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) && !system_going_down()) { 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 if (!drop) 1111 drop = fscrypt_drop_inode(inode); 1112 1113 trace_ext4_drop_inode(inode, drop); 1114 return drop; 1115 } 1116 1117 static void ext4_free_in_core_inode(struct inode *inode) 1118 { 1119 fscrypt_free_inode(inode); 1120 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1121 } 1122 1123 static void ext4_destroy_inode(struct inode *inode) 1124 { 1125 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1126 ext4_msg(inode->i_sb, KERN_ERR, 1127 "Inode %lu (%p): orphan list check failed!", 1128 inode->i_ino, EXT4_I(inode)); 1129 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1130 EXT4_I(inode), sizeof(struct ext4_inode_info), 1131 true); 1132 dump_stack(); 1133 } 1134 } 1135 1136 static void init_once(void *foo) 1137 { 1138 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1139 1140 INIT_LIST_HEAD(&ei->i_orphan); 1141 init_rwsem(&ei->xattr_sem); 1142 init_rwsem(&ei->i_data_sem); 1143 init_rwsem(&ei->i_mmap_sem); 1144 inode_init_once(&ei->vfs_inode); 1145 } 1146 1147 static int __init init_inodecache(void) 1148 { 1149 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1150 sizeof(struct ext4_inode_info), 0, 1151 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1152 SLAB_ACCOUNT), 1153 offsetof(struct ext4_inode_info, i_data), 1154 sizeof_field(struct ext4_inode_info, i_data), 1155 init_once); 1156 if (ext4_inode_cachep == NULL) 1157 return -ENOMEM; 1158 return 0; 1159 } 1160 1161 static void destroy_inodecache(void) 1162 { 1163 /* 1164 * Make sure all delayed rcu free inodes are flushed before we 1165 * destroy cache. 1166 */ 1167 rcu_barrier(); 1168 kmem_cache_destroy(ext4_inode_cachep); 1169 } 1170 1171 void ext4_clear_inode(struct inode *inode) 1172 { 1173 invalidate_inode_buffers(inode); 1174 clear_inode(inode); 1175 dquot_drop(inode); 1176 ext4_discard_preallocations(inode); 1177 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1178 if (EXT4_I(inode)->jinode) { 1179 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1180 EXT4_I(inode)->jinode); 1181 jbd2_free_inode(EXT4_I(inode)->jinode); 1182 EXT4_I(inode)->jinode = NULL; 1183 } 1184 fscrypt_put_encryption_info(inode); 1185 fsverity_cleanup_inode(inode); 1186 } 1187 1188 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1189 u64 ino, u32 generation) 1190 { 1191 struct inode *inode; 1192 1193 /* 1194 * Currently we don't know the generation for parent directory, so 1195 * a generation of 0 means "accept any" 1196 */ 1197 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1198 if (IS_ERR(inode)) 1199 return ERR_CAST(inode); 1200 if (generation && inode->i_generation != generation) { 1201 iput(inode); 1202 return ERR_PTR(-ESTALE); 1203 } 1204 1205 return inode; 1206 } 1207 1208 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1209 int fh_len, int fh_type) 1210 { 1211 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1212 ext4_nfs_get_inode); 1213 } 1214 1215 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1216 int fh_len, int fh_type) 1217 { 1218 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1219 ext4_nfs_get_inode); 1220 } 1221 1222 static int ext4_nfs_commit_metadata(struct inode *inode) 1223 { 1224 struct writeback_control wbc = { 1225 .sync_mode = WB_SYNC_ALL 1226 }; 1227 1228 trace_ext4_nfs_commit_metadata(inode); 1229 return ext4_write_inode(inode, &wbc); 1230 } 1231 1232 /* 1233 * Try to release metadata pages (indirect blocks, directories) which are 1234 * mapped via the block device. Since these pages could have journal heads 1235 * which would prevent try_to_free_buffers() from freeing them, we must use 1236 * jbd2 layer's try_to_free_buffers() function to release them. 1237 */ 1238 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1239 gfp_t wait) 1240 { 1241 journal_t *journal = EXT4_SB(sb)->s_journal; 1242 1243 WARN_ON(PageChecked(page)); 1244 if (!page_has_buffers(page)) 1245 return 0; 1246 if (journal) 1247 return jbd2_journal_try_to_free_buffers(journal, page, 1248 wait & ~__GFP_DIRECT_RECLAIM); 1249 return try_to_free_buffers(page); 1250 } 1251 1252 #ifdef CONFIG_FS_ENCRYPTION 1253 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1254 { 1255 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1256 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1257 } 1258 1259 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1260 void *fs_data) 1261 { 1262 handle_t *handle = fs_data; 1263 int res, res2, credits, retries = 0; 1264 1265 /* 1266 * Encrypting the root directory is not allowed because e2fsck expects 1267 * lost+found to exist and be unencrypted, and encrypting the root 1268 * directory would imply encrypting the lost+found directory as well as 1269 * the filename "lost+found" itself. 1270 */ 1271 if (inode->i_ino == EXT4_ROOT_INO) 1272 return -EPERM; 1273 1274 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 1275 return -EINVAL; 1276 1277 res = ext4_convert_inline_data(inode); 1278 if (res) 1279 return res; 1280 1281 /* 1282 * If a journal handle was specified, then the encryption context is 1283 * being set on a new inode via inheritance and is part of a larger 1284 * transaction to create the inode. Otherwise the encryption context is 1285 * being set on an existing inode in its own transaction. Only in the 1286 * latter case should the "retry on ENOSPC" logic be used. 1287 */ 1288 1289 if (handle) { 1290 res = ext4_xattr_set_handle(handle, inode, 1291 EXT4_XATTR_INDEX_ENCRYPTION, 1292 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1293 ctx, len, 0); 1294 if (!res) { 1295 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1296 ext4_clear_inode_state(inode, 1297 EXT4_STATE_MAY_INLINE_DATA); 1298 /* 1299 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1300 * S_DAX may be disabled 1301 */ 1302 ext4_set_inode_flags(inode); 1303 } 1304 return res; 1305 } 1306 1307 res = dquot_initialize(inode); 1308 if (res) 1309 return res; 1310 retry: 1311 res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1312 &credits); 1313 if (res) 1314 return res; 1315 1316 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1317 if (IS_ERR(handle)) 1318 return PTR_ERR(handle); 1319 1320 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 1321 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1322 ctx, len, 0); 1323 if (!res) { 1324 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1325 /* 1326 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1327 * S_DAX may be disabled 1328 */ 1329 ext4_set_inode_flags(inode); 1330 res = ext4_mark_inode_dirty(handle, inode); 1331 if (res) 1332 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1333 } 1334 res2 = ext4_journal_stop(handle); 1335 1336 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 1337 goto retry; 1338 if (!res) 1339 res = res2; 1340 return res; 1341 } 1342 1343 static bool ext4_dummy_context(struct inode *inode) 1344 { 1345 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); 1346 } 1347 1348 static const struct fscrypt_operations ext4_cryptops = { 1349 .key_prefix = "ext4:", 1350 .get_context = ext4_get_context, 1351 .set_context = ext4_set_context, 1352 .dummy_context = ext4_dummy_context, 1353 .empty_dir = ext4_empty_dir, 1354 .max_namelen = EXT4_NAME_LEN, 1355 }; 1356 #endif 1357 1358 #ifdef CONFIG_QUOTA 1359 static const char * const quotatypes[] = INITQFNAMES; 1360 #define QTYPE2NAME(t) (quotatypes[t]) 1361 1362 static int ext4_write_dquot(struct dquot *dquot); 1363 static int ext4_acquire_dquot(struct dquot *dquot); 1364 static int ext4_release_dquot(struct dquot *dquot); 1365 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1366 static int ext4_write_info(struct super_block *sb, int type); 1367 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1368 const struct path *path); 1369 static int ext4_quota_on_mount(struct super_block *sb, int type); 1370 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1371 size_t len, loff_t off); 1372 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1373 const char *data, size_t len, loff_t off); 1374 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1375 unsigned int flags); 1376 static int ext4_enable_quotas(struct super_block *sb); 1377 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); 1378 1379 static struct dquot **ext4_get_dquots(struct inode *inode) 1380 { 1381 return EXT4_I(inode)->i_dquot; 1382 } 1383 1384 static const struct dquot_operations ext4_quota_operations = { 1385 .get_reserved_space = ext4_get_reserved_space, 1386 .write_dquot = ext4_write_dquot, 1387 .acquire_dquot = ext4_acquire_dquot, 1388 .release_dquot = ext4_release_dquot, 1389 .mark_dirty = ext4_mark_dquot_dirty, 1390 .write_info = ext4_write_info, 1391 .alloc_dquot = dquot_alloc, 1392 .destroy_dquot = dquot_destroy, 1393 .get_projid = ext4_get_projid, 1394 .get_inode_usage = ext4_get_inode_usage, 1395 .get_next_id = ext4_get_next_id, 1396 }; 1397 1398 static const struct quotactl_ops ext4_qctl_operations = { 1399 .quota_on = ext4_quota_on, 1400 .quota_off = ext4_quota_off, 1401 .quota_sync = dquot_quota_sync, 1402 .get_state = dquot_get_state, 1403 .set_info = dquot_set_dqinfo, 1404 .get_dqblk = dquot_get_dqblk, 1405 .set_dqblk = dquot_set_dqblk, 1406 .get_nextdqblk = dquot_get_next_dqblk, 1407 }; 1408 #endif 1409 1410 static const struct super_operations ext4_sops = { 1411 .alloc_inode = ext4_alloc_inode, 1412 .free_inode = ext4_free_in_core_inode, 1413 .destroy_inode = ext4_destroy_inode, 1414 .write_inode = ext4_write_inode, 1415 .dirty_inode = ext4_dirty_inode, 1416 .drop_inode = ext4_drop_inode, 1417 .evict_inode = ext4_evict_inode, 1418 .put_super = ext4_put_super, 1419 .sync_fs = ext4_sync_fs, 1420 .freeze_fs = ext4_freeze, 1421 .unfreeze_fs = ext4_unfreeze, 1422 .statfs = ext4_statfs, 1423 .remount_fs = ext4_remount, 1424 .show_options = ext4_show_options, 1425 #ifdef CONFIG_QUOTA 1426 .quota_read = ext4_quota_read, 1427 .quota_write = ext4_quota_write, 1428 .get_dquots = ext4_get_dquots, 1429 #endif 1430 .bdev_try_to_free_page = bdev_try_to_free_page, 1431 }; 1432 1433 static const struct export_operations ext4_export_ops = { 1434 .fh_to_dentry = ext4_fh_to_dentry, 1435 .fh_to_parent = ext4_fh_to_parent, 1436 .get_parent = ext4_get_parent, 1437 .commit_metadata = ext4_nfs_commit_metadata, 1438 }; 1439 1440 enum { 1441 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1442 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1443 Opt_nouid32, Opt_debug, Opt_removed, 1444 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1445 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1446 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1447 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1448 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1449 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1450 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1451 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1452 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1453 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, 1454 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1455 Opt_nowarn_on_error, Opt_mblk_io_submit, 1456 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 1457 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1458 Opt_inode_readahead_blks, Opt_journal_ioprio, 1459 Opt_dioread_nolock, Opt_dioread_lock, 1460 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1461 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1462 }; 1463 1464 static const match_table_t tokens = { 1465 {Opt_bsd_df, "bsddf"}, 1466 {Opt_minix_df, "minixdf"}, 1467 {Opt_grpid, "grpid"}, 1468 {Opt_grpid, "bsdgroups"}, 1469 {Opt_nogrpid, "nogrpid"}, 1470 {Opt_nogrpid, "sysvgroups"}, 1471 {Opt_resgid, "resgid=%u"}, 1472 {Opt_resuid, "resuid=%u"}, 1473 {Opt_sb, "sb=%u"}, 1474 {Opt_err_cont, "errors=continue"}, 1475 {Opt_err_panic, "errors=panic"}, 1476 {Opt_err_ro, "errors=remount-ro"}, 1477 {Opt_nouid32, "nouid32"}, 1478 {Opt_debug, "debug"}, 1479 {Opt_removed, "oldalloc"}, 1480 {Opt_removed, "orlov"}, 1481 {Opt_user_xattr, "user_xattr"}, 1482 {Opt_nouser_xattr, "nouser_xattr"}, 1483 {Opt_acl, "acl"}, 1484 {Opt_noacl, "noacl"}, 1485 {Opt_noload, "norecovery"}, 1486 {Opt_noload, "noload"}, 1487 {Opt_removed, "nobh"}, 1488 {Opt_removed, "bh"}, 1489 {Opt_commit, "commit=%u"}, 1490 {Opt_min_batch_time, "min_batch_time=%u"}, 1491 {Opt_max_batch_time, "max_batch_time=%u"}, 1492 {Opt_journal_dev, "journal_dev=%u"}, 1493 {Opt_journal_path, "journal_path=%s"}, 1494 {Opt_journal_checksum, "journal_checksum"}, 1495 {Opt_nojournal_checksum, "nojournal_checksum"}, 1496 {Opt_journal_async_commit, "journal_async_commit"}, 1497 {Opt_abort, "abort"}, 1498 {Opt_data_journal, "data=journal"}, 1499 {Opt_data_ordered, "data=ordered"}, 1500 {Opt_data_writeback, "data=writeback"}, 1501 {Opt_data_err_abort, "data_err=abort"}, 1502 {Opt_data_err_ignore, "data_err=ignore"}, 1503 {Opt_offusrjquota, "usrjquota="}, 1504 {Opt_usrjquota, "usrjquota=%s"}, 1505 {Opt_offgrpjquota, "grpjquota="}, 1506 {Opt_grpjquota, "grpjquota=%s"}, 1507 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1508 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1509 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1510 {Opt_grpquota, "grpquota"}, 1511 {Opt_noquota, "noquota"}, 1512 {Opt_quota, "quota"}, 1513 {Opt_usrquota, "usrquota"}, 1514 {Opt_prjquota, "prjquota"}, 1515 {Opt_barrier, "barrier=%u"}, 1516 {Opt_barrier, "barrier"}, 1517 {Opt_nobarrier, "nobarrier"}, 1518 {Opt_i_version, "i_version"}, 1519 {Opt_dax, "dax"}, 1520 {Opt_stripe, "stripe=%u"}, 1521 {Opt_delalloc, "delalloc"}, 1522 {Opt_warn_on_error, "warn_on_error"}, 1523 {Opt_nowarn_on_error, "nowarn_on_error"}, 1524 {Opt_lazytime, "lazytime"}, 1525 {Opt_nolazytime, "nolazytime"}, 1526 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1527 {Opt_nodelalloc, "nodelalloc"}, 1528 {Opt_removed, "mblk_io_submit"}, 1529 {Opt_removed, "nomblk_io_submit"}, 1530 {Opt_block_validity, "block_validity"}, 1531 {Opt_noblock_validity, "noblock_validity"}, 1532 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1533 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1534 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1535 {Opt_auto_da_alloc, "auto_da_alloc"}, 1536 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1537 {Opt_dioread_nolock, "dioread_nolock"}, 1538 {Opt_dioread_lock, "dioread_lock"}, 1539 {Opt_discard, "discard"}, 1540 {Opt_nodiscard, "nodiscard"}, 1541 {Opt_init_itable, "init_itable=%u"}, 1542 {Opt_init_itable, "init_itable"}, 1543 {Opt_noinit_itable, "noinit_itable"}, 1544 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1545 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1546 {Opt_nombcache, "nombcache"}, 1547 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1548 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1549 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1550 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1551 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1552 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1553 {Opt_err, NULL}, 1554 }; 1555 1556 static ext4_fsblk_t get_sb_block(void **data) 1557 { 1558 ext4_fsblk_t sb_block; 1559 char *options = (char *) *data; 1560 1561 if (!options || strncmp(options, "sb=", 3) != 0) 1562 return 1; /* Default location */ 1563 1564 options += 3; 1565 /* TODO: use simple_strtoll with >32bit ext4 */ 1566 sb_block = simple_strtoul(options, &options, 0); 1567 if (*options && *options != ',') { 1568 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1569 (char *) *data); 1570 return 1; 1571 } 1572 if (*options == ',') 1573 options++; 1574 *data = (void *) options; 1575 1576 return sb_block; 1577 } 1578 1579 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1580 static const char deprecated_msg[] = 1581 "Mount option \"%s\" will be removed by %s\n" 1582 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1583 1584 #ifdef CONFIG_QUOTA 1585 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1586 { 1587 struct ext4_sb_info *sbi = EXT4_SB(sb); 1588 char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 1589 int ret = -1; 1590 1591 if (sb_any_quota_loaded(sb) && !old_qname) { 1592 ext4_msg(sb, KERN_ERR, 1593 "Cannot change journaled " 1594 "quota options when quota turned on"); 1595 return -1; 1596 } 1597 if (ext4_has_feature_quota(sb)) { 1598 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1599 "ignored when QUOTA feature is enabled"); 1600 return 1; 1601 } 1602 qname = match_strdup(args); 1603 if (!qname) { 1604 ext4_msg(sb, KERN_ERR, 1605 "Not enough memory for storing quotafile name"); 1606 return -1; 1607 } 1608 if (old_qname) { 1609 if (strcmp(old_qname, qname) == 0) 1610 ret = 1; 1611 else 1612 ext4_msg(sb, KERN_ERR, 1613 "%s quota file already specified", 1614 QTYPE2NAME(qtype)); 1615 goto errout; 1616 } 1617 if (strchr(qname, '/')) { 1618 ext4_msg(sb, KERN_ERR, 1619 "quotafile must be on filesystem root"); 1620 goto errout; 1621 } 1622 rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1623 set_opt(sb, QUOTA); 1624 return 1; 1625 errout: 1626 kfree(qname); 1627 return ret; 1628 } 1629 1630 static int clear_qf_name(struct super_block *sb, int qtype) 1631 { 1632 1633 struct ext4_sb_info *sbi = EXT4_SB(sb); 1634 char *old_qname = get_qf_name(sb, sbi, qtype); 1635 1636 if (sb_any_quota_loaded(sb) && old_qname) { 1637 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1638 " when quota turned on"); 1639 return -1; 1640 } 1641 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 1642 synchronize_rcu(); 1643 kfree(old_qname); 1644 return 1; 1645 } 1646 #endif 1647 1648 #define MOPT_SET 0x0001 1649 #define MOPT_CLEAR 0x0002 1650 #define MOPT_NOSUPPORT 0x0004 1651 #define MOPT_EXPLICIT 0x0008 1652 #define MOPT_CLEAR_ERR 0x0010 1653 #define MOPT_GTE0 0x0020 1654 #ifdef CONFIG_QUOTA 1655 #define MOPT_Q 0 1656 #define MOPT_QFMT 0x0040 1657 #else 1658 #define MOPT_Q MOPT_NOSUPPORT 1659 #define MOPT_QFMT MOPT_NOSUPPORT 1660 #endif 1661 #define MOPT_DATAJ 0x0080 1662 #define MOPT_NO_EXT2 0x0100 1663 #define MOPT_NO_EXT3 0x0200 1664 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1665 #define MOPT_STRING 0x0400 1666 1667 static const struct mount_opts { 1668 int token; 1669 int mount_opt; 1670 int flags; 1671 } ext4_mount_opts[] = { 1672 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1673 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1674 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1675 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1676 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1677 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1678 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1679 MOPT_EXT4_ONLY | MOPT_SET}, 1680 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1681 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1682 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1683 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1684 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1685 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1686 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1687 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1688 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1689 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1690 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1691 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1692 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1693 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1694 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1695 EXT4_MOUNT_JOURNAL_CHECKSUM), 1696 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1697 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1698 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1699 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1700 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1701 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1702 MOPT_NO_EXT2}, 1703 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1704 MOPT_NO_EXT2}, 1705 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1706 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1707 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1708 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1709 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1710 {Opt_commit, 0, MOPT_GTE0}, 1711 {Opt_max_batch_time, 0, MOPT_GTE0}, 1712 {Opt_min_batch_time, 0, MOPT_GTE0}, 1713 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1714 {Opt_init_itable, 0, MOPT_GTE0}, 1715 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 1716 {Opt_stripe, 0, MOPT_GTE0}, 1717 {Opt_resuid, 0, MOPT_GTE0}, 1718 {Opt_resgid, 0, MOPT_GTE0}, 1719 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1720 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1721 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1722 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1723 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1724 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1725 MOPT_NO_EXT2 | MOPT_DATAJ}, 1726 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1727 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1728 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1729 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1730 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1731 #else 1732 {Opt_acl, 0, MOPT_NOSUPPORT}, 1733 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1734 #endif 1735 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1736 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1737 {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 1738 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1739 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1740 MOPT_SET | MOPT_Q}, 1741 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1742 MOPT_SET | MOPT_Q}, 1743 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1744 MOPT_SET | MOPT_Q}, 1745 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1746 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1747 MOPT_CLEAR | MOPT_Q}, 1748 {Opt_usrjquota, 0, MOPT_Q}, 1749 {Opt_grpjquota, 0, MOPT_Q}, 1750 {Opt_offusrjquota, 0, MOPT_Q}, 1751 {Opt_offgrpjquota, 0, MOPT_Q}, 1752 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1753 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1754 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1755 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1756 {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1757 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1758 {Opt_err, 0, 0} 1759 }; 1760 1761 #ifdef CONFIG_UNICODE 1762 static const struct ext4_sb_encodings { 1763 __u16 magic; 1764 char *name; 1765 char *version; 1766 } ext4_sb_encoding_map[] = { 1767 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1768 }; 1769 1770 static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1771 const struct ext4_sb_encodings **encoding, 1772 __u16 *flags) 1773 { 1774 __u16 magic = le16_to_cpu(es->s_encoding); 1775 int i; 1776 1777 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1778 if (magic == ext4_sb_encoding_map[i].magic) 1779 break; 1780 1781 if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1782 return -EINVAL; 1783 1784 *encoding = &ext4_sb_encoding_map[i]; 1785 *flags = le16_to_cpu(es->s_encoding_flags); 1786 1787 return 0; 1788 } 1789 #endif 1790 1791 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1792 substring_t *args, unsigned long *journal_devnum, 1793 unsigned int *journal_ioprio, int is_remount) 1794 { 1795 struct ext4_sb_info *sbi = EXT4_SB(sb); 1796 const struct mount_opts *m; 1797 kuid_t uid; 1798 kgid_t gid; 1799 int arg = 0; 1800 1801 #ifdef CONFIG_QUOTA 1802 if (token == Opt_usrjquota) 1803 return set_qf_name(sb, USRQUOTA, &args[0]); 1804 else if (token == Opt_grpjquota) 1805 return set_qf_name(sb, GRPQUOTA, &args[0]); 1806 else if (token == Opt_offusrjquota) 1807 return clear_qf_name(sb, USRQUOTA); 1808 else if (token == Opt_offgrpjquota) 1809 return clear_qf_name(sb, GRPQUOTA); 1810 #endif 1811 switch (token) { 1812 case Opt_noacl: 1813 case Opt_nouser_xattr: 1814 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1815 break; 1816 case Opt_sb: 1817 return 1; /* handled by get_sb_block() */ 1818 case Opt_removed: 1819 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1820 return 1; 1821 case Opt_abort: 1822 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1823 return 1; 1824 case Opt_i_version: 1825 sb->s_flags |= SB_I_VERSION; 1826 return 1; 1827 case Opt_lazytime: 1828 sb->s_flags |= SB_LAZYTIME; 1829 return 1; 1830 case Opt_nolazytime: 1831 sb->s_flags &= ~SB_LAZYTIME; 1832 return 1; 1833 } 1834 1835 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1836 if (token == m->token) 1837 break; 1838 1839 if (m->token == Opt_err) { 1840 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1841 "or missing value", opt); 1842 return -1; 1843 } 1844 1845 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1846 ext4_msg(sb, KERN_ERR, 1847 "Mount option \"%s\" incompatible with ext2", opt); 1848 return -1; 1849 } 1850 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1851 ext4_msg(sb, KERN_ERR, 1852 "Mount option \"%s\" incompatible with ext3", opt); 1853 return -1; 1854 } 1855 1856 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1857 return -1; 1858 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1859 return -1; 1860 if (m->flags & MOPT_EXPLICIT) { 1861 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 1862 set_opt2(sb, EXPLICIT_DELALLOC); 1863 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 1864 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1865 } else 1866 return -1; 1867 } 1868 if (m->flags & MOPT_CLEAR_ERR) 1869 clear_opt(sb, ERRORS_MASK); 1870 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1871 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1872 "options when quota turned on"); 1873 return -1; 1874 } 1875 1876 if (m->flags & MOPT_NOSUPPORT) { 1877 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1878 } else if (token == Opt_commit) { 1879 if (arg == 0) 1880 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1881 sbi->s_commit_interval = HZ * arg; 1882 } else if (token == Opt_debug_want_extra_isize) { 1883 sbi->s_want_extra_isize = arg; 1884 } else if (token == Opt_max_batch_time) { 1885 sbi->s_max_batch_time = arg; 1886 } else if (token == Opt_min_batch_time) { 1887 sbi->s_min_batch_time = arg; 1888 } else if (token == Opt_inode_readahead_blks) { 1889 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1890 ext4_msg(sb, KERN_ERR, 1891 "EXT4-fs: inode_readahead_blks must be " 1892 "0 or a power of 2 smaller than 2^31"); 1893 return -1; 1894 } 1895 sbi->s_inode_readahead_blks = arg; 1896 } else if (token == Opt_init_itable) { 1897 set_opt(sb, INIT_INODE_TABLE); 1898 if (!args->from) 1899 arg = EXT4_DEF_LI_WAIT_MULT; 1900 sbi->s_li_wait_mult = arg; 1901 } else if (token == Opt_max_dir_size_kb) { 1902 sbi->s_max_dir_size_kb = arg; 1903 } else if (token == Opt_stripe) { 1904 sbi->s_stripe = arg; 1905 } else if (token == Opt_resuid) { 1906 uid = make_kuid(current_user_ns(), arg); 1907 if (!uid_valid(uid)) { 1908 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1909 return -1; 1910 } 1911 sbi->s_resuid = uid; 1912 } else if (token == Opt_resgid) { 1913 gid = make_kgid(current_user_ns(), arg); 1914 if (!gid_valid(gid)) { 1915 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1916 return -1; 1917 } 1918 sbi->s_resgid = gid; 1919 } else if (token == Opt_journal_dev) { 1920 if (is_remount) { 1921 ext4_msg(sb, KERN_ERR, 1922 "Cannot specify journal on remount"); 1923 return -1; 1924 } 1925 *journal_devnum = arg; 1926 } else if (token == Opt_journal_path) { 1927 char *journal_path; 1928 struct inode *journal_inode; 1929 struct path path; 1930 int error; 1931 1932 if (is_remount) { 1933 ext4_msg(sb, KERN_ERR, 1934 "Cannot specify journal on remount"); 1935 return -1; 1936 } 1937 journal_path = match_strdup(&args[0]); 1938 if (!journal_path) { 1939 ext4_msg(sb, KERN_ERR, "error: could not dup " 1940 "journal device string"); 1941 return -1; 1942 } 1943 1944 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1945 if (error) { 1946 ext4_msg(sb, KERN_ERR, "error: could not find " 1947 "journal device path: error %d", error); 1948 kfree(journal_path); 1949 return -1; 1950 } 1951 1952 journal_inode = d_inode(path.dentry); 1953 if (!S_ISBLK(journal_inode->i_mode)) { 1954 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1955 "is not a block device", journal_path); 1956 path_put(&path); 1957 kfree(journal_path); 1958 return -1; 1959 } 1960 1961 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1962 path_put(&path); 1963 kfree(journal_path); 1964 } else if (token == Opt_journal_ioprio) { 1965 if (arg > 7) { 1966 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1967 " (must be 0-7)"); 1968 return -1; 1969 } 1970 *journal_ioprio = 1971 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1972 } else if (token == Opt_test_dummy_encryption) { 1973 #ifdef CONFIG_FS_ENCRYPTION 1974 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 1975 ext4_msg(sb, KERN_WARNING, 1976 "Test dummy encryption mode enabled"); 1977 #else 1978 ext4_msg(sb, KERN_WARNING, 1979 "Test dummy encryption mount option ignored"); 1980 #endif 1981 } else if (m->flags & MOPT_DATAJ) { 1982 if (is_remount) { 1983 if (!sbi->s_journal) 1984 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1985 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 1986 ext4_msg(sb, KERN_ERR, 1987 "Cannot change data mode on remount"); 1988 return -1; 1989 } 1990 } else { 1991 clear_opt(sb, DATA_FLAGS); 1992 sbi->s_mount_opt |= m->mount_opt; 1993 } 1994 #ifdef CONFIG_QUOTA 1995 } else if (m->flags & MOPT_QFMT) { 1996 if (sb_any_quota_loaded(sb) && 1997 sbi->s_jquota_fmt != m->mount_opt) { 1998 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 1999 "quota options when quota turned on"); 2000 return -1; 2001 } 2002 if (ext4_has_feature_quota(sb)) { 2003 ext4_msg(sb, KERN_INFO, 2004 "Quota format mount options ignored " 2005 "when QUOTA feature is enabled"); 2006 return 1; 2007 } 2008 sbi->s_jquota_fmt = m->mount_opt; 2009 #endif 2010 } else if (token == Opt_dax) { 2011 #ifdef CONFIG_FS_DAX 2012 ext4_msg(sb, KERN_WARNING, 2013 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 2014 sbi->s_mount_opt |= m->mount_opt; 2015 #else 2016 ext4_msg(sb, KERN_INFO, "dax option not supported"); 2017 return -1; 2018 #endif 2019 } else if (token == Opt_data_err_abort) { 2020 sbi->s_mount_opt |= m->mount_opt; 2021 } else if (token == Opt_data_err_ignore) { 2022 sbi->s_mount_opt &= ~m->mount_opt; 2023 } else { 2024 if (!args->from) 2025 arg = 1; 2026 if (m->flags & MOPT_CLEAR) 2027 arg = !arg; 2028 else if (unlikely(!(m->flags & MOPT_SET))) { 2029 ext4_msg(sb, KERN_WARNING, 2030 "buggy handling of option %s", opt); 2031 WARN_ON(1); 2032 return -1; 2033 } 2034 if (arg != 0) 2035 sbi->s_mount_opt |= m->mount_opt; 2036 else 2037 sbi->s_mount_opt &= ~m->mount_opt; 2038 } 2039 return 1; 2040 } 2041 2042 static int parse_options(char *options, struct super_block *sb, 2043 unsigned long *journal_devnum, 2044 unsigned int *journal_ioprio, 2045 int is_remount) 2046 { 2047 struct ext4_sb_info *sbi = EXT4_SB(sb); 2048 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2049 substring_t args[MAX_OPT_ARGS]; 2050 int token; 2051 2052 if (!options) 2053 return 1; 2054 2055 while ((p = strsep(&options, ",")) != NULL) { 2056 if (!*p) 2057 continue; 2058 /* 2059 * Initialize args struct so we know whether arg was 2060 * found; some options take optional arguments. 2061 */ 2062 args[0].to = args[0].from = NULL; 2063 token = match_token(p, tokens, args); 2064 if (handle_mount_opt(sb, p, token, args, journal_devnum, 2065 journal_ioprio, is_remount) < 0) 2066 return 0; 2067 } 2068 #ifdef CONFIG_QUOTA 2069 /* 2070 * We do the test below only for project quotas. 'usrquota' and 2071 * 'grpquota' mount options are allowed even without quota feature 2072 * to support legacy quotas in quota files. 2073 */ 2074 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 2075 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 2076 "Cannot enable project quota enforcement."); 2077 return 0; 2078 } 2079 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 2080 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 2081 if (usr_qf_name || grp_qf_name) { 2082 if (test_opt(sb, USRQUOTA) && usr_qf_name) 2083 clear_opt(sb, USRQUOTA); 2084 2085 if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2086 clear_opt(sb, GRPQUOTA); 2087 2088 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2089 ext4_msg(sb, KERN_ERR, "old and new quota " 2090 "format mixing"); 2091 return 0; 2092 } 2093 2094 if (!sbi->s_jquota_fmt) { 2095 ext4_msg(sb, KERN_ERR, "journaled quota format " 2096 "not specified"); 2097 return 0; 2098 } 2099 } 2100 #endif 2101 if (test_opt(sb, DIOREAD_NOLOCK)) { 2102 int blocksize = 2103 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2104 2105 if (blocksize < PAGE_SIZE) { 2106 ext4_msg(sb, KERN_ERR, "can't mount with " 2107 "dioread_nolock if block size != PAGE_SIZE"); 2108 return 0; 2109 } 2110 } 2111 return 1; 2112 } 2113 2114 static inline void ext4_show_quota_options(struct seq_file *seq, 2115 struct super_block *sb) 2116 { 2117 #if defined(CONFIG_QUOTA) 2118 struct ext4_sb_info *sbi = EXT4_SB(sb); 2119 char *usr_qf_name, *grp_qf_name; 2120 2121 if (sbi->s_jquota_fmt) { 2122 char *fmtname = ""; 2123 2124 switch (sbi->s_jquota_fmt) { 2125 case QFMT_VFS_OLD: 2126 fmtname = "vfsold"; 2127 break; 2128 case QFMT_VFS_V0: 2129 fmtname = "vfsv0"; 2130 break; 2131 case QFMT_VFS_V1: 2132 fmtname = "vfsv1"; 2133 break; 2134 } 2135 seq_printf(seq, ",jqfmt=%s", fmtname); 2136 } 2137 2138 rcu_read_lock(); 2139 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2140 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2141 if (usr_qf_name) 2142 seq_show_option(seq, "usrjquota", usr_qf_name); 2143 if (grp_qf_name) 2144 seq_show_option(seq, "grpjquota", grp_qf_name); 2145 rcu_read_unlock(); 2146 #endif 2147 } 2148 2149 static const char *token2str(int token) 2150 { 2151 const struct match_token *t; 2152 2153 for (t = tokens; t->token != Opt_err; t++) 2154 if (t->token == token && !strchr(t->pattern, '=')) 2155 break; 2156 return t->pattern; 2157 } 2158 2159 /* 2160 * Show an option if 2161 * - it's set to a non-default value OR 2162 * - if the per-sb default is different from the global default 2163 */ 2164 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2165 int nodefs) 2166 { 2167 struct ext4_sb_info *sbi = EXT4_SB(sb); 2168 struct ext4_super_block *es = sbi->s_es; 2169 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2170 const struct mount_opts *m; 2171 char sep = nodefs ? '\n' : ','; 2172 2173 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2174 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2175 2176 if (sbi->s_sb_block != 1) 2177 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2178 2179 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2180 int want_set = m->flags & MOPT_SET; 2181 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2182 (m->flags & MOPT_CLEAR_ERR)) 2183 continue; 2184 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2185 continue; /* skip if same as the default */ 2186 if ((want_set && 2187 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2188 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2189 continue; /* select Opt_noFoo vs Opt_Foo */ 2190 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2191 } 2192 2193 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2194 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2195 SEQ_OPTS_PRINT("resuid=%u", 2196 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2197 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2198 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2199 SEQ_OPTS_PRINT("resgid=%u", 2200 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2201 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2202 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2203 SEQ_OPTS_PUTS("errors=remount-ro"); 2204 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2205 SEQ_OPTS_PUTS("errors=continue"); 2206 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2207 SEQ_OPTS_PUTS("errors=panic"); 2208 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2209 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2210 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2211 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2212 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2213 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2214 if (sb->s_flags & SB_I_VERSION) 2215 SEQ_OPTS_PUTS("i_version"); 2216 if (nodefs || sbi->s_stripe) 2217 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2218 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2219 (sbi->s_mount_opt ^ def_mount_opt)) { 2220 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2221 SEQ_OPTS_PUTS("data=journal"); 2222 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2223 SEQ_OPTS_PUTS("data=ordered"); 2224 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2225 SEQ_OPTS_PUTS("data=writeback"); 2226 } 2227 if (nodefs || 2228 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2229 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2230 sbi->s_inode_readahead_blks); 2231 2232 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2233 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2234 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2235 if (nodefs || sbi->s_max_dir_size_kb) 2236 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2237 if (test_opt(sb, DATA_ERR_ABORT)) 2238 SEQ_OPTS_PUTS("data_err=abort"); 2239 if (DUMMY_ENCRYPTION_ENABLED(sbi)) 2240 SEQ_OPTS_PUTS("test_dummy_encryption"); 2241 2242 ext4_show_quota_options(seq, sb); 2243 return 0; 2244 } 2245 2246 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2247 { 2248 return _ext4_show_options(seq, root->d_sb, 0); 2249 } 2250 2251 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2252 { 2253 struct super_block *sb = seq->private; 2254 int rc; 2255 2256 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 2257 rc = _ext4_show_options(seq, sb, 1); 2258 seq_puts(seq, "\n"); 2259 return rc; 2260 } 2261 2262 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2263 int read_only) 2264 { 2265 struct ext4_sb_info *sbi = EXT4_SB(sb); 2266 int err = 0; 2267 2268 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2269 ext4_msg(sb, KERN_ERR, "revision level too high, " 2270 "forcing read-only mode"); 2271 err = -EROFS; 2272 } 2273 if (read_only) 2274 goto done; 2275 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2276 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2277 "running e2fsck is recommended"); 2278 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2279 ext4_msg(sb, KERN_WARNING, 2280 "warning: mounting fs with errors, " 2281 "running e2fsck is recommended"); 2282 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2283 le16_to_cpu(es->s_mnt_count) >= 2284 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2285 ext4_msg(sb, KERN_WARNING, 2286 "warning: maximal mount count reached, " 2287 "running e2fsck is recommended"); 2288 else if (le32_to_cpu(es->s_checkinterval) && 2289 (ext4_get_tstamp(es, s_lastcheck) + 2290 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2291 ext4_msg(sb, KERN_WARNING, 2292 "warning: checktime reached, " 2293 "running e2fsck is recommended"); 2294 if (!sbi->s_journal) 2295 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2296 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2297 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2298 le16_add_cpu(&es->s_mnt_count, 1); 2299 ext4_update_tstamp(es, s_mtime); 2300 if (sbi->s_journal) 2301 ext4_set_feature_journal_needs_recovery(sb); 2302 2303 err = ext4_commit_super(sb, 1); 2304 done: 2305 if (test_opt(sb, DEBUG)) 2306 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2307 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2308 sb->s_blocksize, 2309 sbi->s_groups_count, 2310 EXT4_BLOCKS_PER_GROUP(sb), 2311 EXT4_INODES_PER_GROUP(sb), 2312 sbi->s_mount_opt, sbi->s_mount_opt2); 2313 2314 cleancache_init_fs(sb); 2315 return err; 2316 } 2317 2318 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2319 { 2320 struct ext4_sb_info *sbi = EXT4_SB(sb); 2321 struct flex_groups *new_groups; 2322 int size; 2323 2324 if (!sbi->s_log_groups_per_flex) 2325 return 0; 2326 2327 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2328 if (size <= sbi->s_flex_groups_allocated) 2329 return 0; 2330 2331 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2332 new_groups = kvzalloc(size, GFP_KERNEL); 2333 if (!new_groups) { 2334 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2335 size / (int) sizeof(struct flex_groups)); 2336 return -ENOMEM; 2337 } 2338 2339 if (sbi->s_flex_groups) { 2340 memcpy(new_groups, sbi->s_flex_groups, 2341 (sbi->s_flex_groups_allocated * 2342 sizeof(struct flex_groups))); 2343 kvfree(sbi->s_flex_groups); 2344 } 2345 sbi->s_flex_groups = new_groups; 2346 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2347 return 0; 2348 } 2349 2350 static int ext4_fill_flex_info(struct super_block *sb) 2351 { 2352 struct ext4_sb_info *sbi = EXT4_SB(sb); 2353 struct ext4_group_desc *gdp = NULL; 2354 ext4_group_t flex_group; 2355 int i, err; 2356 2357 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2358 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2359 sbi->s_log_groups_per_flex = 0; 2360 return 1; 2361 } 2362 2363 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2364 if (err) 2365 goto failed; 2366 2367 for (i = 0; i < sbi->s_groups_count; i++) { 2368 gdp = ext4_get_group_desc(sb, i, NULL); 2369 2370 flex_group = ext4_flex_group(sbi, i); 2371 atomic_add(ext4_free_inodes_count(sb, gdp), 2372 &sbi->s_flex_groups[flex_group].free_inodes); 2373 atomic64_add(ext4_free_group_clusters(sb, gdp), 2374 &sbi->s_flex_groups[flex_group].free_clusters); 2375 atomic_add(ext4_used_dirs_count(sb, gdp), 2376 &sbi->s_flex_groups[flex_group].used_dirs); 2377 } 2378 2379 return 1; 2380 failed: 2381 return 0; 2382 } 2383 2384 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2385 struct ext4_group_desc *gdp) 2386 { 2387 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2388 __u16 crc = 0; 2389 __le32 le_group = cpu_to_le32(block_group); 2390 struct ext4_sb_info *sbi = EXT4_SB(sb); 2391 2392 if (ext4_has_metadata_csum(sbi->s_sb)) { 2393 /* Use new metadata_csum algorithm */ 2394 __u32 csum32; 2395 __u16 dummy_csum = 0; 2396 2397 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2398 sizeof(le_group)); 2399 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2400 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2401 sizeof(dummy_csum)); 2402 offset += sizeof(dummy_csum); 2403 if (offset < sbi->s_desc_size) 2404 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2405 sbi->s_desc_size - offset); 2406 2407 crc = csum32 & 0xFFFF; 2408 goto out; 2409 } 2410 2411 /* old crc16 code */ 2412 if (!ext4_has_feature_gdt_csum(sb)) 2413 return 0; 2414 2415 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2416 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2417 crc = crc16(crc, (__u8 *)gdp, offset); 2418 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2419 /* for checksum of struct ext4_group_desc do the rest...*/ 2420 if (ext4_has_feature_64bit(sb) && 2421 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2422 crc = crc16(crc, (__u8 *)gdp + offset, 2423 le16_to_cpu(sbi->s_es->s_desc_size) - 2424 offset); 2425 2426 out: 2427 return cpu_to_le16(crc); 2428 } 2429 2430 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2431 struct ext4_group_desc *gdp) 2432 { 2433 if (ext4_has_group_desc_csum(sb) && 2434 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2435 return 0; 2436 2437 return 1; 2438 } 2439 2440 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2441 struct ext4_group_desc *gdp) 2442 { 2443 if (!ext4_has_group_desc_csum(sb)) 2444 return; 2445 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2446 } 2447 2448 /* Called at mount-time, super-block is locked */ 2449 static int ext4_check_descriptors(struct super_block *sb, 2450 ext4_fsblk_t sb_block, 2451 ext4_group_t *first_not_zeroed) 2452 { 2453 struct ext4_sb_info *sbi = EXT4_SB(sb); 2454 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2455 ext4_fsblk_t last_block; 2456 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2457 ext4_fsblk_t block_bitmap; 2458 ext4_fsblk_t inode_bitmap; 2459 ext4_fsblk_t inode_table; 2460 int flexbg_flag = 0; 2461 ext4_group_t i, grp = sbi->s_groups_count; 2462 2463 if (ext4_has_feature_flex_bg(sb)) 2464 flexbg_flag = 1; 2465 2466 ext4_debug("Checking group descriptors"); 2467 2468 for (i = 0; i < sbi->s_groups_count; i++) { 2469 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2470 2471 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2472 last_block = ext4_blocks_count(sbi->s_es) - 1; 2473 else 2474 last_block = first_block + 2475 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2476 2477 if ((grp == sbi->s_groups_count) && 2478 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2479 grp = i; 2480 2481 block_bitmap = ext4_block_bitmap(sb, gdp); 2482 if (block_bitmap == sb_block) { 2483 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2484 "Block bitmap for group %u overlaps " 2485 "superblock", i); 2486 if (!sb_rdonly(sb)) 2487 return 0; 2488 } 2489 if (block_bitmap >= sb_block + 1 && 2490 block_bitmap <= last_bg_block) { 2491 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2492 "Block bitmap for group %u overlaps " 2493 "block group descriptors", i); 2494 if (!sb_rdonly(sb)) 2495 return 0; 2496 } 2497 if (block_bitmap < first_block || block_bitmap > last_block) { 2498 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2499 "Block bitmap for group %u not in group " 2500 "(block %llu)!", i, block_bitmap); 2501 return 0; 2502 } 2503 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2504 if (inode_bitmap == sb_block) { 2505 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2506 "Inode bitmap for group %u overlaps " 2507 "superblock", i); 2508 if (!sb_rdonly(sb)) 2509 return 0; 2510 } 2511 if (inode_bitmap >= sb_block + 1 && 2512 inode_bitmap <= last_bg_block) { 2513 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2514 "Inode bitmap for group %u overlaps " 2515 "block group descriptors", i); 2516 if (!sb_rdonly(sb)) 2517 return 0; 2518 } 2519 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2520 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2521 "Inode bitmap for group %u not in group " 2522 "(block %llu)!", i, inode_bitmap); 2523 return 0; 2524 } 2525 inode_table = ext4_inode_table(sb, gdp); 2526 if (inode_table == sb_block) { 2527 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2528 "Inode table for group %u overlaps " 2529 "superblock", i); 2530 if (!sb_rdonly(sb)) 2531 return 0; 2532 } 2533 if (inode_table >= sb_block + 1 && 2534 inode_table <= last_bg_block) { 2535 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2536 "Inode table for group %u overlaps " 2537 "block group descriptors", i); 2538 if (!sb_rdonly(sb)) 2539 return 0; 2540 } 2541 if (inode_table < first_block || 2542 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2543 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2544 "Inode table for group %u not in group " 2545 "(block %llu)!", i, inode_table); 2546 return 0; 2547 } 2548 ext4_lock_group(sb, i); 2549 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2550 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2551 "Checksum for group %u failed (%u!=%u)", 2552 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2553 gdp)), le16_to_cpu(gdp->bg_checksum)); 2554 if (!sb_rdonly(sb)) { 2555 ext4_unlock_group(sb, i); 2556 return 0; 2557 } 2558 } 2559 ext4_unlock_group(sb, i); 2560 if (!flexbg_flag) 2561 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2562 } 2563 if (NULL != first_not_zeroed) 2564 *first_not_zeroed = grp; 2565 return 1; 2566 } 2567 2568 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2569 * the superblock) which were deleted from all directories, but held open by 2570 * a process at the time of a crash. We walk the list and try to delete these 2571 * inodes at recovery time (only with a read-write filesystem). 2572 * 2573 * In order to keep the orphan inode chain consistent during traversal (in 2574 * case of crash during recovery), we link each inode into the superblock 2575 * orphan list_head and handle it the same way as an inode deletion during 2576 * normal operation (which journals the operations for us). 2577 * 2578 * We only do an iget() and an iput() on each inode, which is very safe if we 2579 * accidentally point at an in-use or already deleted inode. The worst that 2580 * can happen in this case is that we get a "bit already cleared" message from 2581 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2582 * e2fsck was run on this filesystem, and it must have already done the orphan 2583 * inode cleanup for us, so we can safely abort without any further action. 2584 */ 2585 static void ext4_orphan_cleanup(struct super_block *sb, 2586 struct ext4_super_block *es) 2587 { 2588 unsigned int s_flags = sb->s_flags; 2589 int ret, nr_orphans = 0, nr_truncates = 0; 2590 #ifdef CONFIG_QUOTA 2591 int quota_update = 0; 2592 int i; 2593 #endif 2594 if (!es->s_last_orphan) { 2595 jbd_debug(4, "no orphan inodes to clean up\n"); 2596 return; 2597 } 2598 2599 if (bdev_read_only(sb->s_bdev)) { 2600 ext4_msg(sb, KERN_ERR, "write access " 2601 "unavailable, skipping orphan cleanup"); 2602 return; 2603 } 2604 2605 /* Check if feature set would not allow a r/w mount */ 2606 if (!ext4_feature_set_ok(sb, 0)) { 2607 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2608 "unknown ROCOMPAT features"); 2609 return; 2610 } 2611 2612 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2613 /* don't clear list on RO mount w/ errors */ 2614 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 2615 ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2616 "clearing orphan list.\n"); 2617 es->s_last_orphan = 0; 2618 } 2619 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2620 return; 2621 } 2622 2623 if (s_flags & SB_RDONLY) { 2624 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2625 sb->s_flags &= ~SB_RDONLY; 2626 } 2627 #ifdef CONFIG_QUOTA 2628 /* Needed for iput() to work correctly and not trash data */ 2629 sb->s_flags |= SB_ACTIVE; 2630 2631 /* 2632 * Turn on quotas which were not enabled for read-only mounts if 2633 * filesystem has quota feature, so that they are updated correctly. 2634 */ 2635 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 2636 int ret = ext4_enable_quotas(sb); 2637 2638 if (!ret) 2639 quota_update = 1; 2640 else 2641 ext4_msg(sb, KERN_ERR, 2642 "Cannot turn on quotas: error %d", ret); 2643 } 2644 2645 /* Turn on journaled quotas used for old sytle */ 2646 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2647 if (EXT4_SB(sb)->s_qf_names[i]) { 2648 int ret = ext4_quota_on_mount(sb, i); 2649 2650 if (!ret) 2651 quota_update = 1; 2652 else 2653 ext4_msg(sb, KERN_ERR, 2654 "Cannot turn on journaled " 2655 "quota: type %d: error %d", i, ret); 2656 } 2657 } 2658 #endif 2659 2660 while (es->s_last_orphan) { 2661 struct inode *inode; 2662 2663 /* 2664 * We may have encountered an error during cleanup; if 2665 * so, skip the rest. 2666 */ 2667 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2668 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2669 es->s_last_orphan = 0; 2670 break; 2671 } 2672 2673 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2674 if (IS_ERR(inode)) { 2675 es->s_last_orphan = 0; 2676 break; 2677 } 2678 2679 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2680 dquot_initialize(inode); 2681 if (inode->i_nlink) { 2682 if (test_opt(sb, DEBUG)) 2683 ext4_msg(sb, KERN_DEBUG, 2684 "%s: truncating inode %lu to %lld bytes", 2685 __func__, inode->i_ino, inode->i_size); 2686 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2687 inode->i_ino, inode->i_size); 2688 inode_lock(inode); 2689 truncate_inode_pages(inode->i_mapping, inode->i_size); 2690 ret = ext4_truncate(inode); 2691 if (ret) 2692 ext4_std_error(inode->i_sb, ret); 2693 inode_unlock(inode); 2694 nr_truncates++; 2695 } else { 2696 if (test_opt(sb, DEBUG)) 2697 ext4_msg(sb, KERN_DEBUG, 2698 "%s: deleting unreferenced inode %lu", 2699 __func__, inode->i_ino); 2700 jbd_debug(2, "deleting unreferenced inode %lu\n", 2701 inode->i_ino); 2702 nr_orphans++; 2703 } 2704 iput(inode); /* The delete magic happens here! */ 2705 } 2706 2707 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2708 2709 if (nr_orphans) 2710 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2711 PLURAL(nr_orphans)); 2712 if (nr_truncates) 2713 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2714 PLURAL(nr_truncates)); 2715 #ifdef CONFIG_QUOTA 2716 /* Turn off quotas if they were enabled for orphan cleanup */ 2717 if (quota_update) { 2718 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2719 if (sb_dqopt(sb)->files[i]) 2720 dquot_quota_off(sb, i); 2721 } 2722 } 2723 #endif 2724 sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2725 } 2726 2727 /* 2728 * Maximal extent format file size. 2729 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2730 * extent format containers, within a sector_t, and within i_blocks 2731 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2732 * so that won't be a limiting factor. 2733 * 2734 * However there is other limiting factor. We do store extents in the form 2735 * of starting block and length, hence the resulting length of the extent 2736 * covering maximum file size must fit into on-disk format containers as 2737 * well. Given that length is always by 1 unit bigger than max unit (because 2738 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2739 * 2740 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2741 */ 2742 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2743 { 2744 loff_t res; 2745 loff_t upper_limit = MAX_LFS_FILESIZE; 2746 2747 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 2748 2749 if (!has_huge_files) { 2750 upper_limit = (1LL << 32) - 1; 2751 2752 /* total blocks in file system block size */ 2753 upper_limit >>= (blkbits - 9); 2754 upper_limit <<= blkbits; 2755 } 2756 2757 /* 2758 * 32-bit extent-start container, ee_block. We lower the maxbytes 2759 * by one fs block, so ee_len can cover the extent of maximum file 2760 * size 2761 */ 2762 res = (1LL << 32) - 1; 2763 res <<= blkbits; 2764 2765 /* Sanity check against vm- & vfs- imposed limits */ 2766 if (res > upper_limit) 2767 res = upper_limit; 2768 2769 return res; 2770 } 2771 2772 /* 2773 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2774 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2775 * We need to be 1 filesystem block less than the 2^48 sector limit. 2776 */ 2777 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2778 { 2779 loff_t res = EXT4_NDIR_BLOCKS; 2780 int meta_blocks; 2781 loff_t upper_limit; 2782 /* This is calculated to be the largest file size for a dense, block 2783 * mapped file such that the file's total number of 512-byte sectors, 2784 * including data and all indirect blocks, does not exceed (2^48 - 1). 2785 * 2786 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2787 * number of 512-byte sectors of the file. 2788 */ 2789 2790 if (!has_huge_files) { 2791 /* 2792 * !has_huge_files or implies that the inode i_block field 2793 * represents total file blocks in 2^32 512-byte sectors == 2794 * size of vfs inode i_blocks * 8 2795 */ 2796 upper_limit = (1LL << 32) - 1; 2797 2798 /* total blocks in file system block size */ 2799 upper_limit >>= (bits - 9); 2800 2801 } else { 2802 /* 2803 * We use 48 bit ext4_inode i_blocks 2804 * With EXT4_HUGE_FILE_FL set the i_blocks 2805 * represent total number of blocks in 2806 * file system block size 2807 */ 2808 upper_limit = (1LL << 48) - 1; 2809 2810 } 2811 2812 /* indirect blocks */ 2813 meta_blocks = 1; 2814 /* double indirect blocks */ 2815 meta_blocks += 1 + (1LL << (bits-2)); 2816 /* tripple indirect blocks */ 2817 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2818 2819 upper_limit -= meta_blocks; 2820 upper_limit <<= bits; 2821 2822 res += 1LL << (bits-2); 2823 res += 1LL << (2*(bits-2)); 2824 res += 1LL << (3*(bits-2)); 2825 res <<= bits; 2826 if (res > upper_limit) 2827 res = upper_limit; 2828 2829 if (res > MAX_LFS_FILESIZE) 2830 res = MAX_LFS_FILESIZE; 2831 2832 return res; 2833 } 2834 2835 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2836 ext4_fsblk_t logical_sb_block, int nr) 2837 { 2838 struct ext4_sb_info *sbi = EXT4_SB(sb); 2839 ext4_group_t bg, first_meta_bg; 2840 int has_super = 0; 2841 2842 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2843 2844 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 2845 return logical_sb_block + nr + 1; 2846 bg = sbi->s_desc_per_block * nr; 2847 if (ext4_bg_has_super(sb, bg)) 2848 has_super = 1; 2849 2850 /* 2851 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2852 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2853 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2854 * compensate. 2855 */ 2856 if (sb->s_blocksize == 1024 && nr == 0 && 2857 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 2858 has_super++; 2859 2860 return (has_super + ext4_group_first_block_no(sb, bg)); 2861 } 2862 2863 /** 2864 * ext4_get_stripe_size: Get the stripe size. 2865 * @sbi: In memory super block info 2866 * 2867 * If we have specified it via mount option, then 2868 * use the mount option value. If the value specified at mount time is 2869 * greater than the blocks per group use the super block value. 2870 * If the super block value is greater than blocks per group return 0. 2871 * Allocator needs it be less than blocks per group. 2872 * 2873 */ 2874 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2875 { 2876 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2877 unsigned long stripe_width = 2878 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2879 int ret; 2880 2881 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2882 ret = sbi->s_stripe; 2883 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 2884 ret = stripe_width; 2885 else if (stride && stride <= sbi->s_blocks_per_group) 2886 ret = stride; 2887 else 2888 ret = 0; 2889 2890 /* 2891 * If the stripe width is 1, this makes no sense and 2892 * we set it to 0 to turn off stripe handling code. 2893 */ 2894 if (ret <= 1) 2895 ret = 0; 2896 2897 return ret; 2898 } 2899 2900 /* 2901 * Check whether this filesystem can be mounted based on 2902 * the features present and the RDONLY/RDWR mount requested. 2903 * Returns 1 if this filesystem can be mounted as requested, 2904 * 0 if it cannot be. 2905 */ 2906 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2907 { 2908 if (ext4_has_unknown_ext4_incompat_features(sb)) { 2909 ext4_msg(sb, KERN_ERR, 2910 "Couldn't mount because of " 2911 "unsupported optional features (%x)", 2912 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2913 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2914 return 0; 2915 } 2916 2917 #ifndef CONFIG_UNICODE 2918 if (ext4_has_feature_casefold(sb)) { 2919 ext4_msg(sb, KERN_ERR, 2920 "Filesystem with casefold feature cannot be " 2921 "mounted without CONFIG_UNICODE"); 2922 return 0; 2923 } 2924 #endif 2925 2926 if (readonly) 2927 return 1; 2928 2929 if (ext4_has_feature_readonly(sb)) { 2930 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 2931 sb->s_flags |= SB_RDONLY; 2932 return 1; 2933 } 2934 2935 /* Check that feature set is OK for a read-write mount */ 2936 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2937 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2938 "unsupported optional features (%x)", 2939 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2940 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2941 return 0; 2942 } 2943 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2944 ext4_msg(sb, KERN_ERR, 2945 "Can't support bigalloc feature without " 2946 "extents feature\n"); 2947 return 0; 2948 } 2949 2950 #ifndef CONFIG_QUOTA 2951 if (ext4_has_feature_quota(sb) && !readonly) { 2952 ext4_msg(sb, KERN_ERR, 2953 "Filesystem with quota feature cannot be mounted RDWR " 2954 "without CONFIG_QUOTA"); 2955 return 0; 2956 } 2957 if (ext4_has_feature_project(sb) && !readonly) { 2958 ext4_msg(sb, KERN_ERR, 2959 "Filesystem with project quota feature cannot be mounted RDWR " 2960 "without CONFIG_QUOTA"); 2961 return 0; 2962 } 2963 #endif /* CONFIG_QUOTA */ 2964 return 1; 2965 } 2966 2967 /* 2968 * This function is called once a day if we have errors logged 2969 * on the file system 2970 */ 2971 static void print_daily_error_info(struct timer_list *t) 2972 { 2973 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 2974 struct super_block *sb = sbi->s_sb; 2975 struct ext4_super_block *es = sbi->s_es; 2976 2977 if (es->s_error_count) 2978 /* fsck newer than v1.41.13 is needed to clean this condition. */ 2979 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 2980 le32_to_cpu(es->s_error_count)); 2981 if (es->s_first_error_time) { 2982 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 2983 sb->s_id, 2984 ext4_get_tstamp(es, s_first_error_time), 2985 (int) sizeof(es->s_first_error_func), 2986 es->s_first_error_func, 2987 le32_to_cpu(es->s_first_error_line)); 2988 if (es->s_first_error_ino) 2989 printk(KERN_CONT ": inode %u", 2990 le32_to_cpu(es->s_first_error_ino)); 2991 if (es->s_first_error_block) 2992 printk(KERN_CONT ": block %llu", (unsigned long long) 2993 le64_to_cpu(es->s_first_error_block)); 2994 printk(KERN_CONT "\n"); 2995 } 2996 if (es->s_last_error_time) { 2997 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 2998 sb->s_id, 2999 ext4_get_tstamp(es, s_last_error_time), 3000 (int) sizeof(es->s_last_error_func), 3001 es->s_last_error_func, 3002 le32_to_cpu(es->s_last_error_line)); 3003 if (es->s_last_error_ino) 3004 printk(KERN_CONT ": inode %u", 3005 le32_to_cpu(es->s_last_error_ino)); 3006 if (es->s_last_error_block) 3007 printk(KERN_CONT ": block %llu", (unsigned long long) 3008 le64_to_cpu(es->s_last_error_block)); 3009 printk(KERN_CONT "\n"); 3010 } 3011 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3012 } 3013 3014 /* Find next suitable group and run ext4_init_inode_table */ 3015 static int ext4_run_li_request(struct ext4_li_request *elr) 3016 { 3017 struct ext4_group_desc *gdp = NULL; 3018 ext4_group_t group, ngroups; 3019 struct super_block *sb; 3020 unsigned long timeout = 0; 3021 int ret = 0; 3022 3023 sb = elr->lr_super; 3024 ngroups = EXT4_SB(sb)->s_groups_count; 3025 3026 for (group = elr->lr_next_group; group < ngroups; group++) { 3027 gdp = ext4_get_group_desc(sb, group, NULL); 3028 if (!gdp) { 3029 ret = 1; 3030 break; 3031 } 3032 3033 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3034 break; 3035 } 3036 3037 if (group >= ngroups) 3038 ret = 1; 3039 3040 if (!ret) { 3041 timeout = jiffies; 3042 ret = ext4_init_inode_table(sb, group, 3043 elr->lr_timeout ? 0 : 1); 3044 if (elr->lr_timeout == 0) { 3045 timeout = (jiffies - timeout) * 3046 elr->lr_sbi->s_li_wait_mult; 3047 elr->lr_timeout = timeout; 3048 } 3049 elr->lr_next_sched = jiffies + elr->lr_timeout; 3050 elr->lr_next_group = group + 1; 3051 } 3052 return ret; 3053 } 3054 3055 /* 3056 * Remove lr_request from the list_request and free the 3057 * request structure. Should be called with li_list_mtx held 3058 */ 3059 static void ext4_remove_li_request(struct ext4_li_request *elr) 3060 { 3061 struct ext4_sb_info *sbi; 3062 3063 if (!elr) 3064 return; 3065 3066 sbi = elr->lr_sbi; 3067 3068 list_del(&elr->lr_request); 3069 sbi->s_li_request = NULL; 3070 kfree(elr); 3071 } 3072 3073 static void ext4_unregister_li_request(struct super_block *sb) 3074 { 3075 mutex_lock(&ext4_li_mtx); 3076 if (!ext4_li_info) { 3077 mutex_unlock(&ext4_li_mtx); 3078 return; 3079 } 3080 3081 mutex_lock(&ext4_li_info->li_list_mtx); 3082 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3083 mutex_unlock(&ext4_li_info->li_list_mtx); 3084 mutex_unlock(&ext4_li_mtx); 3085 } 3086 3087 static struct task_struct *ext4_lazyinit_task; 3088 3089 /* 3090 * This is the function where ext4lazyinit thread lives. It walks 3091 * through the request list searching for next scheduled filesystem. 3092 * When such a fs is found, run the lazy initialization request 3093 * (ext4_rn_li_request) and keep track of the time spend in this 3094 * function. Based on that time we compute next schedule time of 3095 * the request. When walking through the list is complete, compute 3096 * next waking time and put itself into sleep. 3097 */ 3098 static int ext4_lazyinit_thread(void *arg) 3099 { 3100 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3101 struct list_head *pos, *n; 3102 struct ext4_li_request *elr; 3103 unsigned long next_wakeup, cur; 3104 3105 BUG_ON(NULL == eli); 3106 3107 cont_thread: 3108 while (true) { 3109 next_wakeup = MAX_JIFFY_OFFSET; 3110 3111 mutex_lock(&eli->li_list_mtx); 3112 if (list_empty(&eli->li_request_list)) { 3113 mutex_unlock(&eli->li_list_mtx); 3114 goto exit_thread; 3115 } 3116 list_for_each_safe(pos, n, &eli->li_request_list) { 3117 int err = 0; 3118 int progress = 0; 3119 elr = list_entry(pos, struct ext4_li_request, 3120 lr_request); 3121 3122 if (time_before(jiffies, elr->lr_next_sched)) { 3123 if (time_before(elr->lr_next_sched, next_wakeup)) 3124 next_wakeup = elr->lr_next_sched; 3125 continue; 3126 } 3127 if (down_read_trylock(&elr->lr_super->s_umount)) { 3128 if (sb_start_write_trylock(elr->lr_super)) { 3129 progress = 1; 3130 /* 3131 * We hold sb->s_umount, sb can not 3132 * be removed from the list, it is 3133 * now safe to drop li_list_mtx 3134 */ 3135 mutex_unlock(&eli->li_list_mtx); 3136 err = ext4_run_li_request(elr); 3137 sb_end_write(elr->lr_super); 3138 mutex_lock(&eli->li_list_mtx); 3139 n = pos->next; 3140 } 3141 up_read((&elr->lr_super->s_umount)); 3142 } 3143 /* error, remove the lazy_init job */ 3144 if (err) { 3145 ext4_remove_li_request(elr); 3146 continue; 3147 } 3148 if (!progress) { 3149 elr->lr_next_sched = jiffies + 3150 (prandom_u32() 3151 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3152 } 3153 if (time_before(elr->lr_next_sched, next_wakeup)) 3154 next_wakeup = elr->lr_next_sched; 3155 } 3156 mutex_unlock(&eli->li_list_mtx); 3157 3158 try_to_freeze(); 3159 3160 cur = jiffies; 3161 if ((time_after_eq(cur, next_wakeup)) || 3162 (MAX_JIFFY_OFFSET == next_wakeup)) { 3163 cond_resched(); 3164 continue; 3165 } 3166 3167 schedule_timeout_interruptible(next_wakeup - cur); 3168 3169 if (kthread_should_stop()) { 3170 ext4_clear_request_list(); 3171 goto exit_thread; 3172 } 3173 } 3174 3175 exit_thread: 3176 /* 3177 * It looks like the request list is empty, but we need 3178 * to check it under the li_list_mtx lock, to prevent any 3179 * additions into it, and of course we should lock ext4_li_mtx 3180 * to atomically free the list and ext4_li_info, because at 3181 * this point another ext4 filesystem could be registering 3182 * new one. 3183 */ 3184 mutex_lock(&ext4_li_mtx); 3185 mutex_lock(&eli->li_list_mtx); 3186 if (!list_empty(&eli->li_request_list)) { 3187 mutex_unlock(&eli->li_list_mtx); 3188 mutex_unlock(&ext4_li_mtx); 3189 goto cont_thread; 3190 } 3191 mutex_unlock(&eli->li_list_mtx); 3192 kfree(ext4_li_info); 3193 ext4_li_info = NULL; 3194 mutex_unlock(&ext4_li_mtx); 3195 3196 return 0; 3197 } 3198 3199 static void ext4_clear_request_list(void) 3200 { 3201 struct list_head *pos, *n; 3202 struct ext4_li_request *elr; 3203 3204 mutex_lock(&ext4_li_info->li_list_mtx); 3205 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3206 elr = list_entry(pos, struct ext4_li_request, 3207 lr_request); 3208 ext4_remove_li_request(elr); 3209 } 3210 mutex_unlock(&ext4_li_info->li_list_mtx); 3211 } 3212 3213 static int ext4_run_lazyinit_thread(void) 3214 { 3215 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3216 ext4_li_info, "ext4lazyinit"); 3217 if (IS_ERR(ext4_lazyinit_task)) { 3218 int err = PTR_ERR(ext4_lazyinit_task); 3219 ext4_clear_request_list(); 3220 kfree(ext4_li_info); 3221 ext4_li_info = NULL; 3222 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3223 "initialization thread\n", 3224 err); 3225 return err; 3226 } 3227 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3228 return 0; 3229 } 3230 3231 /* 3232 * Check whether it make sense to run itable init. thread or not. 3233 * If there is at least one uninitialized inode table, return 3234 * corresponding group number, else the loop goes through all 3235 * groups and return total number of groups. 3236 */ 3237 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3238 { 3239 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3240 struct ext4_group_desc *gdp = NULL; 3241 3242 if (!ext4_has_group_desc_csum(sb)) 3243 return ngroups; 3244 3245 for (group = 0; group < ngroups; group++) { 3246 gdp = ext4_get_group_desc(sb, group, NULL); 3247 if (!gdp) 3248 continue; 3249 3250 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3251 break; 3252 } 3253 3254 return group; 3255 } 3256 3257 static int ext4_li_info_new(void) 3258 { 3259 struct ext4_lazy_init *eli = NULL; 3260 3261 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3262 if (!eli) 3263 return -ENOMEM; 3264 3265 INIT_LIST_HEAD(&eli->li_request_list); 3266 mutex_init(&eli->li_list_mtx); 3267 3268 eli->li_state |= EXT4_LAZYINIT_QUIT; 3269 3270 ext4_li_info = eli; 3271 3272 return 0; 3273 } 3274 3275 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3276 ext4_group_t start) 3277 { 3278 struct ext4_sb_info *sbi = EXT4_SB(sb); 3279 struct ext4_li_request *elr; 3280 3281 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3282 if (!elr) 3283 return NULL; 3284 3285 elr->lr_super = sb; 3286 elr->lr_sbi = sbi; 3287 elr->lr_next_group = start; 3288 3289 /* 3290 * Randomize first schedule time of the request to 3291 * spread the inode table initialization requests 3292 * better. 3293 */ 3294 elr->lr_next_sched = jiffies + (prandom_u32() % 3295 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3296 return elr; 3297 } 3298 3299 int ext4_register_li_request(struct super_block *sb, 3300 ext4_group_t first_not_zeroed) 3301 { 3302 struct ext4_sb_info *sbi = EXT4_SB(sb); 3303 struct ext4_li_request *elr = NULL; 3304 ext4_group_t ngroups = sbi->s_groups_count; 3305 int ret = 0; 3306 3307 mutex_lock(&ext4_li_mtx); 3308 if (sbi->s_li_request != NULL) { 3309 /* 3310 * Reset timeout so it can be computed again, because 3311 * s_li_wait_mult might have changed. 3312 */ 3313 sbi->s_li_request->lr_timeout = 0; 3314 goto out; 3315 } 3316 3317 if (first_not_zeroed == ngroups || sb_rdonly(sb) || 3318 !test_opt(sb, INIT_INODE_TABLE)) 3319 goto out; 3320 3321 elr = ext4_li_request_new(sb, first_not_zeroed); 3322 if (!elr) { 3323 ret = -ENOMEM; 3324 goto out; 3325 } 3326 3327 if (NULL == ext4_li_info) { 3328 ret = ext4_li_info_new(); 3329 if (ret) 3330 goto out; 3331 } 3332 3333 mutex_lock(&ext4_li_info->li_list_mtx); 3334 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3335 mutex_unlock(&ext4_li_info->li_list_mtx); 3336 3337 sbi->s_li_request = elr; 3338 /* 3339 * set elr to NULL here since it has been inserted to 3340 * the request_list and the removal and free of it is 3341 * handled by ext4_clear_request_list from now on. 3342 */ 3343 elr = NULL; 3344 3345 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3346 ret = ext4_run_lazyinit_thread(); 3347 if (ret) 3348 goto out; 3349 } 3350 out: 3351 mutex_unlock(&ext4_li_mtx); 3352 if (ret) 3353 kfree(elr); 3354 return ret; 3355 } 3356 3357 /* 3358 * We do not need to lock anything since this is called on 3359 * module unload. 3360 */ 3361 static void ext4_destroy_lazyinit_thread(void) 3362 { 3363 /* 3364 * If thread exited earlier 3365 * there's nothing to be done. 3366 */ 3367 if (!ext4_li_info || !ext4_lazyinit_task) 3368 return; 3369 3370 kthread_stop(ext4_lazyinit_task); 3371 } 3372 3373 static int set_journal_csum_feature_set(struct super_block *sb) 3374 { 3375 int ret = 1; 3376 int compat, incompat; 3377 struct ext4_sb_info *sbi = EXT4_SB(sb); 3378 3379 if (ext4_has_metadata_csum(sb)) { 3380 /* journal checksum v3 */ 3381 compat = 0; 3382 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3383 } else { 3384 /* journal checksum v1 */ 3385 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3386 incompat = 0; 3387 } 3388 3389 jbd2_journal_clear_features(sbi->s_journal, 3390 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3391 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3392 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3393 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3394 ret = jbd2_journal_set_features(sbi->s_journal, 3395 compat, 0, 3396 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3397 incompat); 3398 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3399 ret = jbd2_journal_set_features(sbi->s_journal, 3400 compat, 0, 3401 incompat); 3402 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3403 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3404 } else { 3405 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3406 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3407 } 3408 3409 return ret; 3410 } 3411 3412 /* 3413 * Note: calculating the overhead so we can be compatible with 3414 * historical BSD practice is quite difficult in the face of 3415 * clusters/bigalloc. This is because multiple metadata blocks from 3416 * different block group can end up in the same allocation cluster. 3417 * Calculating the exact overhead in the face of clustered allocation 3418 * requires either O(all block bitmaps) in memory or O(number of block 3419 * groups**2) in time. We will still calculate the superblock for 3420 * older file systems --- and if we come across with a bigalloc file 3421 * system with zero in s_overhead_clusters the estimate will be close to 3422 * correct especially for very large cluster sizes --- but for newer 3423 * file systems, it's better to calculate this figure once at mkfs 3424 * time, and store it in the superblock. If the superblock value is 3425 * present (even for non-bigalloc file systems), we will use it. 3426 */ 3427 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3428 char *buf) 3429 { 3430 struct ext4_sb_info *sbi = EXT4_SB(sb); 3431 struct ext4_group_desc *gdp; 3432 ext4_fsblk_t first_block, last_block, b; 3433 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3434 int s, j, count = 0; 3435 3436 if (!ext4_has_feature_bigalloc(sb)) 3437 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3438 sbi->s_itb_per_group + 2); 3439 3440 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3441 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3442 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3443 for (i = 0; i < ngroups; i++) { 3444 gdp = ext4_get_group_desc(sb, i, NULL); 3445 b = ext4_block_bitmap(sb, gdp); 3446 if (b >= first_block && b <= last_block) { 3447 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3448 count++; 3449 } 3450 b = ext4_inode_bitmap(sb, gdp); 3451 if (b >= first_block && b <= last_block) { 3452 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3453 count++; 3454 } 3455 b = ext4_inode_table(sb, gdp); 3456 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3457 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3458 int c = EXT4_B2C(sbi, b - first_block); 3459 ext4_set_bit(c, buf); 3460 count++; 3461 } 3462 if (i != grp) 3463 continue; 3464 s = 0; 3465 if (ext4_bg_has_super(sb, grp)) { 3466 ext4_set_bit(s++, buf); 3467 count++; 3468 } 3469 j = ext4_bg_num_gdb(sb, grp); 3470 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3471 ext4_error(sb, "Invalid number of block group " 3472 "descriptor blocks: %d", j); 3473 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3474 } 3475 count += j; 3476 for (; j > 0; j--) 3477 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3478 } 3479 if (!count) 3480 return 0; 3481 return EXT4_CLUSTERS_PER_GROUP(sb) - 3482 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3483 } 3484 3485 /* 3486 * Compute the overhead and stash it in sbi->s_overhead 3487 */ 3488 int ext4_calculate_overhead(struct super_block *sb) 3489 { 3490 struct ext4_sb_info *sbi = EXT4_SB(sb); 3491 struct ext4_super_block *es = sbi->s_es; 3492 struct inode *j_inode; 3493 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3494 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3495 ext4_fsblk_t overhead = 0; 3496 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3497 3498 if (!buf) 3499 return -ENOMEM; 3500 3501 /* 3502 * Compute the overhead (FS structures). This is constant 3503 * for a given filesystem unless the number of block groups 3504 * changes so we cache the previous value until it does. 3505 */ 3506 3507 /* 3508 * All of the blocks before first_data_block are overhead 3509 */ 3510 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3511 3512 /* 3513 * Add the overhead found in each block group 3514 */ 3515 for (i = 0; i < ngroups; i++) { 3516 int blks; 3517 3518 blks = count_overhead(sb, i, buf); 3519 overhead += blks; 3520 if (blks) 3521 memset(buf, 0, PAGE_SIZE); 3522 cond_resched(); 3523 } 3524 3525 /* 3526 * Add the internal journal blocks whether the journal has been 3527 * loaded or not 3528 */ 3529 if (sbi->s_journal && !sbi->journal_bdev) 3530 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3531 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { 3532 j_inode = ext4_get_journal_inode(sb, j_inum); 3533 if (j_inode) { 3534 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3535 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3536 iput(j_inode); 3537 } else { 3538 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3539 } 3540 } 3541 sbi->s_overhead = overhead; 3542 smp_wmb(); 3543 free_page((unsigned long) buf); 3544 return 0; 3545 } 3546 3547 static void ext4_clamp_want_extra_isize(struct super_block *sb) 3548 { 3549 struct ext4_sb_info *sbi = EXT4_SB(sb); 3550 struct ext4_super_block *es = sbi->s_es; 3551 3552 /* determine the minimum size of new large inodes, if present */ 3553 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE && 3554 sbi->s_want_extra_isize == 0) { 3555 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3556 EXT4_GOOD_OLD_INODE_SIZE; 3557 if (ext4_has_feature_extra_isize(sb)) { 3558 if (sbi->s_want_extra_isize < 3559 le16_to_cpu(es->s_want_extra_isize)) 3560 sbi->s_want_extra_isize = 3561 le16_to_cpu(es->s_want_extra_isize); 3562 if (sbi->s_want_extra_isize < 3563 le16_to_cpu(es->s_min_extra_isize)) 3564 sbi->s_want_extra_isize = 3565 le16_to_cpu(es->s_min_extra_isize); 3566 } 3567 } 3568 /* Check if enough inode space is available */ 3569 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 3570 sbi->s_inode_size) { 3571 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3572 EXT4_GOOD_OLD_INODE_SIZE; 3573 ext4_msg(sb, KERN_INFO, 3574 "required extra inode space not available"); 3575 } 3576 } 3577 3578 static void ext4_set_resv_clusters(struct super_block *sb) 3579 { 3580 ext4_fsblk_t resv_clusters; 3581 struct ext4_sb_info *sbi = EXT4_SB(sb); 3582 3583 /* 3584 * There's no need to reserve anything when we aren't using extents. 3585 * The space estimates are exact, there are no unwritten extents, 3586 * hole punching doesn't need new metadata... This is needed especially 3587 * to keep ext2/3 backward compatibility. 3588 */ 3589 if (!ext4_has_feature_extents(sb)) 3590 return; 3591 /* 3592 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3593 * This should cover the situations where we can not afford to run 3594 * out of space like for example punch hole, or converting 3595 * unwritten extents in delalloc path. In most cases such 3596 * allocation would require 1, or 2 blocks, higher numbers are 3597 * very rare. 3598 */ 3599 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3600 sbi->s_cluster_bits); 3601 3602 do_div(resv_clusters, 50); 3603 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3604 3605 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3606 } 3607 3608 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3609 { 3610 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3611 char *orig_data = kstrdup(data, GFP_KERNEL); 3612 struct buffer_head *bh; 3613 struct ext4_super_block *es = NULL; 3614 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3615 ext4_fsblk_t block; 3616 ext4_fsblk_t sb_block = get_sb_block(&data); 3617 ext4_fsblk_t logical_sb_block; 3618 unsigned long offset = 0; 3619 unsigned long journal_devnum = 0; 3620 unsigned long def_mount_opts; 3621 struct inode *root; 3622 const char *descr; 3623 int ret = -ENOMEM; 3624 int blocksize, clustersize; 3625 unsigned int db_count; 3626 unsigned int i; 3627 int needs_recovery, has_huge_files, has_bigalloc; 3628 __u64 blocks_count; 3629 int err = 0; 3630 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3631 ext4_group_t first_not_zeroed; 3632 3633 if ((data && !orig_data) || !sbi) 3634 goto out_free_base; 3635 3636 sbi->s_daxdev = dax_dev; 3637 sbi->s_blockgroup_lock = 3638 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3639 if (!sbi->s_blockgroup_lock) 3640 goto out_free_base; 3641 3642 sb->s_fs_info = sbi; 3643 sbi->s_sb = sb; 3644 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3645 sbi->s_sb_block = sb_block; 3646 if (sb->s_bdev->bd_part) 3647 sbi->s_sectors_written_start = 3648 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3649 3650 /* Cleanup superblock name */ 3651 strreplace(sb->s_id, '/', '!'); 3652 3653 /* -EINVAL is default */ 3654 ret = -EINVAL; 3655 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3656 if (!blocksize) { 3657 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3658 goto out_fail; 3659 } 3660 3661 /* 3662 * The ext4 superblock will not be buffer aligned for other than 1kB 3663 * block sizes. We need to calculate the offset from buffer start. 3664 */ 3665 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3666 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3667 offset = do_div(logical_sb_block, blocksize); 3668 } else { 3669 logical_sb_block = sb_block; 3670 } 3671 3672 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3673 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3674 goto out_fail; 3675 } 3676 /* 3677 * Note: s_es must be initialized as soon as possible because 3678 * some ext4 macro-instructions depend on its value 3679 */ 3680 es = (struct ext4_super_block *) (bh->b_data + offset); 3681 sbi->s_es = es; 3682 sb->s_magic = le16_to_cpu(es->s_magic); 3683 if (sb->s_magic != EXT4_SUPER_MAGIC) 3684 goto cantfind_ext4; 3685 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3686 3687 /* Warn if metadata_csum and gdt_csum are both set. */ 3688 if (ext4_has_feature_metadata_csum(sb) && 3689 ext4_has_feature_gdt_csum(sb)) 3690 ext4_warning(sb, "metadata_csum and uninit_bg are " 3691 "redundant flags; please run fsck."); 3692 3693 /* Check for a known checksum algorithm */ 3694 if (!ext4_verify_csum_type(sb, es)) { 3695 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3696 "unknown checksum algorithm."); 3697 silent = 1; 3698 goto cantfind_ext4; 3699 } 3700 3701 /* Load the checksum driver */ 3702 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3703 if (IS_ERR(sbi->s_chksum_driver)) { 3704 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3705 ret = PTR_ERR(sbi->s_chksum_driver); 3706 sbi->s_chksum_driver = NULL; 3707 goto failed_mount; 3708 } 3709 3710 /* Check superblock checksum */ 3711 if (!ext4_superblock_csum_verify(sb, es)) { 3712 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3713 "invalid superblock checksum. Run e2fsck?"); 3714 silent = 1; 3715 ret = -EFSBADCRC; 3716 goto cantfind_ext4; 3717 } 3718 3719 /* Precompute checksum seed for all metadata */ 3720 if (ext4_has_feature_csum_seed(sb)) 3721 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3722 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3723 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3724 sizeof(es->s_uuid)); 3725 3726 /* Set defaults before we parse the mount options */ 3727 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3728 set_opt(sb, INIT_INODE_TABLE); 3729 if (def_mount_opts & EXT4_DEFM_DEBUG) 3730 set_opt(sb, DEBUG); 3731 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3732 set_opt(sb, GRPID); 3733 if (def_mount_opts & EXT4_DEFM_UID16) 3734 set_opt(sb, NO_UID32); 3735 /* xattr user namespace & acls are now defaulted on */ 3736 set_opt(sb, XATTR_USER); 3737 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3738 set_opt(sb, POSIX_ACL); 3739 #endif 3740 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 3741 if (ext4_has_metadata_csum(sb)) 3742 set_opt(sb, JOURNAL_CHECKSUM); 3743 3744 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3745 set_opt(sb, JOURNAL_DATA); 3746 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3747 set_opt(sb, ORDERED_DATA); 3748 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3749 set_opt(sb, WRITEBACK_DATA); 3750 3751 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3752 set_opt(sb, ERRORS_PANIC); 3753 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3754 set_opt(sb, ERRORS_CONT); 3755 else 3756 set_opt(sb, ERRORS_RO); 3757 /* block_validity enabled by default; disable with noblock_validity */ 3758 set_opt(sb, BLOCK_VALIDITY); 3759 if (def_mount_opts & EXT4_DEFM_DISCARD) 3760 set_opt(sb, DISCARD); 3761 3762 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3763 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3764 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3765 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3766 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3767 3768 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3769 set_opt(sb, BARRIER); 3770 3771 /* 3772 * enable delayed allocation by default 3773 * Use -o nodelalloc to turn it off 3774 */ 3775 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3776 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3777 set_opt(sb, DELALLOC); 3778 3779 /* 3780 * set default s_li_wait_mult for lazyinit, for the case there is 3781 * no mount option specified. 3782 */ 3783 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3784 3785 if (sbi->s_es->s_mount_opts[0]) { 3786 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 3787 sizeof(sbi->s_es->s_mount_opts), 3788 GFP_KERNEL); 3789 if (!s_mount_opts) 3790 goto failed_mount; 3791 if (!parse_options(s_mount_opts, sb, &journal_devnum, 3792 &journal_ioprio, 0)) { 3793 ext4_msg(sb, KERN_WARNING, 3794 "failed to parse options in superblock: %s", 3795 s_mount_opts); 3796 } 3797 kfree(s_mount_opts); 3798 } 3799 sbi->s_def_mount_opt = sbi->s_mount_opt; 3800 if (!parse_options((char *) data, sb, &journal_devnum, 3801 &journal_ioprio, 0)) 3802 goto failed_mount; 3803 3804 #ifdef CONFIG_UNICODE 3805 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 3806 const struct ext4_sb_encodings *encoding_info; 3807 struct unicode_map *encoding; 3808 __u16 encoding_flags; 3809 3810 if (ext4_has_feature_encrypt(sb)) { 3811 ext4_msg(sb, KERN_ERR, 3812 "Can't mount with encoding and encryption"); 3813 goto failed_mount; 3814 } 3815 3816 if (ext4_sb_read_encoding(es, &encoding_info, 3817 &encoding_flags)) { 3818 ext4_msg(sb, KERN_ERR, 3819 "Encoding requested by superblock is unknown"); 3820 goto failed_mount; 3821 } 3822 3823 encoding = utf8_load(encoding_info->version); 3824 if (IS_ERR(encoding)) { 3825 ext4_msg(sb, KERN_ERR, 3826 "can't mount with superblock charset: %s-%s " 3827 "not supported by the kernel. flags: 0x%x.", 3828 encoding_info->name, encoding_info->version, 3829 encoding_flags); 3830 goto failed_mount; 3831 } 3832 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 3833 "%s-%s with flags 0x%hx", encoding_info->name, 3834 encoding_info->version?:"\b", encoding_flags); 3835 3836 sbi->s_encoding = encoding; 3837 sbi->s_encoding_flags = encoding_flags; 3838 } 3839 #endif 3840 3841 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3842 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3843 "with data=journal disables delayed " 3844 "allocation and O_DIRECT support!\n"); 3845 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3846 ext4_msg(sb, KERN_ERR, "can't mount with " 3847 "both data=journal and delalloc"); 3848 goto failed_mount; 3849 } 3850 if (test_opt(sb, DIOREAD_NOLOCK)) { 3851 ext4_msg(sb, KERN_ERR, "can't mount with " 3852 "both data=journal and dioread_nolock"); 3853 goto failed_mount; 3854 } 3855 if (test_opt(sb, DAX)) { 3856 ext4_msg(sb, KERN_ERR, "can't mount with " 3857 "both data=journal and dax"); 3858 goto failed_mount; 3859 } 3860 if (ext4_has_feature_encrypt(sb)) { 3861 ext4_msg(sb, KERN_WARNING, 3862 "encrypted files will use data=ordered " 3863 "instead of data journaling mode"); 3864 } 3865 if (test_opt(sb, DELALLOC)) 3866 clear_opt(sb, DELALLOC); 3867 } else { 3868 sb->s_iflags |= SB_I_CGROUPWB; 3869 } 3870 3871 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3872 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 3873 3874 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3875 (ext4_has_compat_features(sb) || 3876 ext4_has_ro_compat_features(sb) || 3877 ext4_has_incompat_features(sb))) 3878 ext4_msg(sb, KERN_WARNING, 3879 "feature flags set on rev 0 fs, " 3880 "running e2fsck is recommended"); 3881 3882 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3883 set_opt2(sb, HURD_COMPAT); 3884 if (ext4_has_feature_64bit(sb)) { 3885 ext4_msg(sb, KERN_ERR, 3886 "The Hurd can't support 64-bit file systems"); 3887 goto failed_mount; 3888 } 3889 3890 /* 3891 * ea_inode feature uses l_i_version field which is not 3892 * available in HURD_COMPAT mode. 3893 */ 3894 if (ext4_has_feature_ea_inode(sb)) { 3895 ext4_msg(sb, KERN_ERR, 3896 "ea_inode feature is not supported for Hurd"); 3897 goto failed_mount; 3898 } 3899 } 3900 3901 if (IS_EXT2_SB(sb)) { 3902 if (ext2_feature_set_ok(sb)) 3903 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3904 "using the ext4 subsystem"); 3905 else { 3906 /* 3907 * If we're probing be silent, if this looks like 3908 * it's actually an ext[34] filesystem. 3909 */ 3910 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3911 goto failed_mount; 3912 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3913 "to feature incompatibilities"); 3914 goto failed_mount; 3915 } 3916 } 3917 3918 if (IS_EXT3_SB(sb)) { 3919 if (ext3_feature_set_ok(sb)) 3920 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3921 "using the ext4 subsystem"); 3922 else { 3923 /* 3924 * If we're probing be silent, if this looks like 3925 * it's actually an ext4 filesystem. 3926 */ 3927 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3928 goto failed_mount; 3929 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3930 "to feature incompatibilities"); 3931 goto failed_mount; 3932 } 3933 } 3934 3935 /* 3936 * Check feature flags regardless of the revision level, since we 3937 * previously didn't change the revision level when setting the flags, 3938 * so there is a chance incompat flags are set on a rev 0 filesystem. 3939 */ 3940 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 3941 goto failed_mount; 3942 3943 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3944 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3945 blocksize > EXT4_MAX_BLOCK_SIZE) { 3946 ext4_msg(sb, KERN_ERR, 3947 "Unsupported filesystem blocksize %d (%d log_block_size)", 3948 blocksize, le32_to_cpu(es->s_log_block_size)); 3949 goto failed_mount; 3950 } 3951 if (le32_to_cpu(es->s_log_block_size) > 3952 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3953 ext4_msg(sb, KERN_ERR, 3954 "Invalid log block size: %u", 3955 le32_to_cpu(es->s_log_block_size)); 3956 goto failed_mount; 3957 } 3958 if (le32_to_cpu(es->s_log_cluster_size) > 3959 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3960 ext4_msg(sb, KERN_ERR, 3961 "Invalid log cluster size: %u", 3962 le32_to_cpu(es->s_log_cluster_size)); 3963 goto failed_mount; 3964 } 3965 3966 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 3967 ext4_msg(sb, KERN_ERR, 3968 "Number of reserved GDT blocks insanely large: %d", 3969 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 3970 goto failed_mount; 3971 } 3972 3973 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 3974 if (ext4_has_feature_inline_data(sb)) { 3975 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 3976 " that may contain inline data"); 3977 goto failed_mount; 3978 } 3979 if (!bdev_dax_supported(sb->s_bdev, blocksize)) { 3980 ext4_msg(sb, KERN_ERR, 3981 "DAX unsupported by block device."); 3982 goto failed_mount; 3983 } 3984 } 3985 3986 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 3987 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 3988 es->s_encryption_level); 3989 goto failed_mount; 3990 } 3991 3992 if (sb->s_blocksize != blocksize) { 3993 /* Validate the filesystem blocksize */ 3994 if (!sb_set_blocksize(sb, blocksize)) { 3995 ext4_msg(sb, KERN_ERR, "bad block size %d", 3996 blocksize); 3997 goto failed_mount; 3998 } 3999 4000 brelse(bh); 4001 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 4002 offset = do_div(logical_sb_block, blocksize); 4003 bh = sb_bread_unmovable(sb, logical_sb_block); 4004 if (!bh) { 4005 ext4_msg(sb, KERN_ERR, 4006 "Can't read superblock on 2nd try"); 4007 goto failed_mount; 4008 } 4009 es = (struct ext4_super_block *)(bh->b_data + offset); 4010 sbi->s_es = es; 4011 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4012 ext4_msg(sb, KERN_ERR, 4013 "Magic mismatch, very weird!"); 4014 goto failed_mount; 4015 } 4016 } 4017 4018 has_huge_files = ext4_has_feature_huge_file(sb); 4019 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4020 has_huge_files); 4021 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4022 4023 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4024 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4025 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4026 } else { 4027 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4028 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4029 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4030 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4031 sbi->s_first_ino); 4032 goto failed_mount; 4033 } 4034 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4035 (!is_power_of_2(sbi->s_inode_size)) || 4036 (sbi->s_inode_size > blocksize)) { 4037 ext4_msg(sb, KERN_ERR, 4038 "unsupported inode size: %d", 4039 sbi->s_inode_size); 4040 goto failed_mount; 4041 } 4042 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 4043 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 4044 } 4045 4046 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4047 if (ext4_has_feature_64bit(sb)) { 4048 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4049 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4050 !is_power_of_2(sbi->s_desc_size)) { 4051 ext4_msg(sb, KERN_ERR, 4052 "unsupported descriptor size %lu", 4053 sbi->s_desc_size); 4054 goto failed_mount; 4055 } 4056 } else 4057 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4058 4059 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4060 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4061 4062 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4063 if (sbi->s_inodes_per_block == 0) 4064 goto cantfind_ext4; 4065 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4066 sbi->s_inodes_per_group > blocksize * 8) { 4067 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4068 sbi->s_blocks_per_group); 4069 goto failed_mount; 4070 } 4071 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4072 sbi->s_inodes_per_block; 4073 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4074 sbi->s_sbh = bh; 4075 sbi->s_mount_state = le16_to_cpu(es->s_state); 4076 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4077 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4078 4079 for (i = 0; i < 4; i++) 4080 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4081 sbi->s_def_hash_version = es->s_def_hash_version; 4082 if (ext4_has_feature_dir_index(sb)) { 4083 i = le32_to_cpu(es->s_flags); 4084 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4085 sbi->s_hash_unsigned = 3; 4086 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4087 #ifdef __CHAR_UNSIGNED__ 4088 if (!sb_rdonly(sb)) 4089 es->s_flags |= 4090 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4091 sbi->s_hash_unsigned = 3; 4092 #else 4093 if (!sb_rdonly(sb)) 4094 es->s_flags |= 4095 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4096 #endif 4097 } 4098 } 4099 4100 /* Handle clustersize */ 4101 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4102 has_bigalloc = ext4_has_feature_bigalloc(sb); 4103 if (has_bigalloc) { 4104 if (clustersize < blocksize) { 4105 ext4_msg(sb, KERN_ERR, 4106 "cluster size (%d) smaller than " 4107 "block size (%d)", clustersize, blocksize); 4108 goto failed_mount; 4109 } 4110 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4111 le32_to_cpu(es->s_log_block_size); 4112 sbi->s_clusters_per_group = 4113 le32_to_cpu(es->s_clusters_per_group); 4114 if (sbi->s_clusters_per_group > blocksize * 8) { 4115 ext4_msg(sb, KERN_ERR, 4116 "#clusters per group too big: %lu", 4117 sbi->s_clusters_per_group); 4118 goto failed_mount; 4119 } 4120 if (sbi->s_blocks_per_group != 4121 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4122 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4123 "clusters per group (%lu) inconsistent", 4124 sbi->s_blocks_per_group, 4125 sbi->s_clusters_per_group); 4126 goto failed_mount; 4127 } 4128 } else { 4129 if (clustersize != blocksize) { 4130 ext4_msg(sb, KERN_ERR, 4131 "fragment/cluster size (%d) != " 4132 "block size (%d)", clustersize, blocksize); 4133 goto failed_mount; 4134 } 4135 if (sbi->s_blocks_per_group > blocksize * 8) { 4136 ext4_msg(sb, KERN_ERR, 4137 "#blocks per group too big: %lu", 4138 sbi->s_blocks_per_group); 4139 goto failed_mount; 4140 } 4141 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4142 sbi->s_cluster_bits = 0; 4143 } 4144 sbi->s_cluster_ratio = clustersize / blocksize; 4145 4146 /* Do we have standard group size of clustersize * 8 blocks ? */ 4147 if (sbi->s_blocks_per_group == clustersize << 3) 4148 set_opt2(sb, STD_GROUP_SIZE); 4149 4150 /* 4151 * Test whether we have more sectors than will fit in sector_t, 4152 * and whether the max offset is addressable by the page cache. 4153 */ 4154 err = generic_check_addressable(sb->s_blocksize_bits, 4155 ext4_blocks_count(es)); 4156 if (err) { 4157 ext4_msg(sb, KERN_ERR, "filesystem" 4158 " too large to mount safely on this system"); 4159 goto failed_mount; 4160 } 4161 4162 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4163 goto cantfind_ext4; 4164 4165 /* check blocks count against device size */ 4166 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 4167 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4168 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4169 "exceeds size of device (%llu blocks)", 4170 ext4_blocks_count(es), blocks_count); 4171 goto failed_mount; 4172 } 4173 4174 /* 4175 * It makes no sense for the first data block to be beyond the end 4176 * of the filesystem. 4177 */ 4178 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4179 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4180 "block %u is beyond end of filesystem (%llu)", 4181 le32_to_cpu(es->s_first_data_block), 4182 ext4_blocks_count(es)); 4183 goto failed_mount; 4184 } 4185 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4186 (sbi->s_cluster_ratio == 1)) { 4187 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4188 "block is 0 with a 1k block and cluster size"); 4189 goto failed_mount; 4190 } 4191 4192 blocks_count = (ext4_blocks_count(es) - 4193 le32_to_cpu(es->s_first_data_block) + 4194 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4195 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4196 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4197 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 4198 "(block count %llu, first data block %u, " 4199 "blocks per group %lu)", sbi->s_groups_count, 4200 ext4_blocks_count(es), 4201 le32_to_cpu(es->s_first_data_block), 4202 EXT4_BLOCKS_PER_GROUP(sb)); 4203 goto failed_mount; 4204 } 4205 sbi->s_groups_count = blocks_count; 4206 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4207 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4208 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4209 le32_to_cpu(es->s_inodes_count)) { 4210 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4211 le32_to_cpu(es->s_inodes_count), 4212 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4213 ret = -EINVAL; 4214 goto failed_mount; 4215 } 4216 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4217 EXT4_DESC_PER_BLOCK(sb); 4218 if (ext4_has_feature_meta_bg(sb)) { 4219 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4220 ext4_msg(sb, KERN_WARNING, 4221 "first meta block group too large: %u " 4222 "(group descriptor block count %u)", 4223 le32_to_cpu(es->s_first_meta_bg), db_count); 4224 goto failed_mount; 4225 } 4226 } 4227 sbi->s_group_desc = kvmalloc_array(db_count, 4228 sizeof(struct buffer_head *), 4229 GFP_KERNEL); 4230 if (sbi->s_group_desc == NULL) { 4231 ext4_msg(sb, KERN_ERR, "not enough memory"); 4232 ret = -ENOMEM; 4233 goto failed_mount; 4234 } 4235 4236 bgl_lock_init(sbi->s_blockgroup_lock); 4237 4238 /* Pre-read the descriptors into the buffer cache */ 4239 for (i = 0; i < db_count; i++) { 4240 block = descriptor_loc(sb, logical_sb_block, i); 4241 sb_breadahead(sb, block); 4242 } 4243 4244 for (i = 0; i < db_count; i++) { 4245 block = descriptor_loc(sb, logical_sb_block, i); 4246 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 4247 if (!sbi->s_group_desc[i]) { 4248 ext4_msg(sb, KERN_ERR, 4249 "can't read group descriptor %d", i); 4250 db_count = i; 4251 goto failed_mount2; 4252 } 4253 } 4254 sbi->s_gdb_count = db_count; 4255 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4256 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4257 ret = -EFSCORRUPTED; 4258 goto failed_mount2; 4259 } 4260 4261 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4262 4263 /* Register extent status tree shrinker */ 4264 if (ext4_es_register_shrinker(sbi)) 4265 goto failed_mount3; 4266 4267 sbi->s_stripe = ext4_get_stripe_size(sbi); 4268 sbi->s_extent_max_zeroout_kb = 32; 4269 4270 /* 4271 * set up enough so that it can read an inode 4272 */ 4273 sb->s_op = &ext4_sops; 4274 sb->s_export_op = &ext4_export_ops; 4275 sb->s_xattr = ext4_xattr_handlers; 4276 #ifdef CONFIG_FS_ENCRYPTION 4277 sb->s_cop = &ext4_cryptops; 4278 #endif 4279 #ifdef CONFIG_FS_VERITY 4280 sb->s_vop = &ext4_verityops; 4281 #endif 4282 #ifdef CONFIG_QUOTA 4283 sb->dq_op = &ext4_quota_operations; 4284 if (ext4_has_feature_quota(sb)) 4285 sb->s_qcop = &dquot_quotactl_sysfile_ops; 4286 else 4287 sb->s_qcop = &ext4_qctl_operations; 4288 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4289 #endif 4290 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4291 4292 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 4293 mutex_init(&sbi->s_orphan_lock); 4294 4295 sb->s_root = NULL; 4296 4297 needs_recovery = (es->s_last_orphan != 0 || 4298 ext4_has_feature_journal_needs_recovery(sb)); 4299 4300 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4301 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 4302 goto failed_mount3a; 4303 4304 /* 4305 * The first inode we look at is the journal inode. Don't try 4306 * root first: it may be modified in the journal! 4307 */ 4308 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 4309 err = ext4_load_journal(sb, es, journal_devnum); 4310 if (err) 4311 goto failed_mount3a; 4312 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4313 ext4_has_feature_journal_needs_recovery(sb)) { 4314 ext4_msg(sb, KERN_ERR, "required journal recovery " 4315 "suppressed and not mounted read-only"); 4316 goto failed_mount_wq; 4317 } else { 4318 /* Nojournal mode, all journal mount options are illegal */ 4319 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 4320 ext4_msg(sb, KERN_ERR, "can't mount with " 4321 "journal_checksum, fs mounted w/o journal"); 4322 goto failed_mount_wq; 4323 } 4324 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4325 ext4_msg(sb, KERN_ERR, "can't mount with " 4326 "journal_async_commit, fs mounted w/o journal"); 4327 goto failed_mount_wq; 4328 } 4329 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 4330 ext4_msg(sb, KERN_ERR, "can't mount with " 4331 "commit=%lu, fs mounted w/o journal", 4332 sbi->s_commit_interval / HZ); 4333 goto failed_mount_wq; 4334 } 4335 if (EXT4_MOUNT_DATA_FLAGS & 4336 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 4337 ext4_msg(sb, KERN_ERR, "can't mount with " 4338 "data=, fs mounted w/o journal"); 4339 goto failed_mount_wq; 4340 } 4341 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 4342 clear_opt(sb, JOURNAL_CHECKSUM); 4343 clear_opt(sb, DATA_FLAGS); 4344 sbi->s_journal = NULL; 4345 needs_recovery = 0; 4346 goto no_journal; 4347 } 4348 4349 if (ext4_has_feature_64bit(sb) && 4350 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4351 JBD2_FEATURE_INCOMPAT_64BIT)) { 4352 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4353 goto failed_mount_wq; 4354 } 4355 4356 if (!set_journal_csum_feature_set(sb)) { 4357 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4358 "feature set"); 4359 goto failed_mount_wq; 4360 } 4361 4362 /* We have now updated the journal if required, so we can 4363 * validate the data journaling mode. */ 4364 switch (test_opt(sb, DATA_FLAGS)) { 4365 case 0: 4366 /* No mode set, assume a default based on the journal 4367 * capabilities: ORDERED_DATA if the journal can 4368 * cope, else JOURNAL_DATA 4369 */ 4370 if (jbd2_journal_check_available_features 4371 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4372 set_opt(sb, ORDERED_DATA); 4373 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4374 } else { 4375 set_opt(sb, JOURNAL_DATA); 4376 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4377 } 4378 break; 4379 4380 case EXT4_MOUNT_ORDERED_DATA: 4381 case EXT4_MOUNT_WRITEBACK_DATA: 4382 if (!jbd2_journal_check_available_features 4383 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4384 ext4_msg(sb, KERN_ERR, "Journal does not support " 4385 "requested data journaling mode"); 4386 goto failed_mount_wq; 4387 } 4388 default: 4389 break; 4390 } 4391 4392 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4393 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4394 ext4_msg(sb, KERN_ERR, "can't mount with " 4395 "journal_async_commit in data=ordered mode"); 4396 goto failed_mount_wq; 4397 } 4398 4399 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4400 4401 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4402 4403 no_journal: 4404 if (!test_opt(sb, NO_MBCACHE)) { 4405 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 4406 if (!sbi->s_ea_block_cache) { 4407 ext4_msg(sb, KERN_ERR, 4408 "Failed to create ea_block_cache"); 4409 goto failed_mount_wq; 4410 } 4411 4412 if (ext4_has_feature_ea_inode(sb)) { 4413 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4414 if (!sbi->s_ea_inode_cache) { 4415 ext4_msg(sb, KERN_ERR, 4416 "Failed to create ea_inode_cache"); 4417 goto failed_mount_wq; 4418 } 4419 } 4420 } 4421 4422 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && 4423 (blocksize != PAGE_SIZE)) { 4424 ext4_msg(sb, KERN_ERR, 4425 "Unsupported blocksize for fs encryption"); 4426 goto failed_mount_wq; 4427 } 4428 4429 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4430 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4431 goto failed_mount_wq; 4432 } 4433 4434 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4435 !ext4_has_feature_encrypt(sb)) { 4436 ext4_set_feature_encrypt(sb); 4437 ext4_commit_super(sb, 1); 4438 } 4439 4440 /* 4441 * Get the # of file system overhead blocks from the 4442 * superblock if present. 4443 */ 4444 if (es->s_overhead_clusters) 4445 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4446 else { 4447 err = ext4_calculate_overhead(sb); 4448 if (err) 4449 goto failed_mount_wq; 4450 } 4451 4452 /* 4453 * The maximum number of concurrent works can be high and 4454 * concurrency isn't really necessary. Limit it to 1. 4455 */ 4456 EXT4_SB(sb)->rsv_conversion_wq = 4457 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4458 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4459 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4460 ret = -ENOMEM; 4461 goto failed_mount4; 4462 } 4463 4464 /* 4465 * The jbd2_journal_load will have done any necessary log recovery, 4466 * so we can safely mount the rest of the filesystem now. 4467 */ 4468 4469 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 4470 if (IS_ERR(root)) { 4471 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4472 ret = PTR_ERR(root); 4473 root = NULL; 4474 goto failed_mount4; 4475 } 4476 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4477 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4478 iput(root); 4479 goto failed_mount4; 4480 } 4481 4482 #ifdef CONFIG_UNICODE 4483 if (sbi->s_encoding) 4484 sb->s_d_op = &ext4_dentry_ops; 4485 #endif 4486 4487 sb->s_root = d_make_root(root); 4488 if (!sb->s_root) { 4489 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4490 ret = -ENOMEM; 4491 goto failed_mount4; 4492 } 4493 4494 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4495 if (ret == -EROFS) { 4496 sb->s_flags |= SB_RDONLY; 4497 ret = 0; 4498 } else if (ret) 4499 goto failed_mount4a; 4500 4501 ext4_clamp_want_extra_isize(sb); 4502 4503 ext4_set_resv_clusters(sb); 4504 4505 err = ext4_setup_system_zone(sb); 4506 if (err) { 4507 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4508 "zone (%d)", err); 4509 goto failed_mount4a; 4510 } 4511 4512 ext4_ext_init(sb); 4513 err = ext4_mb_init(sb); 4514 if (err) { 4515 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4516 err); 4517 goto failed_mount5; 4518 } 4519 4520 block = ext4_count_free_clusters(sb); 4521 ext4_free_blocks_count_set(sbi->s_es, 4522 EXT4_C2B(sbi, block)); 4523 ext4_superblock_csum_set(sb); 4524 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4525 GFP_KERNEL); 4526 if (!err) { 4527 unsigned long freei = ext4_count_free_inodes(sb); 4528 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4529 ext4_superblock_csum_set(sb); 4530 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4531 GFP_KERNEL); 4532 } 4533 if (!err) 4534 err = percpu_counter_init(&sbi->s_dirs_counter, 4535 ext4_count_dirs(sb), GFP_KERNEL); 4536 if (!err) 4537 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4538 GFP_KERNEL); 4539 if (!err) 4540 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 4541 4542 if (err) { 4543 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4544 goto failed_mount6; 4545 } 4546 4547 if (ext4_has_feature_flex_bg(sb)) 4548 if (!ext4_fill_flex_info(sb)) { 4549 ext4_msg(sb, KERN_ERR, 4550 "unable to initialize " 4551 "flex_bg meta info!"); 4552 goto failed_mount6; 4553 } 4554 4555 err = ext4_register_li_request(sb, first_not_zeroed); 4556 if (err) 4557 goto failed_mount6; 4558 4559 err = ext4_register_sysfs(sb); 4560 if (err) 4561 goto failed_mount7; 4562 4563 #ifdef CONFIG_QUOTA 4564 /* Enable quota usage during mount. */ 4565 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 4566 err = ext4_enable_quotas(sb); 4567 if (err) 4568 goto failed_mount8; 4569 } 4570 #endif /* CONFIG_QUOTA */ 4571 4572 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4573 ext4_orphan_cleanup(sb, es); 4574 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4575 if (needs_recovery) { 4576 ext4_msg(sb, KERN_INFO, "recovery complete"); 4577 ext4_mark_recovery_complete(sb, es); 4578 } 4579 if (EXT4_SB(sb)->s_journal) { 4580 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4581 descr = " journalled data mode"; 4582 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4583 descr = " ordered data mode"; 4584 else 4585 descr = " writeback data mode"; 4586 } else 4587 descr = "out journal"; 4588 4589 if (test_opt(sb, DISCARD)) { 4590 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4591 if (!blk_queue_discard(q)) 4592 ext4_msg(sb, KERN_WARNING, 4593 "mounting with \"discard\" option, but " 4594 "the device does not support discard"); 4595 } 4596 4597 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4598 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4599 "Opts: %.*s%s%s", descr, 4600 (int) sizeof(sbi->s_es->s_mount_opts), 4601 sbi->s_es->s_mount_opts, 4602 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4603 4604 if (es->s_error_count) 4605 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4606 4607 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4608 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4609 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4610 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4611 4612 kfree(orig_data); 4613 return 0; 4614 4615 cantfind_ext4: 4616 if (!silent) 4617 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4618 goto failed_mount; 4619 4620 #ifdef CONFIG_QUOTA 4621 failed_mount8: 4622 ext4_unregister_sysfs(sb); 4623 #endif 4624 failed_mount7: 4625 ext4_unregister_li_request(sb); 4626 failed_mount6: 4627 ext4_mb_release(sb); 4628 if (sbi->s_flex_groups) 4629 kvfree(sbi->s_flex_groups); 4630 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4631 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4632 percpu_counter_destroy(&sbi->s_dirs_counter); 4633 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4634 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 4635 failed_mount5: 4636 ext4_ext_release(sb); 4637 ext4_release_system_zone(sb); 4638 failed_mount4a: 4639 dput(sb->s_root); 4640 sb->s_root = NULL; 4641 failed_mount4: 4642 ext4_msg(sb, KERN_ERR, "mount failed"); 4643 if (EXT4_SB(sb)->rsv_conversion_wq) 4644 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4645 failed_mount_wq: 4646 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4647 sbi->s_ea_inode_cache = NULL; 4648 4649 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 4650 sbi->s_ea_block_cache = NULL; 4651 4652 if (sbi->s_journal) { 4653 jbd2_journal_destroy(sbi->s_journal); 4654 sbi->s_journal = NULL; 4655 } 4656 failed_mount3a: 4657 ext4_es_unregister_shrinker(sbi); 4658 failed_mount3: 4659 del_timer_sync(&sbi->s_err_report); 4660 if (sbi->s_mmp_tsk) 4661 kthread_stop(sbi->s_mmp_tsk); 4662 failed_mount2: 4663 for (i = 0; i < db_count; i++) 4664 brelse(sbi->s_group_desc[i]); 4665 kvfree(sbi->s_group_desc); 4666 failed_mount: 4667 if (sbi->s_chksum_driver) 4668 crypto_free_shash(sbi->s_chksum_driver); 4669 4670 #ifdef CONFIG_UNICODE 4671 utf8_unload(sbi->s_encoding); 4672 #endif 4673 4674 #ifdef CONFIG_QUOTA 4675 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4676 kfree(get_qf_name(sb, sbi, i)); 4677 #endif 4678 ext4_blkdev_remove(sbi); 4679 brelse(bh); 4680 out_fail: 4681 sb->s_fs_info = NULL; 4682 kfree(sbi->s_blockgroup_lock); 4683 out_free_base: 4684 kfree(sbi); 4685 kfree(orig_data); 4686 fs_put_dax(dax_dev); 4687 return err ? err : ret; 4688 } 4689 4690 /* 4691 * Setup any per-fs journal parameters now. We'll do this both on 4692 * initial mount, once the journal has been initialised but before we've 4693 * done any recovery; and again on any subsequent remount. 4694 */ 4695 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4696 { 4697 struct ext4_sb_info *sbi = EXT4_SB(sb); 4698 4699 journal->j_commit_interval = sbi->s_commit_interval; 4700 journal->j_min_batch_time = sbi->s_min_batch_time; 4701 journal->j_max_batch_time = sbi->s_max_batch_time; 4702 4703 write_lock(&journal->j_state_lock); 4704 if (test_opt(sb, BARRIER)) 4705 journal->j_flags |= JBD2_BARRIER; 4706 else 4707 journal->j_flags &= ~JBD2_BARRIER; 4708 if (test_opt(sb, DATA_ERR_ABORT)) 4709 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4710 else 4711 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4712 write_unlock(&journal->j_state_lock); 4713 } 4714 4715 static struct inode *ext4_get_journal_inode(struct super_block *sb, 4716 unsigned int journal_inum) 4717 { 4718 struct inode *journal_inode; 4719 4720 /* 4721 * Test for the existence of a valid inode on disk. Bad things 4722 * happen if we iget() an unused inode, as the subsequent iput() 4723 * will try to delete it. 4724 */ 4725 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 4726 if (IS_ERR(journal_inode)) { 4727 ext4_msg(sb, KERN_ERR, "no journal found"); 4728 return NULL; 4729 } 4730 if (!journal_inode->i_nlink) { 4731 make_bad_inode(journal_inode); 4732 iput(journal_inode); 4733 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4734 return NULL; 4735 } 4736 4737 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4738 journal_inode, journal_inode->i_size); 4739 if (!S_ISREG(journal_inode->i_mode)) { 4740 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4741 iput(journal_inode); 4742 return NULL; 4743 } 4744 return journal_inode; 4745 } 4746 4747 static journal_t *ext4_get_journal(struct super_block *sb, 4748 unsigned int journal_inum) 4749 { 4750 struct inode *journal_inode; 4751 journal_t *journal; 4752 4753 BUG_ON(!ext4_has_feature_journal(sb)); 4754 4755 journal_inode = ext4_get_journal_inode(sb, journal_inum); 4756 if (!journal_inode) 4757 return NULL; 4758 4759 journal = jbd2_journal_init_inode(journal_inode); 4760 if (!journal) { 4761 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4762 iput(journal_inode); 4763 return NULL; 4764 } 4765 journal->j_private = sb; 4766 ext4_init_journal_params(sb, journal); 4767 return journal; 4768 } 4769 4770 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4771 dev_t j_dev) 4772 { 4773 struct buffer_head *bh; 4774 journal_t *journal; 4775 ext4_fsblk_t start; 4776 ext4_fsblk_t len; 4777 int hblock, blocksize; 4778 ext4_fsblk_t sb_block; 4779 unsigned long offset; 4780 struct ext4_super_block *es; 4781 struct block_device *bdev; 4782 4783 BUG_ON(!ext4_has_feature_journal(sb)); 4784 4785 bdev = ext4_blkdev_get(j_dev, sb); 4786 if (bdev == NULL) 4787 return NULL; 4788 4789 blocksize = sb->s_blocksize; 4790 hblock = bdev_logical_block_size(bdev); 4791 if (blocksize < hblock) { 4792 ext4_msg(sb, KERN_ERR, 4793 "blocksize too small for journal device"); 4794 goto out_bdev; 4795 } 4796 4797 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4798 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4799 set_blocksize(bdev, blocksize); 4800 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4801 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4802 "external journal"); 4803 goto out_bdev; 4804 } 4805 4806 es = (struct ext4_super_block *) (bh->b_data + offset); 4807 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4808 !(le32_to_cpu(es->s_feature_incompat) & 4809 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4810 ext4_msg(sb, KERN_ERR, "external journal has " 4811 "bad superblock"); 4812 brelse(bh); 4813 goto out_bdev; 4814 } 4815 4816 if ((le32_to_cpu(es->s_feature_ro_compat) & 4817 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4818 es->s_checksum != ext4_superblock_csum(sb, es)) { 4819 ext4_msg(sb, KERN_ERR, "external journal has " 4820 "corrupt superblock"); 4821 brelse(bh); 4822 goto out_bdev; 4823 } 4824 4825 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4826 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4827 brelse(bh); 4828 goto out_bdev; 4829 } 4830 4831 len = ext4_blocks_count(es); 4832 start = sb_block + 1; 4833 brelse(bh); /* we're done with the superblock */ 4834 4835 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4836 start, len, blocksize); 4837 if (!journal) { 4838 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4839 goto out_bdev; 4840 } 4841 journal->j_private = sb; 4842 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4843 wait_on_buffer(journal->j_sb_buffer); 4844 if (!buffer_uptodate(journal->j_sb_buffer)) { 4845 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4846 goto out_journal; 4847 } 4848 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4849 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4850 "user (unsupported) - %d", 4851 be32_to_cpu(journal->j_superblock->s_nr_users)); 4852 goto out_journal; 4853 } 4854 EXT4_SB(sb)->journal_bdev = bdev; 4855 ext4_init_journal_params(sb, journal); 4856 return journal; 4857 4858 out_journal: 4859 jbd2_journal_destroy(journal); 4860 out_bdev: 4861 ext4_blkdev_put(bdev); 4862 return NULL; 4863 } 4864 4865 static int ext4_load_journal(struct super_block *sb, 4866 struct ext4_super_block *es, 4867 unsigned long journal_devnum) 4868 { 4869 journal_t *journal; 4870 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4871 dev_t journal_dev; 4872 int err = 0; 4873 int really_read_only; 4874 4875 BUG_ON(!ext4_has_feature_journal(sb)); 4876 4877 if (journal_devnum && 4878 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4879 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4880 "numbers have changed"); 4881 journal_dev = new_decode_dev(journal_devnum); 4882 } else 4883 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4884 4885 really_read_only = bdev_read_only(sb->s_bdev); 4886 4887 /* 4888 * Are we loading a blank journal or performing recovery after a 4889 * crash? For recovery, we need to check in advance whether we 4890 * can get read-write access to the device. 4891 */ 4892 if (ext4_has_feature_journal_needs_recovery(sb)) { 4893 if (sb_rdonly(sb)) { 4894 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4895 "required on readonly filesystem"); 4896 if (really_read_only) { 4897 ext4_msg(sb, KERN_ERR, "write access " 4898 "unavailable, cannot proceed " 4899 "(try mounting with noload)"); 4900 return -EROFS; 4901 } 4902 ext4_msg(sb, KERN_INFO, "write access will " 4903 "be enabled during recovery"); 4904 } 4905 } 4906 4907 if (journal_inum && journal_dev) { 4908 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4909 "and inode journals!"); 4910 return -EINVAL; 4911 } 4912 4913 if (journal_inum) { 4914 if (!(journal = ext4_get_journal(sb, journal_inum))) 4915 return -EINVAL; 4916 } else { 4917 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4918 return -EINVAL; 4919 } 4920 4921 if (!(journal->j_flags & JBD2_BARRIER)) 4922 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4923 4924 if (!ext4_has_feature_journal_needs_recovery(sb)) 4925 err = jbd2_journal_wipe(journal, !really_read_only); 4926 if (!err) { 4927 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4928 if (save) 4929 memcpy(save, ((char *) es) + 4930 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4931 err = jbd2_journal_load(journal); 4932 if (save) 4933 memcpy(((char *) es) + EXT4_S_ERR_START, 4934 save, EXT4_S_ERR_LEN); 4935 kfree(save); 4936 } 4937 4938 if (err) { 4939 ext4_msg(sb, KERN_ERR, "error loading journal"); 4940 jbd2_journal_destroy(journal); 4941 return err; 4942 } 4943 4944 EXT4_SB(sb)->s_journal = journal; 4945 ext4_clear_journal_err(sb, es); 4946 4947 if (!really_read_only && journal_devnum && 4948 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4949 es->s_journal_dev = cpu_to_le32(journal_devnum); 4950 4951 /* Make sure we flush the recovery flag to disk. */ 4952 ext4_commit_super(sb, 1); 4953 } 4954 4955 return 0; 4956 } 4957 4958 static int ext4_commit_super(struct super_block *sb, int sync) 4959 { 4960 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4961 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4962 int error = 0; 4963 4964 if (!sbh || block_device_ejected(sb)) 4965 return error; 4966 4967 /* 4968 * The superblock bh should be mapped, but it might not be if the 4969 * device was hot-removed. Not much we can do but fail the I/O. 4970 */ 4971 if (!buffer_mapped(sbh)) 4972 return error; 4973 4974 /* 4975 * If the file system is mounted read-only, don't update the 4976 * superblock write time. This avoids updating the superblock 4977 * write time when we are mounting the root file system 4978 * read/only but we need to replay the journal; at that point, 4979 * for people who are east of GMT and who make their clock 4980 * tick in localtime for Windows bug-for-bug compatibility, 4981 * the clock is set in the future, and this will cause e2fsck 4982 * to complain and force a full file system check. 4983 */ 4984 if (!(sb->s_flags & SB_RDONLY)) 4985 ext4_update_tstamp(es, s_wtime); 4986 if (sb->s_bdev->bd_part) 4987 es->s_kbytes_written = 4988 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4989 ((part_stat_read(sb->s_bdev->bd_part, 4990 sectors[STAT_WRITE]) - 4991 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4992 else 4993 es->s_kbytes_written = 4994 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4995 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 4996 ext4_free_blocks_count_set(es, 4997 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4998 &EXT4_SB(sb)->s_freeclusters_counter))); 4999 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 5000 es->s_free_inodes_count = 5001 cpu_to_le32(percpu_counter_sum_positive( 5002 &EXT4_SB(sb)->s_freeinodes_counter)); 5003 BUFFER_TRACE(sbh, "marking dirty"); 5004 ext4_superblock_csum_set(sb); 5005 if (sync) 5006 lock_buffer(sbh); 5007 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 5008 /* 5009 * Oh, dear. A previous attempt to write the 5010 * superblock failed. This could happen because the 5011 * USB device was yanked out. Or it could happen to 5012 * be a transient write error and maybe the block will 5013 * be remapped. Nothing we can do but to retry the 5014 * write and hope for the best. 5015 */ 5016 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5017 "superblock detected"); 5018 clear_buffer_write_io_error(sbh); 5019 set_buffer_uptodate(sbh); 5020 } 5021 mark_buffer_dirty(sbh); 5022 if (sync) { 5023 unlock_buffer(sbh); 5024 error = __sync_dirty_buffer(sbh, 5025 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5026 if (buffer_write_io_error(sbh)) { 5027 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5028 "superblock"); 5029 clear_buffer_write_io_error(sbh); 5030 set_buffer_uptodate(sbh); 5031 } 5032 } 5033 return error; 5034 } 5035 5036 /* 5037 * Have we just finished recovery? If so, and if we are mounting (or 5038 * remounting) the filesystem readonly, then we will end up with a 5039 * consistent fs on disk. Record that fact. 5040 */ 5041 static void ext4_mark_recovery_complete(struct super_block *sb, 5042 struct ext4_super_block *es) 5043 { 5044 journal_t *journal = EXT4_SB(sb)->s_journal; 5045 5046 if (!ext4_has_feature_journal(sb)) { 5047 BUG_ON(journal != NULL); 5048 return; 5049 } 5050 jbd2_journal_lock_updates(journal); 5051 if (jbd2_journal_flush(journal) < 0) 5052 goto out; 5053 5054 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5055 ext4_clear_feature_journal_needs_recovery(sb); 5056 ext4_commit_super(sb, 1); 5057 } 5058 5059 out: 5060 jbd2_journal_unlock_updates(journal); 5061 } 5062 5063 /* 5064 * If we are mounting (or read-write remounting) a filesystem whose journal 5065 * has recorded an error from a previous lifetime, move that error to the 5066 * main filesystem now. 5067 */ 5068 static void ext4_clear_journal_err(struct super_block *sb, 5069 struct ext4_super_block *es) 5070 { 5071 journal_t *journal; 5072 int j_errno; 5073 const char *errstr; 5074 5075 BUG_ON(!ext4_has_feature_journal(sb)); 5076 5077 journal = EXT4_SB(sb)->s_journal; 5078 5079 /* 5080 * Now check for any error status which may have been recorded in the 5081 * journal by a prior ext4_error() or ext4_abort() 5082 */ 5083 5084 j_errno = jbd2_journal_errno(journal); 5085 if (j_errno) { 5086 char nbuf[16]; 5087 5088 errstr = ext4_decode_error(sb, j_errno, nbuf); 5089 ext4_warning(sb, "Filesystem error recorded " 5090 "from previous mount: %s", errstr); 5091 ext4_warning(sb, "Marking fs in need of filesystem check."); 5092 5093 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5094 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5095 ext4_commit_super(sb, 1); 5096 5097 jbd2_journal_clear_err(journal); 5098 jbd2_journal_update_sb_errno(journal); 5099 } 5100 } 5101 5102 /* 5103 * Force the running and committing transactions to commit, 5104 * and wait on the commit. 5105 */ 5106 int ext4_force_commit(struct super_block *sb) 5107 { 5108 journal_t *journal; 5109 5110 if (sb_rdonly(sb)) 5111 return 0; 5112 5113 journal = EXT4_SB(sb)->s_journal; 5114 return ext4_journal_force_commit(journal); 5115 } 5116 5117 static int ext4_sync_fs(struct super_block *sb, int wait) 5118 { 5119 int ret = 0; 5120 tid_t target; 5121 bool needs_barrier = false; 5122 struct ext4_sb_info *sbi = EXT4_SB(sb); 5123 5124 if (unlikely(ext4_forced_shutdown(sbi))) 5125 return 0; 5126 5127 trace_ext4_sync_fs(sb, wait); 5128 flush_workqueue(sbi->rsv_conversion_wq); 5129 /* 5130 * Writeback quota in non-journalled quota case - journalled quota has 5131 * no dirty dquots 5132 */ 5133 dquot_writeback_dquots(sb, -1); 5134 /* 5135 * Data writeback is possible w/o journal transaction, so barrier must 5136 * being sent at the end of the function. But we can skip it if 5137 * transaction_commit will do it for us. 5138 */ 5139 if (sbi->s_journal) { 5140 target = jbd2_get_latest_transaction(sbi->s_journal); 5141 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 5142 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 5143 needs_barrier = true; 5144 5145 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5146 if (wait) 5147 ret = jbd2_log_wait_commit(sbi->s_journal, 5148 target); 5149 } 5150 } else if (wait && test_opt(sb, BARRIER)) 5151 needs_barrier = true; 5152 if (needs_barrier) { 5153 int err; 5154 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 5155 if (!ret) 5156 ret = err; 5157 } 5158 5159 return ret; 5160 } 5161 5162 /* 5163 * LVM calls this function before a (read-only) snapshot is created. This 5164 * gives us a chance to flush the journal completely and mark the fs clean. 5165 * 5166 * Note that only this function cannot bring a filesystem to be in a clean 5167 * state independently. It relies on upper layer to stop all data & metadata 5168 * modifications. 5169 */ 5170 static int ext4_freeze(struct super_block *sb) 5171 { 5172 int error = 0; 5173 journal_t *journal; 5174 5175 if (sb_rdonly(sb)) 5176 return 0; 5177 5178 journal = EXT4_SB(sb)->s_journal; 5179 5180 if (journal) { 5181 /* Now we set up the journal barrier. */ 5182 jbd2_journal_lock_updates(journal); 5183 5184 /* 5185 * Don't clear the needs_recovery flag if we failed to 5186 * flush the journal. 5187 */ 5188 error = jbd2_journal_flush(journal); 5189 if (error < 0) 5190 goto out; 5191 5192 /* Journal blocked and flushed, clear needs_recovery flag. */ 5193 ext4_clear_feature_journal_needs_recovery(sb); 5194 } 5195 5196 error = ext4_commit_super(sb, 1); 5197 out: 5198 if (journal) 5199 /* we rely on upper layer to stop further updates */ 5200 jbd2_journal_unlock_updates(journal); 5201 return error; 5202 } 5203 5204 /* 5205 * Called by LVM after the snapshot is done. We need to reset the RECOVER 5206 * flag here, even though the filesystem is not technically dirty yet. 5207 */ 5208 static int ext4_unfreeze(struct super_block *sb) 5209 { 5210 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 5211 return 0; 5212 5213 if (EXT4_SB(sb)->s_journal) { 5214 /* Reset the needs_recovery flag before the fs is unlocked. */ 5215 ext4_set_feature_journal_needs_recovery(sb); 5216 } 5217 5218 ext4_commit_super(sb, 1); 5219 return 0; 5220 } 5221 5222 /* 5223 * Structure to save mount options for ext4_remount's benefit 5224 */ 5225 struct ext4_mount_options { 5226 unsigned long s_mount_opt; 5227 unsigned long s_mount_opt2; 5228 kuid_t s_resuid; 5229 kgid_t s_resgid; 5230 unsigned long s_commit_interval; 5231 u32 s_min_batch_time, s_max_batch_time; 5232 #ifdef CONFIG_QUOTA 5233 int s_jquota_fmt; 5234 char *s_qf_names[EXT4_MAXQUOTAS]; 5235 #endif 5236 }; 5237 5238 static int ext4_remount(struct super_block *sb, int *flags, char *data) 5239 { 5240 struct ext4_super_block *es; 5241 struct ext4_sb_info *sbi = EXT4_SB(sb); 5242 unsigned long old_sb_flags; 5243 struct ext4_mount_options old_opts; 5244 int enable_quota = 0; 5245 ext4_group_t g; 5246 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5247 int err = 0; 5248 #ifdef CONFIG_QUOTA 5249 int i, j; 5250 char *to_free[EXT4_MAXQUOTAS]; 5251 #endif 5252 char *orig_data = kstrdup(data, GFP_KERNEL); 5253 5254 if (data && !orig_data) 5255 return -ENOMEM; 5256 5257 /* Store the original options */ 5258 old_sb_flags = sb->s_flags; 5259 old_opts.s_mount_opt = sbi->s_mount_opt; 5260 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5261 old_opts.s_resuid = sbi->s_resuid; 5262 old_opts.s_resgid = sbi->s_resgid; 5263 old_opts.s_commit_interval = sbi->s_commit_interval; 5264 old_opts.s_min_batch_time = sbi->s_min_batch_time; 5265 old_opts.s_max_batch_time = sbi->s_max_batch_time; 5266 #ifdef CONFIG_QUOTA 5267 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5268 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5269 if (sbi->s_qf_names[i]) { 5270 char *qf_name = get_qf_name(sb, sbi, i); 5271 5272 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 5273 if (!old_opts.s_qf_names[i]) { 5274 for (j = 0; j < i; j++) 5275 kfree(old_opts.s_qf_names[j]); 5276 kfree(orig_data); 5277 return -ENOMEM; 5278 } 5279 } else 5280 old_opts.s_qf_names[i] = NULL; 5281 #endif 5282 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5283 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5284 5285 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5286 err = -EINVAL; 5287 goto restore_opts; 5288 } 5289 5290 ext4_clamp_want_extra_isize(sb); 5291 5292 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 5293 test_opt(sb, JOURNAL_CHECKSUM)) { 5294 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 5295 "during remount not supported; ignoring"); 5296 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5297 } 5298 5299 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5300 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5301 ext4_msg(sb, KERN_ERR, "can't mount with " 5302 "both data=journal and delalloc"); 5303 err = -EINVAL; 5304 goto restore_opts; 5305 } 5306 if (test_opt(sb, DIOREAD_NOLOCK)) { 5307 ext4_msg(sb, KERN_ERR, "can't mount with " 5308 "both data=journal and dioread_nolock"); 5309 err = -EINVAL; 5310 goto restore_opts; 5311 } 5312 if (test_opt(sb, DAX)) { 5313 ext4_msg(sb, KERN_ERR, "can't mount with " 5314 "both data=journal and dax"); 5315 err = -EINVAL; 5316 goto restore_opts; 5317 } 5318 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5319 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5320 ext4_msg(sb, KERN_ERR, "can't mount with " 5321 "journal_async_commit in data=ordered mode"); 5322 err = -EINVAL; 5323 goto restore_opts; 5324 } 5325 } 5326 5327 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5328 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5329 err = -EINVAL; 5330 goto restore_opts; 5331 } 5332 5333 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 5334 ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 5335 "dax flag with busy inodes while remounting"); 5336 sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 5337 } 5338 5339 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 5340 ext4_abort(sb, "Abort forced by user"); 5341 5342 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5343 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5344 5345 es = sbi->s_es; 5346 5347 if (sbi->s_journal) { 5348 ext4_init_journal_params(sb, sbi->s_journal); 5349 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5350 } 5351 5352 if (*flags & SB_LAZYTIME) 5353 sb->s_flags |= SB_LAZYTIME; 5354 5355 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 5356 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5357 err = -EROFS; 5358 goto restore_opts; 5359 } 5360 5361 if (*flags & SB_RDONLY) { 5362 err = sync_filesystem(sb); 5363 if (err < 0) 5364 goto restore_opts; 5365 err = dquot_suspend(sb, -1); 5366 if (err < 0) 5367 goto restore_opts; 5368 5369 /* 5370 * First of all, the unconditional stuff we have to do 5371 * to disable replay of the journal when we next remount 5372 */ 5373 sb->s_flags |= SB_RDONLY; 5374 5375 /* 5376 * OK, test if we are remounting a valid rw partition 5377 * readonly, and if so set the rdonly flag and then 5378 * mark the partition as valid again. 5379 */ 5380 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5381 (sbi->s_mount_state & EXT4_VALID_FS)) 5382 es->s_state = cpu_to_le16(sbi->s_mount_state); 5383 5384 if (sbi->s_journal) 5385 ext4_mark_recovery_complete(sb, es); 5386 if (sbi->s_mmp_tsk) 5387 kthread_stop(sbi->s_mmp_tsk); 5388 } else { 5389 /* Make sure we can mount this feature set readwrite */ 5390 if (ext4_has_feature_readonly(sb) || 5391 !ext4_feature_set_ok(sb, 0)) { 5392 err = -EROFS; 5393 goto restore_opts; 5394 } 5395 /* 5396 * Make sure the group descriptor checksums 5397 * are sane. If they aren't, refuse to remount r/w. 5398 */ 5399 for (g = 0; g < sbi->s_groups_count; g++) { 5400 struct ext4_group_desc *gdp = 5401 ext4_get_group_desc(sb, g, NULL); 5402 5403 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5404 ext4_msg(sb, KERN_ERR, 5405 "ext4_remount: Checksum for group %u failed (%u!=%u)", 5406 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 5407 le16_to_cpu(gdp->bg_checksum)); 5408 err = -EFSBADCRC; 5409 goto restore_opts; 5410 } 5411 } 5412 5413 /* 5414 * If we have an unprocessed orphan list hanging 5415 * around from a previously readonly bdev mount, 5416 * require a full umount/remount for now. 5417 */ 5418 if (es->s_last_orphan) { 5419 ext4_msg(sb, KERN_WARNING, "Couldn't " 5420 "remount RDWR because of unprocessed " 5421 "orphan inode list. Please " 5422 "umount/remount instead"); 5423 err = -EINVAL; 5424 goto restore_opts; 5425 } 5426 5427 /* 5428 * Mounting a RDONLY partition read-write, so reread 5429 * and store the current valid flag. (It may have 5430 * been changed by e2fsck since we originally mounted 5431 * the partition.) 5432 */ 5433 if (sbi->s_journal) 5434 ext4_clear_journal_err(sb, es); 5435 sbi->s_mount_state = le16_to_cpu(es->s_state); 5436 5437 err = ext4_setup_super(sb, es, 0); 5438 if (err) 5439 goto restore_opts; 5440 5441 sb->s_flags &= ~SB_RDONLY; 5442 if (ext4_has_feature_mmp(sb)) 5443 if (ext4_multi_mount_protect(sb, 5444 le64_to_cpu(es->s_mmp_block))) { 5445 err = -EROFS; 5446 goto restore_opts; 5447 } 5448 enable_quota = 1; 5449 } 5450 } 5451 5452 /* 5453 * Reinitialize lazy itable initialization thread based on 5454 * current settings 5455 */ 5456 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5457 ext4_unregister_li_request(sb); 5458 else { 5459 ext4_group_t first_not_zeroed; 5460 first_not_zeroed = ext4_has_uninit_itable(sb); 5461 ext4_register_li_request(sb, first_not_zeroed); 5462 } 5463 5464 ext4_setup_system_zone(sb); 5465 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5466 err = ext4_commit_super(sb, 1); 5467 if (err) 5468 goto restore_opts; 5469 } 5470 5471 #ifdef CONFIG_QUOTA 5472 /* Release old quota file names */ 5473 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5474 kfree(old_opts.s_qf_names[i]); 5475 if (enable_quota) { 5476 if (sb_any_quota_suspended(sb)) 5477 dquot_resume(sb, -1); 5478 else if (ext4_has_feature_quota(sb)) { 5479 err = ext4_enable_quotas(sb); 5480 if (err) 5481 goto restore_opts; 5482 } 5483 } 5484 #endif 5485 5486 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 5487 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5488 kfree(orig_data); 5489 return 0; 5490 5491 restore_opts: 5492 sb->s_flags = old_sb_flags; 5493 sbi->s_mount_opt = old_opts.s_mount_opt; 5494 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5495 sbi->s_resuid = old_opts.s_resuid; 5496 sbi->s_resgid = old_opts.s_resgid; 5497 sbi->s_commit_interval = old_opts.s_commit_interval; 5498 sbi->s_min_batch_time = old_opts.s_min_batch_time; 5499 sbi->s_max_batch_time = old_opts.s_max_batch_time; 5500 #ifdef CONFIG_QUOTA 5501 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5502 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 5503 to_free[i] = get_qf_name(sb, sbi, i); 5504 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5505 } 5506 synchronize_rcu(); 5507 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5508 kfree(to_free[i]); 5509 #endif 5510 kfree(orig_data); 5511 return err; 5512 } 5513 5514 #ifdef CONFIG_QUOTA 5515 static int ext4_statfs_project(struct super_block *sb, 5516 kprojid_t projid, struct kstatfs *buf) 5517 { 5518 struct kqid qid; 5519 struct dquot *dquot; 5520 u64 limit; 5521 u64 curblock; 5522 5523 qid = make_kqid_projid(projid); 5524 dquot = dqget(sb, qid); 5525 if (IS_ERR(dquot)) 5526 return PTR_ERR(dquot); 5527 spin_lock(&dquot->dq_dqb_lock); 5528 5529 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 5530 dquot->dq_dqb.dqb_bsoftlimit : 5531 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 5532 if (limit && buf->f_blocks > limit) { 5533 curblock = (dquot->dq_dqb.dqb_curspace + 5534 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5535 buf->f_blocks = limit; 5536 buf->f_bfree = buf->f_bavail = 5537 (buf->f_blocks > curblock) ? 5538 (buf->f_blocks - curblock) : 0; 5539 } 5540 5541 limit = dquot->dq_dqb.dqb_isoftlimit ? 5542 dquot->dq_dqb.dqb_isoftlimit : 5543 dquot->dq_dqb.dqb_ihardlimit; 5544 if (limit && buf->f_files > limit) { 5545 buf->f_files = limit; 5546 buf->f_ffree = 5547 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5548 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5549 } 5550 5551 spin_unlock(&dquot->dq_dqb_lock); 5552 dqput(dquot); 5553 return 0; 5554 } 5555 #endif 5556 5557 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5558 { 5559 struct super_block *sb = dentry->d_sb; 5560 struct ext4_sb_info *sbi = EXT4_SB(sb); 5561 struct ext4_super_block *es = sbi->s_es; 5562 ext4_fsblk_t overhead = 0, resv_blocks; 5563 u64 fsid; 5564 s64 bfree; 5565 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5566 5567 if (!test_opt(sb, MINIX_DF)) 5568 overhead = sbi->s_overhead; 5569 5570 buf->f_type = EXT4_SUPER_MAGIC; 5571 buf->f_bsize = sb->s_blocksize; 5572 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5573 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5574 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5575 /* prevent underflow in case that few free space is available */ 5576 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5577 buf->f_bavail = buf->f_bfree - 5578 (ext4_r_blocks_count(es) + resv_blocks); 5579 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5580 buf->f_bavail = 0; 5581 buf->f_files = le32_to_cpu(es->s_inodes_count); 5582 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5583 buf->f_namelen = EXT4_NAME_LEN; 5584 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5585 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5586 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5587 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5588 5589 #ifdef CONFIG_QUOTA 5590 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5591 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5592 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5593 #endif 5594 return 0; 5595 } 5596 5597 5598 #ifdef CONFIG_QUOTA 5599 5600 /* 5601 * Helper functions so that transaction is started before we acquire dqio_sem 5602 * to keep correct lock ordering of transaction > dqio_sem 5603 */ 5604 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5605 { 5606 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5607 } 5608 5609 static int ext4_write_dquot(struct dquot *dquot) 5610 { 5611 int ret, err; 5612 handle_t *handle; 5613 struct inode *inode; 5614 5615 inode = dquot_to_inode(dquot); 5616 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5617 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5618 if (IS_ERR(handle)) 5619 return PTR_ERR(handle); 5620 ret = dquot_commit(dquot); 5621 err = ext4_journal_stop(handle); 5622 if (!ret) 5623 ret = err; 5624 return ret; 5625 } 5626 5627 static int ext4_acquire_dquot(struct dquot *dquot) 5628 { 5629 int ret, err; 5630 handle_t *handle; 5631 5632 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5633 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5634 if (IS_ERR(handle)) 5635 return PTR_ERR(handle); 5636 ret = dquot_acquire(dquot); 5637 err = ext4_journal_stop(handle); 5638 if (!ret) 5639 ret = err; 5640 return ret; 5641 } 5642 5643 static int ext4_release_dquot(struct dquot *dquot) 5644 { 5645 int ret, err; 5646 handle_t *handle; 5647 5648 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5649 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5650 if (IS_ERR(handle)) { 5651 /* Release dquot anyway to avoid endless cycle in dqput() */ 5652 dquot_release(dquot); 5653 return PTR_ERR(handle); 5654 } 5655 ret = dquot_release(dquot); 5656 err = ext4_journal_stop(handle); 5657 if (!ret) 5658 ret = err; 5659 return ret; 5660 } 5661 5662 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5663 { 5664 struct super_block *sb = dquot->dq_sb; 5665 struct ext4_sb_info *sbi = EXT4_SB(sb); 5666 5667 /* Are we journaling quotas? */ 5668 if (ext4_has_feature_quota(sb) || 5669 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5670 dquot_mark_dquot_dirty(dquot); 5671 return ext4_write_dquot(dquot); 5672 } else { 5673 return dquot_mark_dquot_dirty(dquot); 5674 } 5675 } 5676 5677 static int ext4_write_info(struct super_block *sb, int type) 5678 { 5679 int ret, err; 5680 handle_t *handle; 5681 5682 /* Data block + inode block */ 5683 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5684 if (IS_ERR(handle)) 5685 return PTR_ERR(handle); 5686 ret = dquot_commit_info(sb, type); 5687 err = ext4_journal_stop(handle); 5688 if (!ret) 5689 ret = err; 5690 return ret; 5691 } 5692 5693 /* 5694 * Turn on quotas during mount time - we need to find 5695 * the quota file and such... 5696 */ 5697 static int ext4_quota_on_mount(struct super_block *sb, int type) 5698 { 5699 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 5700 EXT4_SB(sb)->s_jquota_fmt, type); 5701 } 5702 5703 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5704 { 5705 struct ext4_inode_info *ei = EXT4_I(inode); 5706 5707 /* The first argument of lockdep_set_subclass has to be 5708 * *exactly* the same as the argument to init_rwsem() --- in 5709 * this case, in init_once() --- or lockdep gets unhappy 5710 * because the name of the lock is set using the 5711 * stringification of the argument to init_rwsem(). 5712 */ 5713 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5714 lockdep_set_subclass(&ei->i_data_sem, subclass); 5715 } 5716 5717 /* 5718 * Standard function to be called on quota_on 5719 */ 5720 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5721 const struct path *path) 5722 { 5723 int err; 5724 5725 if (!test_opt(sb, QUOTA)) 5726 return -EINVAL; 5727 5728 /* Quotafile not on the same filesystem? */ 5729 if (path->dentry->d_sb != sb) 5730 return -EXDEV; 5731 /* Journaling quota? */ 5732 if (EXT4_SB(sb)->s_qf_names[type]) { 5733 /* Quotafile not in fs root? */ 5734 if (path->dentry->d_parent != sb->s_root) 5735 ext4_msg(sb, KERN_WARNING, 5736 "Quota file not on filesystem root. " 5737 "Journaled quota will not work"); 5738 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 5739 } else { 5740 /* 5741 * Clear the flag just in case mount options changed since 5742 * last time. 5743 */ 5744 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 5745 } 5746 5747 /* 5748 * When we journal data on quota file, we have to flush journal to see 5749 * all updates to the file when we bypass pagecache... 5750 */ 5751 if (EXT4_SB(sb)->s_journal && 5752 ext4_should_journal_data(d_inode(path->dentry))) { 5753 /* 5754 * We don't need to lock updates but journal_flush() could 5755 * otherwise be livelocked... 5756 */ 5757 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5758 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5759 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5760 if (err) 5761 return err; 5762 } 5763 5764 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5765 err = dquot_quota_on(sb, type, format_id, path); 5766 if (err) { 5767 lockdep_set_quota_inode(path->dentry->d_inode, 5768 I_DATA_SEM_NORMAL); 5769 } else { 5770 struct inode *inode = d_inode(path->dentry); 5771 handle_t *handle; 5772 5773 /* 5774 * Set inode flags to prevent userspace from messing with quota 5775 * files. If this fails, we return success anyway since quotas 5776 * are already enabled and this is not a hard failure. 5777 */ 5778 inode_lock(inode); 5779 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5780 if (IS_ERR(handle)) 5781 goto unlock_inode; 5782 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 5783 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 5784 S_NOATIME | S_IMMUTABLE); 5785 ext4_mark_inode_dirty(handle, inode); 5786 ext4_journal_stop(handle); 5787 unlock_inode: 5788 inode_unlock(inode); 5789 } 5790 return err; 5791 } 5792 5793 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5794 unsigned int flags) 5795 { 5796 int err; 5797 struct inode *qf_inode; 5798 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5799 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5800 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5801 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5802 }; 5803 5804 BUG_ON(!ext4_has_feature_quota(sb)); 5805 5806 if (!qf_inums[type]) 5807 return -EPERM; 5808 5809 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 5810 if (IS_ERR(qf_inode)) { 5811 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5812 return PTR_ERR(qf_inode); 5813 } 5814 5815 /* Don't account quota for quota files to avoid recursion */ 5816 qf_inode->i_flags |= S_NOQUOTA; 5817 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 5818 err = dquot_enable(qf_inode, type, format_id, flags); 5819 if (err) 5820 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 5821 iput(qf_inode); 5822 5823 return err; 5824 } 5825 5826 /* Enable usage tracking for all quota types. */ 5827 static int ext4_enable_quotas(struct super_block *sb) 5828 { 5829 int type, err = 0; 5830 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5831 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5832 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5833 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5834 }; 5835 bool quota_mopt[EXT4_MAXQUOTAS] = { 5836 test_opt(sb, USRQUOTA), 5837 test_opt(sb, GRPQUOTA), 5838 test_opt(sb, PRJQUOTA), 5839 }; 5840 5841 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 5842 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 5843 if (qf_inums[type]) { 5844 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5845 DQUOT_USAGE_ENABLED | 5846 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 5847 if (err) { 5848 ext4_warning(sb, 5849 "Failed to enable quota tracking " 5850 "(type=%d, err=%d). Please run " 5851 "e2fsck to fix.", type, err); 5852 for (type--; type >= 0; type--) 5853 dquot_quota_off(sb, type); 5854 5855 return err; 5856 } 5857 } 5858 } 5859 return 0; 5860 } 5861 5862 static int ext4_quota_off(struct super_block *sb, int type) 5863 { 5864 struct inode *inode = sb_dqopt(sb)->files[type]; 5865 handle_t *handle; 5866 int err; 5867 5868 /* Force all delayed allocation blocks to be allocated. 5869 * Caller already holds s_umount sem */ 5870 if (test_opt(sb, DELALLOC)) 5871 sync_filesystem(sb); 5872 5873 if (!inode || !igrab(inode)) 5874 goto out; 5875 5876 err = dquot_quota_off(sb, type); 5877 if (err || ext4_has_feature_quota(sb)) 5878 goto out_put; 5879 5880 inode_lock(inode); 5881 /* 5882 * Update modification times of quota files when userspace can 5883 * start looking at them. If we fail, we return success anyway since 5884 * this is not a hard failure and quotas are already disabled. 5885 */ 5886 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5887 if (IS_ERR(handle)) 5888 goto out_unlock; 5889 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 5890 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 5891 inode->i_mtime = inode->i_ctime = current_time(inode); 5892 ext4_mark_inode_dirty(handle, inode); 5893 ext4_journal_stop(handle); 5894 out_unlock: 5895 inode_unlock(inode); 5896 out_put: 5897 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 5898 iput(inode); 5899 return err; 5900 out: 5901 return dquot_quota_off(sb, type); 5902 } 5903 5904 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5905 * acquiring the locks... As quota files are never truncated and quota code 5906 * itself serializes the operations (and no one else should touch the files) 5907 * we don't have to be afraid of races */ 5908 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5909 size_t len, loff_t off) 5910 { 5911 struct inode *inode = sb_dqopt(sb)->files[type]; 5912 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5913 int offset = off & (sb->s_blocksize - 1); 5914 int tocopy; 5915 size_t toread; 5916 struct buffer_head *bh; 5917 loff_t i_size = i_size_read(inode); 5918 5919 if (off > i_size) 5920 return 0; 5921 if (off+len > i_size) 5922 len = i_size-off; 5923 toread = len; 5924 while (toread > 0) { 5925 tocopy = sb->s_blocksize - offset < toread ? 5926 sb->s_blocksize - offset : toread; 5927 bh = ext4_bread(NULL, inode, blk, 0); 5928 if (IS_ERR(bh)) 5929 return PTR_ERR(bh); 5930 if (!bh) /* A hole? */ 5931 memset(data, 0, tocopy); 5932 else 5933 memcpy(data, bh->b_data+offset, tocopy); 5934 brelse(bh); 5935 offset = 0; 5936 toread -= tocopy; 5937 data += tocopy; 5938 blk++; 5939 } 5940 return len; 5941 } 5942 5943 /* Write to quotafile (we know the transaction is already started and has 5944 * enough credits) */ 5945 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5946 const char *data, size_t len, loff_t off) 5947 { 5948 struct inode *inode = sb_dqopt(sb)->files[type]; 5949 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5950 int err, offset = off & (sb->s_blocksize - 1); 5951 int retries = 0; 5952 struct buffer_head *bh; 5953 handle_t *handle = journal_current_handle(); 5954 5955 if (EXT4_SB(sb)->s_journal && !handle) { 5956 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5957 " cancelled because transaction is not started", 5958 (unsigned long long)off, (unsigned long long)len); 5959 return -EIO; 5960 } 5961 /* 5962 * Since we account only one data block in transaction credits, 5963 * then it is impossible to cross a block boundary. 5964 */ 5965 if (sb->s_blocksize - offset < len) { 5966 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5967 " cancelled because not block aligned", 5968 (unsigned long long)off, (unsigned long long)len); 5969 return -EIO; 5970 } 5971 5972 do { 5973 bh = ext4_bread(handle, inode, blk, 5974 EXT4_GET_BLOCKS_CREATE | 5975 EXT4_GET_BLOCKS_METADATA_NOFAIL); 5976 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 5977 ext4_should_retry_alloc(inode->i_sb, &retries)); 5978 if (IS_ERR(bh)) 5979 return PTR_ERR(bh); 5980 if (!bh) 5981 goto out; 5982 BUFFER_TRACE(bh, "get write access"); 5983 err = ext4_journal_get_write_access(handle, bh); 5984 if (err) { 5985 brelse(bh); 5986 return err; 5987 } 5988 lock_buffer(bh); 5989 memcpy(bh->b_data+offset, data, len); 5990 flush_dcache_page(bh->b_page); 5991 unlock_buffer(bh); 5992 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5993 brelse(bh); 5994 out: 5995 if (inode->i_size < off + len) { 5996 i_size_write(inode, off + len); 5997 EXT4_I(inode)->i_disksize = inode->i_size; 5998 ext4_mark_inode_dirty(handle, inode); 5999 } 6000 return len; 6001 } 6002 6003 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) 6004 { 6005 const struct quota_format_ops *ops; 6006 6007 if (!sb_has_quota_loaded(sb, qid->type)) 6008 return -ESRCH; 6009 ops = sb_dqopt(sb)->ops[qid->type]; 6010 if (!ops || !ops->get_next_id) 6011 return -ENOSYS; 6012 return dquot_get_next_id(sb, qid); 6013 } 6014 #endif 6015 6016 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6017 const char *dev_name, void *data) 6018 { 6019 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6020 } 6021 6022 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 6023 static inline void register_as_ext2(void) 6024 { 6025 int err = register_filesystem(&ext2_fs_type); 6026 if (err) 6027 printk(KERN_WARNING 6028 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 6029 } 6030 6031 static inline void unregister_as_ext2(void) 6032 { 6033 unregister_filesystem(&ext2_fs_type); 6034 } 6035 6036 static inline int ext2_feature_set_ok(struct super_block *sb) 6037 { 6038 if (ext4_has_unknown_ext2_incompat_features(sb)) 6039 return 0; 6040 if (sb_rdonly(sb)) 6041 return 1; 6042 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 6043 return 0; 6044 return 1; 6045 } 6046 #else 6047 static inline void register_as_ext2(void) { } 6048 static inline void unregister_as_ext2(void) { } 6049 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 6050 #endif 6051 6052 static inline void register_as_ext3(void) 6053 { 6054 int err = register_filesystem(&ext3_fs_type); 6055 if (err) 6056 printk(KERN_WARNING 6057 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 6058 } 6059 6060 static inline void unregister_as_ext3(void) 6061 { 6062 unregister_filesystem(&ext3_fs_type); 6063 } 6064 6065 static inline int ext3_feature_set_ok(struct super_block *sb) 6066 { 6067 if (ext4_has_unknown_ext3_incompat_features(sb)) 6068 return 0; 6069 if (!ext4_has_feature_journal(sb)) 6070 return 0; 6071 if (sb_rdonly(sb)) 6072 return 1; 6073 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 6074 return 0; 6075 return 1; 6076 } 6077 6078 static struct file_system_type ext4_fs_type = { 6079 .owner = THIS_MODULE, 6080 .name = "ext4", 6081 .mount = ext4_mount, 6082 .kill_sb = kill_block_super, 6083 .fs_flags = FS_REQUIRES_DEV, 6084 }; 6085 MODULE_ALIAS_FS("ext4"); 6086 6087 /* Shared across all ext4 file systems */ 6088 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6089 6090 static int __init ext4_init_fs(void) 6091 { 6092 int i, err; 6093 6094 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 6095 ext4_li_info = NULL; 6096 mutex_init(&ext4_li_mtx); 6097 6098 /* Build-time check for flags consistency */ 6099 ext4_check_flag_values(); 6100 6101 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6102 init_waitqueue_head(&ext4__ioend_wq[i]); 6103 6104 err = ext4_init_es(); 6105 if (err) 6106 return err; 6107 6108 err = ext4_init_pending(); 6109 if (err) 6110 goto out7; 6111 6112 err = ext4_init_post_read_processing(); 6113 if (err) 6114 goto out6; 6115 6116 err = ext4_init_pageio(); 6117 if (err) 6118 goto out5; 6119 6120 err = ext4_init_system_zone(); 6121 if (err) 6122 goto out4; 6123 6124 err = ext4_init_sysfs(); 6125 if (err) 6126 goto out3; 6127 6128 err = ext4_init_mballoc(); 6129 if (err) 6130 goto out2; 6131 err = init_inodecache(); 6132 if (err) 6133 goto out1; 6134 register_as_ext3(); 6135 register_as_ext2(); 6136 err = register_filesystem(&ext4_fs_type); 6137 if (err) 6138 goto out; 6139 6140 return 0; 6141 out: 6142 unregister_as_ext2(); 6143 unregister_as_ext3(); 6144 destroy_inodecache(); 6145 out1: 6146 ext4_exit_mballoc(); 6147 out2: 6148 ext4_exit_sysfs(); 6149 out3: 6150 ext4_exit_system_zone(); 6151 out4: 6152 ext4_exit_pageio(); 6153 out5: 6154 ext4_exit_post_read_processing(); 6155 out6: 6156 ext4_exit_pending(); 6157 out7: 6158 ext4_exit_es(); 6159 6160 return err; 6161 } 6162 6163 static void __exit ext4_exit_fs(void) 6164 { 6165 ext4_destroy_lazyinit_thread(); 6166 unregister_as_ext2(); 6167 unregister_as_ext3(); 6168 unregister_filesystem(&ext4_fs_type); 6169 destroy_inodecache(); 6170 ext4_exit_mballoc(); 6171 ext4_exit_sysfs(); 6172 ext4_exit_system_zone(); 6173 ext4_exit_pageio(); 6174 ext4_exit_post_read_processing(); 6175 ext4_exit_es(); 6176 ext4_exit_pending(); 6177 } 6178 6179 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 6180 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6181 MODULE_LICENSE("GPL"); 6182 MODULE_SOFTDEP("pre: crc32c"); 6183 module_init(ext4_init_fs) 6184 module_exit(ext4_exit_fs) 6185