); 5902 int tocopy; 5903 size_t toread; 5904 struct buffer_head *bh; 5905 loff_t i_size = i_size_read(inode); 5906 5907 if (off > i_size) 5908 return 0; 5909 if (off+len > i_size) 5910 len = i_size-off; 5911 toread = len; 5912 while (toread > 0) { 5913 tocopy = sb->s_blocksize - offset < toread ? 5914 sb->s_blocksize - offset : toread; 5915 bh = ext4_bread(NULL, inode, blk, 0); 5916 if (IS_ERR(bh)) 5917 return PTR_ERR(bh); 5918 if (!bh) /* A hole? */ 5919 memset(data, 0, tocopy); 5920 else 5921 memcpy(data, bh->b_data+offset, tocopy); 5922 brelse(bh); 5923 offset = 0; 5924 toread -= tocopy; 5925 data += tocopy; 5926 blk++; 5927 } 5928 return len; 5929 } 5930 5931 /* Write to quotafile (we know the transaction is already started and has 5932 * enough credits) */ 5933 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5934 const char *data, size_t len, loff_t off) 5935 { 5936 struct inode *inode = sb_dqopt(sb)->files[type]; 5937 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5938 int err, offset = off & (sb->s_blocksize - 1); 5939 int retries = 0; 5940 struct buffer_head *bh; 5941 handle_t *handle = journal_current_handle(); 5942 5943 if (EXT4_SB(sb)->s_journal && !handle) { 5944 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5945 " cancelled because transaction is not started", 5946 (unsigned long long)off, (unsigned long long)len); 5947 return -EIO; 5948 } 5949 /* 5950 * Since we account only one data block in transaction credits, 5951 * then it is impossible to cross a block boundary. 5952 */ 5953 if (sb->s_blocksize - offset < len) { 5954 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5955 " cancelled because not block aligned", 5956 (unsigned long long)off, (unsigned long long)len); 5957 return -EIO; 5958 } 5959 5960 do { 5961 bh = ext4_bread(handle, inode, blk, 5962 EXT4_GET_BLOCKS_CREATE | 5963 EXT4_GET_BLOCKS_METADATA_NOFAIL); 5964 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 5965 ext4_should_retry_alloc(inode->i_sb, &retries)); 5966 if (IS_ERR(bh)) 5967 return PTR_ERR(bh); 5968 if (!bh) 5969 goto out; 5970 BUFFER_TRACE(bh, "get write access"); 5971 err = ext4_journal_get_write_access(handle, bh); 5972 if (err) { 5973 brelse(bh); 5974 return err; 5975 } 5976 lock_buffer(bh); 5977 memcpy(bh->b_data+offset, data, len); 5978 flush_dcache_page(bh->b_page); 5979 unlock_buffer(bh); 5980 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5981 brelse(bh); 5982 out: 5983 if (inode->i_size < off + len) { 5984 i_size_write(inode, off + len); 5985 EXT4_I(inode)->i_disksize = inode->i_size; 5986 ext4_mark_inode_dirty(handle, inode); 5987 } 5988 return len; 5989 } 5990 5991 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) 5992 { 5993 const struct quota_format_ops *ops; 5994 5995 if (!sb_has_quota_loaded(sb, qid->type)) 5996 return -ESRCH; 5997 ops = sb_dqopt(sb)->ops[qid->type]; 5998 if (!ops || !ops->get_next_id) 5999 return -ENOSYS; 6000 return dquot_get_next_id(sb, qid); 6001 } 6002 #endif 6003 6004 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6005 const char *dev_name, void *data) 6006 { 6007 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6008 } 6009 6010 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 6011 static inline void register_as_ext2(void) 6012 { 6013 int err = register_filesystem(&ext2_fs_type); 6014 if (err) 6015 printk(KERN_WARNING 6016 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 6017 } 6018 6019 static inline void unregister_as_ext2(void) 6020 { 6021 unregister_filesystem(&ext2_fs_type); 6022 } 6023 6024 static inline int ext2_feature_set_ok(struct super_block *sb) 6025 { 6026 if (ext4_has_unknown_ext2_incompat_features(sb)) 6027 return 0; 6028 if (sb_rdonly(sb)) 6029 return 1; 6030 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 6031 return 0; 6032 return 1; 6033 } 6034 #else 6035 static inline void register_as_ext2(void) { } 6036 static inline void unregister_as_ext2(void) { } 6037 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 6038 #endif 6039 6040 static inline void register_as_ext3(void) 6041 { 6042 int err = register_filesystem(&ext3_fs_type); 6043 if (err) 6044 printk(KERN_WARNING 6045 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 6046 } 6047 6048 static inline void unregister_as_ext3(void) 6049 { 6050 unregister_filesystem(&ext3_fs_type); 6051 } 6052 6053 static inline int ext3_feature_set_ok(struct super_block *sb) 6054 { 6055 if (ext4_has_unknown_ext3_incompat_features(sb)) 6056 return 0; 6057 if (!ext4_has_feature_journal(sb)) 6058 return 0; 6059 if (sb_rdonly(sb)) 6060 return 1; 6061 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 6062 return 0; 6063 return 1; 6064 } 6065 6066 static struct file_system_type ext4_fs_type = { 6067 .owner = THIS_MODULE, 6068 .name = "ext4", 6069 .mount = ext4_mount, 6070 .kill_sb = kill_block_super, 6071 .fs_flags = FS_REQUIRES_DEV, 6072 }; 6073 MODULE_ALIAS_FS("ext4"); 6074 6075 /* Shared across all ext4 file systems */ 6076 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6077 6078 static int __init ext4_init_fs(void) 6079 { 6080 int i, err; 6081 6082 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 6083 ext4_li_info = NULL; 6084 mutex_init(&ext4_li_mtx); 6085 6086 /* Build-time check for flags consistency */ 6087 ext4_check_flag_values(); 6088 6089 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6090 init_waitqueue_head(&ext4__ioend_wq[i]); 6091 6092 err = ext4_init_es(); 6093 if (err) 6094 return err; 6095 6096 err = ext4_init_pending(); 6097 if (err) 6098 goto out6; 6099 6100 err = ext4_init_pageio(); 6101 if (err) 6102 goto out5; 6103 6104 err = ext4_init_system_zone(); 6105 if (err) 6106 goto out4; 6107 6108 err = ext4_init_sysfs(); 6109 if (err) 6110 goto out3; 6111 6112 err = ext4_init_mballoc(); 6113 if (err) 6114 goto out2; 6115 err = init_inodecache(); 6116 if (err) 6117 goto out1; 6118 register_as_ext3(); 6119 register_as_ext2(); 6120 err = register_filesystem(&ext4_fs_type); 6121 if (err) 6122 goto out; 6123 6124 return 0; 6125 out: 6126 unregister_as_ext2(); 6127 unregister_as_ext3(); 6128 destroy_inodecache(); 6129 out1: 6130 ext4_exit_mballoc(); 6131 out2: 6132 ext4_exit_sysfs(); 6133 out3: 6134 ext4_exit_system_zone(); 6135 out4: 6136 ext4_exit_pageio(); 6137 out5: 6138 ext4_exit_pending(); 6139 out6: 6140 ext4_exit_es(); 6141 6142 return err; 6143 } 6144 6145 static void __exit ext4_exit_fs(void) 6146 { 6147 ext4_destroy_lazyinit_thread(); 6148 unregister_as_ext2(); 6149 unregister_as_ext3(); 6150 unregister_filesystem(&ext4_fs_type); 6151 destroy_inodecache(); 6152 ext4_exit_mballoc(); 6153 ext4_exit_sysfs(); 6154 ext4_exit_system_zone(); 6155 ext4_exit_pageio(); 6156 ext4_exit_es(); 6157 ext4_exit_pending(); 6158 } 6159 6160 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 6161 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6162 MODULE_LICENSE("GPL"); 6163 MODULE_SOFTDEP("pre: crc32c"); 6164 module_init(ext4_init_fs) 6165 module_exit(ext4_exit_fs) 6166