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 bool ext4_has_stable_inodes(struct super_block *sb) 1349 { 1350 return ext4_has_feature_stable_inodes(sb); 1351 } 1352 1353 static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1354 int *ino_bits_ret, int *lblk_bits_ret) 1355 { 1356 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1357 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1358 } 1359 1360 static const struct fscrypt_operations ext4_cryptops = { 1361 .key_prefix = "ext4:", 1362 .get_context = ext4_get_context, 1363 .set_context = ext4_set_context, 1364 .dummy_context = ext4_dummy_context, 1365 .empty_dir = ext4_empty_dir, 1366 .max_namelen = EXT4_NAME_LEN, 1367 .has_stable_inodes = ext4_has_stable_inodes, 1368 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1369 }; 1370 #endif 1371 1372 #ifdef CONFIG_QUOTA 1373 static const char * const quotatypes[] = INITQFNAMES; 1374 #define QTYPE2NAME(t) (quotatypes[t]) 1375 1376 static int ext4_write_dquot(struct dquot *dquot); 1377 static int ext4_acquire_dquot(struct dquot *dquot); 1378 static int ext4_release_dquot(struct dquot *dquot); 1379 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1380 static int ext4_write_info(struct super_block *sb, int type); 1381 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1382 const struct path *path); 1383 static int ext4_quota_on_mount(struct super_block *sb, int type); 1384 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1385 size_t len, loff_t off); 1386 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1387 const char *data, size_t len, loff_t off); 1388 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1389 unsigned int flags); 1390 static int ext4_enable_quotas(struct super_block *sb); 1391 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); 1392 1393 static struct dquot **ext4_get_dquots(struct inode *inode) 1394 { 1395 return EXT4_I(inode)->i_dquot; 1396 } 1397 1398 static const struct dquot_operations ext4_quota_operations = { 1399 .get_reserved_space = ext4_get_reserved_space, 1400 .write_dquot = ext4_write_dquot, 1401 .acquire_dquot = ext4_acquire_dquot, 1402 .release_dquot = ext4_release_dquot, 1403 .mark_dirty = ext4_mark_dquot_dirty, 1404 .write_info = ext4_write_info, 1405 .alloc_dquot = dquot_alloc, 1406 .destroy_dquot = dquot_destroy, 1407 .get_projid = ext4_get_projid, 1408 .get_inode_usage = ext4_get_inode_usage, 1409 .get_next_id = ext4_get_next_id, 1410 }; 1411 1412 static const struct quotactl_ops ext4_qctl_operations = { 1413 .quota_on = ext4_quota_on, 1414 .quota_off = ext4_quota_off, 1415 .quota_sync = dquot_quota_sync, 1416 .get_state = dquot_get_state, 1417 .set_info = dquot_set_dqinfo, 1418 .get_dqblk = dquot_get_dqblk, 1419 .set_dqblk = dquot_set_dqblk, 1420 .get_nextdqblk = dquot_get_next_dqblk, 1421 }; 1422 #endif 1423 1424 static const struct super_operations ext4_sops = { 1425 .alloc_inode = ext4_alloc_inode, 1426 .free_inode = ext4_free_in_core_inode, 1427 .destroy_inode = ext4_destroy_inode, 1428 .write_inode = ext4_write_inode, 1429 .dirty_inode = ext4_dirty_inode, 1430 .drop_inode = ext4_drop_inode, 1431 .evict_inode = ext4_evict_inode, 1432 .put_super = ext4_put_super, 1433 .sync_fs = ext4_sync_fs, 1434 .freeze_fs = ext4_freeze, 1435 .unfreeze_fs = ext4_unfreeze, 1436 .statfs = ext4_statfs, 1437 .remount_fs = ext4_remount, 1438 .show_options = ext4_show_options, 1439 #ifdef CONFIG_QUOTA 1440 .quota_read = ext4_quota_read, 1441 .quota_write = ext4_quota_write, 1442 .get_dquots = ext4_get_dquots, 1443 #endif 1444 .bdev_try_to_free_page = bdev_try_to_free_page, 1445 }; 1446 1447 static const struct export_operations ext4_export_ops = { 1448 .fh_to_dentry = ext4_fh_to_dentry, 1449 .fh_to_parent = ext4_fh_to_parent, 1450 .get_parent = ext4_get_parent, 1451 .commit_metadata = ext4_nfs_commit_metadata, 1452 }; 1453 1454 enum { 1455 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1456 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1457 Opt_nouid32, Opt_debug, Opt_removed, 1458 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1459 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1460 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1461 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1462 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1463 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1464 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1465 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1466 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1467 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, 1468 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1469 Opt_nowarn_on_error, Opt_mblk_io_submit, 1470 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 1471 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1472 Opt_inode_readahead_blks, Opt_journal_ioprio, 1473 Opt_dioread_nolock, Opt_dioread_lock, 1474 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1475 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1476 }; 1477 1478 static const match_table_t tokens = { 1479 {Opt_bsd_df, "bsddf"}, 1480 {Opt_minix_df, "minixdf"}, 1481 {Opt_grpid, "grpid"}, 1482 {Opt_grpid, "bsdgroups"}, 1483 {Opt_nogrpid, "nogrpid"}, 1484 {Opt_nogrpid, "sysvgroups"}, 1485 {Opt_resgid, "resgid=%u"}, 1486 {Opt_resuid, "resuid=%u"}, 1487 {Opt_sb, "sb=%u"}, 1488 {Opt_err_cont, "errors=continue"}, 1489 {Opt_err_panic, "errors=panic"}, 1490 {Opt_err_ro, "errors=remount-ro"}, 1491 {Opt_nouid32, "nouid32"}, 1492 {Opt_debug, "debug"}, 1493 {Opt_removed, "oldalloc"}, 1494 {Opt_removed, "orlov"}, 1495 {Opt_user_xattr, "user_xattr"}, 1496 {Opt_nouser_xattr, "nouser_xattr"}, 1497 {Opt_acl, "acl"}, 1498 {Opt_noacl, "noacl"}, 1499 {Opt_noload, "norecovery"}, 1500 {Opt_noload, "noload"}, 1501 {Opt_removed, "nobh"}, 1502 {Opt_removed, "bh"}, 1503 {Opt_commit, "commit=%u"}, 1504 {Opt_min_batch_time, "min_batch_time=%u"}, 1505 {Opt_max_batch_time, "max_batch_time=%u"}, 1506 {Opt_journal_dev, "journal_dev=%u"}, 1507 {Opt_journal_path, "journal_path=%s"}, 1508 {Opt_journal_checksum, "journal_checksum"}, 1509 {Opt_nojournal_checksum, "nojournal_checksum"}, 1510 {Opt_journal_async_commit, "journal_async_commit"}, 1511 {Opt_abort, "abort"}, 1512 {Opt_data_journal, "data=journal"}, 1513 {Opt_data_ordered, "data=ordered"}, 1514 {Opt_data_writeback, "data=writeback"}, 1515 {Opt_data_err_abort, "data_err=abort"}, 1516 {Opt_data_err_ignore, "data_err=ignore"}, 1517 {Opt_offusrjquota, "usrjquota="}, 1518 {Opt_usrjquota, "usrjquota=%s"}, 1519 {Opt_offgrpjquota, "grpjquota="}, 1520 {Opt_grpjquota, "grpjquota=%s"}, 1521 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1522 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1523 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1524 {Opt_grpquota, "grpquota"}, 1525 {Opt_noquota, "noquota"}, 1526 {Opt_quota, "quota"}, 1527 {Opt_usrquota, "usrquota"}, 1528 {Opt_prjquota, "prjquota"}, 1529 {Opt_barrier, "barrier=%u"}, 1530 {Opt_barrier, "barrier"}, 1531 {Opt_nobarrier, "nobarrier"}, 1532 {Opt_i_version, "i_version"}, 1533 {Opt_dax, "dax"}, 1534 {Opt_stripe, "stripe=%u"}, 1535 {Opt_delalloc, "delalloc"}, 1536 {Opt_warn_on_error, "warn_on_error"}, 1537 {Opt_nowarn_on_error, "nowarn_on_error"}, 1538 {Opt_lazytime, "lazytime"}, 1539 {Opt_nolazytime, "nolazytime"}, 1540 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1541 {Opt_nodelalloc, "nodelalloc"}, 1542 {Opt_removed, "mblk_io_submit"}, 1543 {Opt_removed, "nomblk_io_submit"}, 1544 {Opt_block_validity, "block_validity"}, 1545 {Opt_noblock_validity, "noblock_validity"}, 1546 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1547 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1548 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1549 {Opt_auto_da_alloc, "auto_da_alloc"}, 1550 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1551 {Opt_dioread_nolock, "dioread_nolock"}, 1552 {Opt_dioread_lock, "dioread_lock"}, 1553 {Opt_discard, "discard"}, 1554 {Opt_nodiscard, "nodiscard"}, 1555 {Opt_init_itable, "init_itable=%u"}, 1556 {Opt_init_itable, "init_itable"}, 1557 {Opt_noinit_itable, "noinit_itable"}, 1558 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1559 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1560 {Opt_nombcache, "nombcache"}, 1561 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1562 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1563 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1564 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1565 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1566 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1567 {Opt_err, NULL}, 1568 }; 1569 1570 static ext4_fsblk_t get_sb_block(void **data) 1571 { 1572 ext4_fsblk_t sb_block; 1573 char *options = (char *) *data; 1574 1575 if (!options || strncmp(options, "sb=", 3) != 0) 1576 return 1; /* Default location */ 1577 1578 options += 3; 1579 /* TODO: use simple_strtoll with >32bit ext4 */ 1580 sb_block = simple_strtoul(options, &options, 0); 1581 if (*options && *options != ',') { 1582 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1583 (char *) *data); 1584 return 1; 1585 } 1586 if (*options == ',') 1587 options++; 1588 *data = (void *) options; 1589 1590 return sb_block; 1591 } 1592 1593 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1594 static const char deprecated_msg[] = 1595 "Mount option \"%s\" will be removed by %s\n" 1596 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1597 1598 #ifdef CONFIG_QUOTA 1599 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1600 { 1601 struct ext4_sb_info *sbi = EXT4_SB(sb); 1602 char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 1603 int ret = -1; 1604 1605 if (sb_any_quota_loaded(sb) && !old_qname) { 1606 ext4_msg(sb, KERN_ERR, 1607 "Cannot change journaled " 1608 "quota options when quota turned on"); 1609 return -1; 1610 } 1611 if (ext4_has_feature_quota(sb)) { 1612 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1613 "ignored when QUOTA feature is enabled"); 1614 return 1; 1615 } 1616 qname = match_strdup(args); 1617 if (!qname) { 1618 ext4_msg(sb, KERN_ERR, 1619 "Not enough memory for storing quotafile name"); 1620 return -1; 1621 } 1622 if (old_qname) { 1623 if (strcmp(old_qname, qname) == 0) 1624 ret = 1; 1625 else 1626 ext4_msg(sb, KERN_ERR, 1627 "%s quota file already specified", 1628 QTYPE2NAME(qtype)); 1629 goto errout; 1630 } 1631 if (strchr(qname, '/')) { 1632 ext4_msg(sb, KERN_ERR, 1633 "quotafile must be on filesystem root"); 1634 goto errout; 1635 } 1636 rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1637 set_opt(sb, QUOTA); 1638 return 1; 1639 errout: 1640 kfree(qname); 1641 return ret; 1642 } 1643 1644 static int clear_qf_name(struct super_block *sb, int qtype) 1645 { 1646 1647 struct ext4_sb_info *sbi = EXT4_SB(sb); 1648 char *old_qname = get_qf_name(sb, sbi, qtype); 1649 1650 if (sb_any_quota_loaded(sb) && old_qname) { 1651 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1652 " when quota turned on"); 1653 return -1; 1654 } 1655 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 1656 synchronize_rcu(); 1657 kfree(old_qname); 1658 return 1; 1659 } 1660 #endif 1661 1662 #define MOPT_SET 0x0001 1663 #define MOPT_CLEAR 0x0002 1664 #define MOPT_NOSUPPORT 0x0004 1665 #define MOPT_EXPLICIT 0x0008 1666 #define MOPT_CLEAR_ERR 0x0010 1667 #define MOPT_GTE0 0x0020 1668 #ifdef CONFIG_QUOTA 1669 #define MOPT_Q 0 1670 #define MOPT_QFMT 0x0040 1671 #else 1672 #define MOPT_Q MOPT_NOSUPPORT 1673 #define MOPT_QFMT MOPT_NOSUPPORT 1674 #endif 1675 #define MOPT_DATAJ 0x0080 1676 #define MOPT_NO_EXT2 0x0100 1677 #define MOPT_NO_EXT3 0x0200 1678 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1679 #define MOPT_STRING 0x0400 1680 1681 static const struct mount_opts { 1682 int token; 1683 int mount_opt; 1684 int flags; 1685 } ext4_mount_opts[] = { 1686 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1687 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1688 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1689 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1690 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1691 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1692 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1693 MOPT_EXT4_ONLY | MOPT_SET}, 1694 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1695 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1696 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1697 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1698 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1699 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1700 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1701 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1702 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1703 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1704 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1705 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1706 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1707 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1708 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1709 EXT4_MOUNT_JOURNAL_CHECKSUM), 1710 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1711 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1712 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1713 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1714 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1715 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1716 MOPT_NO_EXT2}, 1717 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1718 MOPT_NO_EXT2}, 1719 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1720 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1721 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1722 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1723 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1724 {Opt_commit, 0, MOPT_GTE0}, 1725 {Opt_max_batch_time, 0, MOPT_GTE0}, 1726 {Opt_min_batch_time, 0, MOPT_GTE0}, 1727 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1728 {Opt_init_itable, 0, MOPT_GTE0}, 1729 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 1730 {Opt_stripe, 0, MOPT_GTE0}, 1731 {Opt_resuid, 0, MOPT_GTE0}, 1732 {Opt_resgid, 0, MOPT_GTE0}, 1733 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1734 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1735 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1736 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1737 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1738 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1739 MOPT_NO_EXT2 | MOPT_DATAJ}, 1740 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1741 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1742 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1743 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1744 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1745 #else 1746 {Opt_acl, 0, MOPT_NOSUPPORT}, 1747 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1748 #endif 1749 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1750 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1751 {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 1752 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1753 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1754 MOPT_SET | MOPT_Q}, 1755 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1756 MOPT_SET | MOPT_Q}, 1757 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1758 MOPT_SET | MOPT_Q}, 1759 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1760 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1761 MOPT_CLEAR | MOPT_Q}, 1762 {Opt_usrjquota, 0, MOPT_Q}, 1763 {Opt_grpjquota, 0, MOPT_Q}, 1764 {Opt_offusrjquota, 0, MOPT_Q}, 1765 {Opt_offgrpjquota, 0, MOPT_Q}, 1766 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1767 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1768 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1769 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1770 {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1771 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1772 {Opt_err, 0, 0} 1773 }; 1774 1775 #ifdef CONFIG_UNICODE 1776 static const struct ext4_sb_encodings { 1777 __u16 magic; 1778 char *name; 1779 char *version; 1780 } ext4_sb_encoding_map[] = { 1781 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1782 }; 1783 1784 static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1785 const struct ext4_sb_encodings **encoding, 1786 __u16 *flags) 1787 { 1788 __u16 magic = le16_to_cpu(es->s_encoding); 1789 int i; 1790 1791 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1792 if (magic == ext4_sb_encoding_map[i].magic) 1793 break; 1794 1795 if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1796 return -EINVAL; 1797 1798 *encoding = &ext4_sb_encoding_map[i]; 1799 *flags = le16_to_cpu(es->s_encoding_flags); 1800 1801 return 0; 1802 } 1803 #endif 1804 1805 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1806 substring_t *args, unsigned long *journal_devnum, 1807 unsigned int *journal_ioprio, int is_remount) 1808 { 1809 struct ext4_sb_info *sbi = EXT4_SB(sb); 1810 const struct mount_opts *m; 1811 kuid_t uid; 1812 kgid_t gid; 1813 int arg = 0; 1814 1815 #ifdef CONFIG_QUOTA 1816 if (token == Opt_usrjquota) 1817 return set_qf_name(sb, USRQUOTA, &args[0]); 1818 else if (token == Opt_grpjquota) 1819 return set_qf_name(sb, GRPQUOTA, &args[0]); 1820 else if (token == Opt_offusrjquota) 1821 return clear_qf_name(sb, USRQUOTA); 1822 else if (token == Opt_offgrpjquota) 1823 return clear_qf_name(sb, GRPQUOTA); 1824 #endif 1825 switch (token) { 1826 case Opt_noacl: 1827 case Opt_nouser_xattr: 1828 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1829 break; 1830 case Opt_sb: 1831 return 1; /* handled by get_sb_block() */ 1832 case Opt_removed: 1833 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1834 return 1; 1835 case Opt_abort: 1836 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1837 return 1; 1838 case Opt_i_version: 1839 sb->s_flags |= SB_I_VERSION; 1840 return 1; 1841 case Opt_lazytime: 1842 sb->s_flags |= SB_LAZYTIME; 1843 return 1; 1844 case Opt_nolazytime: 1845 sb->s_flags &= ~SB_LAZYTIME; 1846 return 1; 1847 } 1848 1849 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1850 if (token == m->token) 1851 break; 1852 1853 if (m->token == Opt_err) { 1854 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1855 "or missing value", opt); 1856 return -1; 1857 } 1858 1859 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1860 ext4_msg(sb, KERN_ERR, 1861 "Mount option \"%s\" incompatible with ext2", opt); 1862 return -1; 1863 } 1864 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1865 ext4_msg(sb, KERN_ERR, 1866 "Mount option \"%s\" incompatible with ext3", opt); 1867 return -1; 1868 } 1869 1870 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1871 return -1; 1872 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1873 return -1; 1874 if (m->flags & MOPT_EXPLICIT) { 1875 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 1876 set_opt2(sb, EXPLICIT_DELALLOC); 1877 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 1878 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1879 } else 1880 return -1; 1881 } 1882 if (m->flags & MOPT_CLEAR_ERR) 1883 clear_opt(sb, ERRORS_MASK); 1884 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1885 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1886 "options when quota turned on"); 1887 return -1; 1888 } 1889 1890 if (m->flags & MOPT_NOSUPPORT) { 1891 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1892 } else if (token == Opt_commit) { 1893 if (arg == 0) 1894 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1895 else if (arg > INT_MAX / HZ) { 1896 ext4_msg(sb, KERN_ERR, 1897 "Invalid commit interval %d, " 1898 "must be smaller than %d", 1899 arg, INT_MAX / HZ); 1900 return -1; 1901 } 1902 sbi->s_commit_interval = HZ * arg; 1903 } else if (token == Opt_debug_want_extra_isize) { 1904 sbi->s_want_extra_isize = arg; 1905 } else if (token == Opt_max_batch_time) { 1906 sbi->s_max_batch_time = arg; 1907 } else if (token == Opt_min_batch_time) { 1908 sbi->s_min_batch_time = arg; 1909 } else if (token == Opt_inode_readahead_blks) { 1910 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1911 ext4_msg(sb, KERN_ERR, 1912 "EXT4-fs: inode_readahead_blks must be " 1913 "0 or a power of 2 smaller than 2^31"); 1914 return -1; 1915 } 1916 sbi->s_inode_readahead_blks = arg; 1917 } else if (token == Opt_init_itable) { 1918 set_opt(sb, INIT_INODE_TABLE); 1919 if (!args->from) 1920 arg = EXT4_DEF_LI_WAIT_MULT; 1921 sbi->s_li_wait_mult = arg; 1922 } else if (token == Opt_max_dir_size_kb) { 1923 sbi->s_max_dir_size_kb = arg; 1924 } else if (token == Opt_stripe) { 1925 sbi->s_stripe = arg; 1926 } else if (token == Opt_resuid) { 1927 uid = make_kuid(current_user_ns(), arg); 1928 if (!uid_valid(uid)) { 1929 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1930 return -1; 1931 } 1932 sbi->s_resuid = uid; 1933 } else if (token == Opt_resgid) { 1934 gid = make_kgid(current_user_ns(), arg); 1935 if (!gid_valid(gid)) { 1936 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1937 return -1; 1938 } 1939 sbi->s_resgid = gid; 1940 } else if (token == Opt_journal_dev) { 1941 if (is_remount) { 1942 ext4_msg(sb, KERN_ERR, 1943 "Cannot specify journal on remount"); 1944 return -1; 1945 } 1946 *journal_devnum = arg; 1947 } else if (token == Opt_journal_path) { 1948 char *journal_path; 1949 struct inode *journal_inode; 1950 struct path path; 1951 int error; 1952 1953 if (is_remount) { 1954 ext4_msg(sb, KERN_ERR, 1955 "Cannot specify journal on remount"); 1956 return -1; 1957 } 1958 journal_path = match_strdup(&args[0]); 1959 if (!journal_path) { 1960 ext4_msg(sb, KERN_ERR, "error: could not dup " 1961 "journal device string"); 1962 return -1; 1963 } 1964 1965 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1966 if (error) { 1967 ext4_msg(sb, KERN_ERR, "error: could not find " 1968 "journal device path: error %d", error); 1969 kfree(journal_path); 1970 return -1; 1971 } 1972 1973 journal_inode = d_inode(path.dentry); 1974 if (!S_ISBLK(journal_inode->i_mode)) { 1975 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1976 "is not a block device", journal_path); 1977 path_put(&path); 1978 kfree(journal_path); 1979 return -1; 1980 } 1981 1982 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1983 path_put(&path); 1984 kfree(journal_path); 1985 } else if (token == Opt_journal_ioprio) { 1986 if (arg > 7) { 1987 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1988 " (must be 0-7)"); 1989 return -1; 1990 } 1991 *journal_ioprio = 1992 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1993 } else if (token == Opt_test_dummy_encryption) { 1994 #ifdef CONFIG_FS_ENCRYPTION 1995 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 1996 ext4_msg(sb, KERN_WARNING, 1997 "Test dummy encryption mode enabled"); 1998 #else 1999 ext4_msg(sb, KERN_WARNING, 2000 "Test dummy encryption mount option ignored"); 2001 #endif 2002 } else if (m->flags & MOPT_DATAJ) { 2003 if (is_remount) { 2004 if (!sbi->s_journal) 2005 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 2006 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 2007 ext4_msg(sb, KERN_ERR, 2008 "Cannot change data mode on remount"); 2009 return -1; 2010 } 2011 } else { 2012 clear_opt(sb, DATA_FLAGS); 2013 sbi->s_mount_opt |= m->mount_opt; 2014 } 2015 #ifdef CONFIG_QUOTA 2016 } else if (m->flags & MOPT_QFMT) { 2017 if (sb_any_quota_loaded(sb) && 2018 sbi->s_jquota_fmt != m->mount_opt) { 2019 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 2020 "quota options when quota turned on"); 2021 return -1; 2022 } 2023 if (ext4_has_feature_quota(sb)) { 2024 ext4_msg(sb, KERN_INFO, 2025 "Quota format mount options ignored " 2026 "when QUOTA feature is enabled"); 2027 return 1; 2028 } 2029 sbi->s_jquota_fmt = m->mount_opt; 2030 #endif 2031 } else if (token == Opt_dax) { 2032 #ifdef CONFIG_FS_DAX 2033 ext4_msg(sb, KERN_WARNING, 2034 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 2035 sbi->s_mount_opt |= m->mount_opt; 2036 #else 2037 ext4_msg(sb, KERN_INFO, "dax option not supported"); 2038 return -1; 2039 #endif 2040 } else if (token == Opt_data_err_abort) { 2041 sbi->s_mount_opt |= m->mount_opt; 2042 } else if (token == Opt_data_err_ignore) { 2043 sbi->s_mount_opt &= ~m->mount_opt; 2044 } else { 2045 if (!args->from) 2046 arg = 1; 2047 if (m->flags & MOPT_CLEAR) 2048 arg = !arg; 2049 else if (unlikely(!(m->flags & MOPT_SET))) { 2050 ext4_msg(sb, KERN_WARNING, 2051 "buggy handling of option %s", opt); 2052 WARN_ON(1); 2053 return -1; 2054 } 2055 if (arg != 0) 2056 sbi->s_mount_opt |= m->mount_opt; 2057 else 2058 sbi->s_mount_opt &= ~m->mount_opt; 2059 } 2060 return 1; 2061 } 2062 2063 static int parse_options(char *options, struct super_block *sb, 2064 unsigned long *journal_devnum, 2065 unsigned int *journal_ioprio, 2066 int is_remount) 2067 { 2068 struct ext4_sb_info *sbi = EXT4_SB(sb); 2069 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2070 substring_t args[MAX_OPT_ARGS]; 2071 int token; 2072 2073 if (!options) 2074 return 1; 2075 2076 while ((p = strsep(&options, ",")) != NULL) { 2077 if (!*p) 2078 continue; 2079 /* 2080 * Initialize args struct so we know whether arg was 2081 * found; some options take optional arguments. 2082 */ 2083 args[0].to = args[0].from = NULL; 2084 token = match_token(p, tokens, args); 2085 if (handle_mount_opt(sb, p, token, args, journal_devnum, 2086 journal_ioprio, is_remount) < 0) 2087 return 0; 2088 } 2089 #ifdef CONFIG_QUOTA 2090 /* 2091 * We do the test below only for project quotas. 'usrquota' and 2092 * 'grpquota' mount options are allowed even without quota feature 2093 * to support legacy quotas in quota files. 2094 */ 2095 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 2096 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 2097 "Cannot enable project quota enforcement."); 2098 return 0; 2099 } 2100 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 2101 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 2102 if (usr_qf_name || grp_qf_name) { 2103 if (test_opt(sb, USRQUOTA) && usr_qf_name) 2104 clear_opt(sb, USRQUOTA); 2105 2106 if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2107 clear_opt(sb, GRPQUOTA); 2108 2109 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2110 ext4_msg(sb, KERN_ERR, "old and new quota " 2111 "format mixing"); 2112 return 0; 2113 } 2114 2115 if (!sbi->s_jquota_fmt) { 2116 ext4_msg(sb, KERN_ERR, "journaled quota format " 2117 "not specified"); 2118 return 0; 2119 } 2120 } 2121 #endif 2122 if (test_opt(sb, DIOREAD_NOLOCK)) { 2123 int blocksize = 2124 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2125 2126 if (blocksize < PAGE_SIZE) { 2127 ext4_msg(sb, KERN_ERR, "can't mount with " 2128 "dioread_nolock if block size != PAGE_SIZE"); 2129 return 0; 2130 } 2131 } 2132 return 1; 2133 } 2134 2135 static inline void ext4_show_quota_options(struct seq_file *seq, 2136 struct super_block *sb) 2137 { 2138 #if defined(CONFIG_QUOTA) 2139 struct ext4_sb_info *sbi = EXT4_SB(sb); 2140 char *usr_qf_name, *grp_qf_name; 2141 2142 if (sbi->s_jquota_fmt) { 2143 char *fmtname = ""; 2144 2145 switch (sbi->s_jquota_fmt) { 2146 case QFMT_VFS_OLD: 2147 fmtname = "vfsold"; 2148 break; 2149 case QFMT_VFS_V0: 2150 fmtname = "vfsv0"; 2151 break; 2152 case QFMT_VFS_V1: 2153 fmtname = "vfsv1"; 2154 break; 2155 } 2156 seq_printf(seq, ",jqfmt=%s", fmtname); 2157 } 2158 2159 rcu_read_lock(); 2160 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2161 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2162 if (usr_qf_name) 2163 seq_show_option(seq, "usrjquota", usr_qf_name); 2164 if (grp_qf_name) 2165 seq_show_option(seq, "grpjquota", grp_qf_name); 2166 rcu_read_unlock(); 2167 #endif 2168 } 2169 2170 static const char *token2str(int token) 2171 { 2172 const struct match_token *t; 2173 2174 for (t = tokens; t->token != Opt_err; t++) 2175 if (t->token == token && !strchr(t->pattern, '=')) 2176 break; 2177 return t->pattern; 2178 } 2179 2180 /* 2181 * Show an option if 2182 * - it's set to a non-default value OR 2183 * - if the per-sb default is different from the global default 2184 */ 2185 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2186 int nodefs) 2187 { 2188 struct ext4_sb_info *sbi = EXT4_SB(sb); 2189 struct ext4_super_block *es = sbi->s_es; 2190 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2191 const struct mount_opts *m; 2192 char sep = nodefs ? '\n' : ','; 2193 2194 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2195 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2196 2197 if (sbi->s_sb_block != 1) 2198 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2199 2200 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2201 int want_set = m->flags & MOPT_SET; 2202 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2203 (m->flags & MOPT_CLEAR_ERR)) 2204 continue; 2205 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2206 continue; /* skip if same as the default */ 2207 if ((want_set && 2208 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2209 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2210 continue; /* select Opt_noFoo vs Opt_Foo */ 2211 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2212 } 2213 2214 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2215 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2216 SEQ_OPTS_PRINT("resuid=%u", 2217 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2218 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2219 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2220 SEQ_OPTS_PRINT("resgid=%u", 2221 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2222 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2223 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2224 SEQ_OPTS_PUTS("errors=remount-ro"); 2225 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2226 SEQ_OPTS_PUTS("errors=continue"); 2227 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2228 SEQ_OPTS_PUTS("errors=panic"); 2229 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2230 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2231 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2232 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2233 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2234 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2235 if (sb->s_flags & SB_I_VERSION) 2236 SEQ_OPTS_PUTS("i_version"); 2237 if (nodefs || sbi->s_stripe) 2238 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2239 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2240 (sbi->s_mount_opt ^ def_mount_opt)) { 2241 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2242 SEQ_OPTS_PUTS("data=journal"); 2243 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2244 SEQ_OPTS_PUTS("data=ordered"); 2245 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2246 SEQ_OPTS_PUTS("data=writeback"); 2247 } 2248 if (nodefs || 2249 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2250 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2251 sbi->s_inode_readahead_blks); 2252 2253 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2254 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2255 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2256 if (nodefs || sbi->s_max_dir_size_kb) 2257 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2258 if (test_opt(sb, DATA_ERR_ABORT)) 2259 SEQ_OPTS_PUTS("data_err=abort"); 2260 if (DUMMY_ENCRYPTION_ENABLED(sbi)) 2261 SEQ_OPTS_PUTS("test_dummy_encryption"); 2262 2263 ext4_show_quota_options(seq, sb); 2264 return 0; 2265 } 2266 2267 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2268 { 2269 return _ext4_show_options(seq, root->d_sb, 0); 2270 } 2271 2272 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2273 { 2274 struct super_block *sb = seq->private; 2275 int rc; 2276 2277 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 2278 rc = _ext4_show_options(seq, sb, 1); 2279 seq_puts(seq, "\n"); 2280 return rc; 2281 } 2282 2283 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2284 int read_only) 2285 { 2286 struct ext4_sb_info *sbi = EXT4_SB(sb); 2287 int err = 0; 2288 2289 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2290 ext4_msg(sb, KERN_ERR, "revision level too high, " 2291 "forcing read-only mode"); 2292 err = -EROFS; 2293 } 2294 if (read_only) 2295 goto done; 2296 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2297 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2298 "running e2fsck is recommended"); 2299 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2300 ext4_msg(sb, KERN_WARNING, 2301 "warning: mounting fs with errors, " 2302 "running e2fsck is recommended"); 2303 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2304 le16_to_cpu(es->s_mnt_count) >= 2305 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2306 ext4_msg(sb, KERN_WARNING, 2307 "warning: maximal mount count reached, " 2308 "running e2fsck is recommended"); 2309 else if (le32_to_cpu(es->s_checkinterval) && 2310 (ext4_get_tstamp(es, s_lastcheck) + 2311 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2312 ext4_msg(sb, KERN_WARNING, 2313 "warning: checktime reached, " 2314 "running e2fsck is recommended"); 2315 if (!sbi->s_journal) 2316 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2317 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2318 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2319 le16_add_cpu(&es->s_mnt_count, 1); 2320 ext4_update_tstamp(es, s_mtime); 2321 if (sbi->s_journal) 2322 ext4_set_feature_journal_needs_recovery(sb); 2323 2324 err = ext4_commit_super(sb, 1); 2325 done: 2326 if (test_opt(sb, DEBUG)) 2327 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2328 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2329 sb->s_blocksize, 2330 sbi->s_groups_count, 2331 EXT4_BLOCKS_PER_GROUP(sb), 2332 EXT4_INODES_PER_GROUP(sb), 2333 sbi->s_mount_opt, sbi->s_mount_opt2); 2334 2335 cleancache_init_fs(sb); 2336 return err; 2337 } 2338 2339 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2340 { 2341 struct ext4_sb_info *sbi = EXT4_SB(sb); 2342 struct flex_groups *new_groups; 2343 int size; 2344 2345 if (!sbi->s_log_groups_per_flex) 2346 return 0; 2347 2348 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2349 if (size <= sbi->s_flex_groups_allocated) 2350 return 0; 2351 2352 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2353 new_groups = kvzalloc(size, GFP_KERNEL); 2354 if (!new_groups) { 2355 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2356 size / (int) sizeof(struct flex_groups)); 2357 return -ENOMEM; 2358 } 2359 2360 if (sbi->s_flex_groups) { 2361 memcpy(new_groups, sbi->s_flex_groups, 2362 (sbi->s_flex_groups_allocated * 2363 sizeof(struct flex_groups))); 2364 kvfree(sbi->s_flex_groups); 2365 } 2366 sbi->s_flex_groups = new_groups; 2367 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2368 return 0; 2369 } 2370 2371 static int ext4_fill_flex_info(struct super_block *sb) 2372 { 2373 struct ext4_sb_info *sbi = EXT4_SB(sb); 2374 struct ext4_group_desc *gdp = NULL; 2375 ext4_group_t flex_group; 2376 int i, err; 2377 2378 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2379 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2380 sbi->s_log_groups_per_flex = 0; 2381 return 1; 2382 } 2383 2384 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2385 if (err) 2386 goto failed; 2387 2388 for (i = 0; i < sbi->s_groups_count; i++) { 2389 gdp = ext4_get_group_desc(sb, i, NULL); 2390 2391 flex_group = ext4_flex_group(sbi, i); 2392 atomic_add(ext4_free_inodes_count(sb, gdp), 2393 &sbi->s_flex_groups[flex_group].free_inodes); 2394 atomic64_add(ext4_free_group_clusters(sb, gdp), 2395 &sbi->s_flex_groups[flex_group].free_clusters); 2396 atomic_add(ext4_used_dirs_count(sb, gdp), 2397 &sbi->s_flex_groups[flex_group].used_dirs); 2398 } 2399 2400 return 1; 2401 failed: 2402 return 0; 2403 } 2404 2405 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2406 struct ext4_group_desc *gdp) 2407 { 2408 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2409 __u16 crc = 0; 2410 __le32 le_group = cpu_to_le32(block_group); 2411 struct ext4_sb_info *sbi = EXT4_SB(sb); 2412 2413 if (ext4_has_metadata_csum(sbi->s_sb)) { 2414 /* Use new metadata_csum algorithm */ 2415 __u32 csum32; 2416 __u16 dummy_csum = 0; 2417 2418 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2419 sizeof(le_group)); 2420 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2421 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2422 sizeof(dummy_csum)); 2423 offset += sizeof(dummy_csum); 2424 if (offset < sbi->s_desc_size) 2425 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2426 sbi->s_desc_size - offset); 2427 2428 crc = csum32 & 0xFFFF; 2429 goto out; 2430 } 2431 2432 /* old crc16 code */ 2433 if (!ext4_has_feature_gdt_csum(sb)) 2434 return 0; 2435 2436 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2437 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2438 crc = crc16(crc, (__u8 *)gdp, offset); 2439 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2440 /* for checksum of struct ext4_group_desc do the rest...*/ 2441 if (ext4_has_feature_64bit(sb) && 2442 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2443 crc = crc16(crc, (__u8 *)gdp + offset, 2444 le16_to_cpu(sbi->s_es->s_desc_size) - 2445 offset); 2446 2447 out: 2448 return cpu_to_le16(crc); 2449 } 2450 2451 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2452 struct ext4_group_desc *gdp) 2453 { 2454 if (ext4_has_group_desc_csum(sb) && 2455 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2456 return 0; 2457 2458 return 1; 2459 } 2460 2461 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2462 struct ext4_group_desc *gdp) 2463 { 2464 if (!ext4_has_group_desc_csum(sb)) 2465 return; 2466 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2467 } 2468 2469 /* Called at mount-time, super-block is locked */ 2470 static int ext4_check_descriptors(struct super_block *sb, 2471 ext4_fsblk_t sb_block, 2472 ext4_group_t *first_not_zeroed) 2473 { 2474 struct ext4_sb_info *sbi = EXT4_SB(sb); 2475 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2476 ext4_fsblk_t last_block; 2477 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2478 ext4_fsblk_t block_bitmap; 2479 ext4_fsblk_t inode_bitmap; 2480 ext4_fsblk_t inode_table; 2481 int flexbg_flag = 0; 2482 ext4_group_t i, grp = sbi->s_groups_count; 2483 2484 if (ext4_has_feature_flex_bg(sb)) 2485 flexbg_flag = 1; 2486 2487 ext4_debug("Checking group descriptors"); 2488 2489 for (i = 0; i < sbi->s_groups_count; i++) { 2490 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2491 2492 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2493 last_block = ext4_blocks_count(sbi->s_es) - 1; 2494 else 2495 last_block = first_block + 2496 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2497 2498 if ((grp == sbi->s_groups_count) && 2499 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2500 grp = i; 2501 2502 block_bitmap = ext4_block_bitmap(sb, gdp); 2503 if (block_bitmap == sb_block) { 2504 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2505 "Block bitmap for group %u overlaps " 2506 "superblock", i); 2507 if (!sb_rdonly(sb)) 2508 return 0; 2509 } 2510 if (block_bitmap >= sb_block + 1 && 2511 block_bitmap <= last_bg_block) { 2512 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2513 "Block bitmap for group %u overlaps " 2514 "block group descriptors", i); 2515 if (!sb_rdonly(sb)) 2516 return 0; 2517 } 2518 if (block_bitmap < first_block || block_bitmap > last_block) { 2519 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2520 "Block bitmap for group %u not in group " 2521 "(block %llu)!", i, block_bitmap); 2522 return 0; 2523 } 2524 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2525 if (inode_bitmap == sb_block) { 2526 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2527 "Inode bitmap for group %u overlaps " 2528 "superblock", i); 2529 if (!sb_rdonly(sb)) 2530 return 0; 2531 } 2532 if (inode_bitmap >= sb_block + 1 && 2533 inode_bitmap <= last_bg_block) { 2534 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2535 "Inode bitmap for group %u overlaps " 2536 "block group descriptors", i); 2537 if (!sb_rdonly(sb)) 2538 return 0; 2539 } 2540 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2541 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2542 "Inode bitmap for group %u not in group " 2543 "(block %llu)!", i, inode_bitmap); 2544 return 0; 2545 } 2546 inode_table = ext4_inode_table(sb, gdp); 2547 if (inode_table == sb_block) { 2548 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2549 "Inode table for group %u overlaps " 2550 "superblock", i); 2551 if (!sb_rdonly(sb)) 2552 return 0; 2553 } 2554 if (inode_table >= sb_block + 1 && 2555 inode_table <= last_bg_block) { 2556 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2557 "Inode table for group %u overlaps " 2558 "block group descriptors", i); 2559 if (!sb_rdonly(sb)) 2560 return 0; 2561 } 2562 if (inode_table < first_block || 2563 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2564 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2565 "Inode table for group %u not in group " 2566 "(block %llu)!", i, inode_table); 2567 return 0; 2568 } 2569 ext4_lock_group(sb, i); 2570 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2571 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2572 "Checksum for group %u failed (%u!=%u)", 2573 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2574 gdp)), le16_to_cpu(gdp->bg_checksum)); 2575 if (!sb_rdonly(sb)) { 2576 ext4_unlock_group(sb, i); 2577 return 0; 2578 } 2579 } 2580 ext4_unlock_group(sb, i); 2581 if (!flexbg_flag) 2582 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2583 } 2584 if (NULL != first_not_zeroed) 2585 *first_not_zeroed = grp; 2586 return 1; 2587 } 2588 2589 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2590 * the superblock) which were deleted from all directories, but held open by 2591 * a process at the time of a crash. We walk the list and try to delete these 2592 * inodes at recovery time (only with a read-write filesystem). 2593 * 2594 * In order to keep the orphan inode chain consistent during traversal (in 2595 * case of crash during recovery), we link each inode into the superblock 2596 * orphan list_head and handle it the same way as an inode deletion during 2597 * normal operation (which journals the operations for us). 2598 * 2599 * We only do an iget() and an iput() on each inode, which is very safe if we 2600 * accidentally point at an in-use or already deleted inode. The worst that 2601 * can happen in this case is that we get a "bit already cleared" message from 2602 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2603 * e2fsck was run on this filesystem, and it must have already done the orphan 2604 * inode cleanup for us, so we can safely abort without any further action. 2605 */ 2606 static void ext4_orphan_cleanup(struct super_block *sb, 2607 struct ext4_super_block *es) 2608 { 2609 unsigned int s_flags = sb->s_flags; 2610 int ret, nr_orphans = 0, nr_truncates = 0; 2611 #ifdef CONFIG_QUOTA 2612 int quota_update = 0; 2613 int i; 2614 #endif 2615 if (!es->s_last_orphan) { 2616 jbd_debug(4, "no orphan inodes to clean up\n"); 2617 return; 2618 } 2619 2620 if (bdev_read_only(sb->s_bdev)) { 2621 ext4_msg(sb, KERN_ERR, "write access " 2622 "unavailable, skipping orphan cleanup"); 2623 return; 2624 } 2625 2626 /* Check if feature set would not allow a r/w mount */ 2627 if (!ext4_feature_set_ok(sb, 0)) { 2628 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2629 "unknown ROCOMPAT features"); 2630 return; 2631 } 2632 2633 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2634 /* don't clear list on RO mount w/ errors */ 2635 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 2636 ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2637 "clearing orphan list.\n"); 2638 es->s_last_orphan = 0; 2639 } 2640 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2641 return; 2642 } 2643 2644 if (s_flags & SB_RDONLY) { 2645 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2646 sb->s_flags &= ~SB_RDONLY; 2647 } 2648 #ifdef CONFIG_QUOTA 2649 /* Needed for iput() to work correctly and not trash data */ 2650 sb->s_flags |= SB_ACTIVE; 2651 2652 /* 2653 * Turn on quotas which were not enabled for read-only mounts if 2654 * filesystem has quota feature, so that they are updated correctly. 2655 */ 2656 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 2657 int ret = ext4_enable_quotas(sb); 2658 2659 if (!ret) 2660 quota_update = 1; 2661 else 2662 ext4_msg(sb, KERN_ERR, 2663 "Cannot turn on quotas: error %d", ret); 2664 } 2665 2666 /* Turn on journaled quotas used for old sytle */ 2667 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2668 if (EXT4_SB(sb)->s_qf_names[i]) { 2669 int ret = ext4_quota_on_mount(sb, i); 2670 2671 if (!ret) 2672 quota_update = 1; 2673 else 2674 ext4_msg(sb, KERN_ERR, 2675 "Cannot turn on journaled " 2676 "quota: type %d: error %d", i, ret); 2677 } 2678 } 2679 #endif 2680 2681 while (es->s_last_orphan) { 2682 struct inode *inode; 2683 2684 /* 2685 * We may have encountered an error during cleanup; if 2686 * so, skip the rest. 2687 */ 2688 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2689 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2690 es->s_last_orphan = 0; 2691 break; 2692 } 2693 2694 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2695 if (IS_ERR(inode)) { 2696 es->s_last_orphan = 0; 2697 break; 2698 } 2699 2700 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2701 dquot_initialize(inode); 2702 if (inode->i_nlink) { 2703 if (test_opt(sb, DEBUG)) 2704 ext4_msg(sb, KERN_DEBUG, 2705 "%s: truncating inode %lu to %lld bytes", 2706 __func__, inode->i_ino, inode->i_size); 2707 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2708 inode->i_ino, inode->i_size); 2709 inode_lock(inode); 2710 truncate_inode_pages(inode->i_mapping, inode->i_size); 2711 ret = ext4_truncate(inode); 2712 if (ret) 2713 ext4_std_error(inode->i_sb, ret); 2714 inode_unlock(inode); 2715 nr_truncates++; 2716 } else { 2717 if (test_opt(sb, DEBUG)) 2718 ext4_msg(sb, KERN_DEBUG, 2719 "%s: deleting unreferenced inode %lu", 2720 __func__, inode->i_ino); 2721 jbd_debug(2, "deleting unreferenced inode %lu\n", 2722 inode->i_ino); 2723 nr_orphans++; 2724 } 2725 iput(inode); /* The delete magic happens here! */ 2726 } 2727 2728 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2729 2730 if (nr_orphans) 2731 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2732 PLURAL(nr_orphans)); 2733 if (nr_truncates) 2734 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2735 PLURAL(nr_truncates)); 2736 #ifdef CONFIG_QUOTA 2737 /* Turn off quotas if they were enabled for orphan cleanup */ 2738 if (quota_update) { 2739 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2740 if (sb_dqopt(sb)->files[i]) 2741 dquot_quota_off(sb, i); 2742 } 2743 } 2744 #endif 2745 sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2746 } 2747 2748 /* 2749 * Maximal extent format file size. 2750 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2751 * extent format containers, within a sector_t, and within i_blocks 2752 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2753 * so that won't be a limiting factor. 2754 * 2755 * However there is other limiting factor. We do store extents in the form 2756 * of starting block and length, hence the resulting length of the extent 2757 * covering maximum file size must fit into on-disk format containers as 2758 * well. Given that length is always by 1 unit bigger than max unit (because 2759 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2760 * 2761 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2762 */ 2763 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2764 { 2765 loff_t res; 2766 loff_t upper_limit = MAX_LFS_FILESIZE; 2767 2768 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 2769 2770 if (!has_huge_files) { 2771 upper_limit = (1LL << 32) - 1; 2772 2773 /* total blocks in file system block size */ 2774 upper_limit >>= (blkbits - 9); 2775 upper_limit <<= blkbits; 2776 } 2777 2778 /* 2779 * 32-bit extent-start container, ee_block. We lower the maxbytes 2780 * by one fs block, so ee_len can cover the extent of maximum file 2781 * size 2782 */ 2783 res = (1LL << 32) - 1; 2784 res <<= blkbits; 2785 2786 /* Sanity check against vm- & vfs- imposed limits */ 2787 if (res > upper_limit) 2788 res = upper_limit; 2789 2790 return res; 2791 } 2792 2793 /* 2794 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2795 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2796 * We need to be 1 filesystem block less than the 2^48 sector limit. 2797 */ 2798 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2799 { 2800 loff_t res = EXT4_NDIR_BLOCKS; 2801 int meta_blocks; 2802 loff_t upper_limit; 2803 /* This is calculated to be the largest file size for a dense, block 2804 * mapped file such that the file's total number of 512-byte sectors, 2805 * including data and all indirect blocks, does not exceed (2^48 - 1). 2806 * 2807 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2808 * number of 512-byte sectors of the file. 2809 */ 2810 2811 if (!has_huge_files) { 2812 /* 2813 * !has_huge_files or implies that the inode i_block field 2814 * represents total file blocks in 2^32 512-byte sectors == 2815 * size of vfs inode i_blocks * 8 2816 */ 2817 upper_limit = (1LL << 32) - 1; 2818 2819 /* total blocks in file system block size */ 2820 upper_limit >>= (bits - 9); 2821 2822 } else { 2823 /* 2824 * We use 48 bit ext4_inode i_blocks 2825 * With EXT4_HUGE_FILE_FL set the i_blocks 2826 * represent total number of blocks in 2827 * file system block size 2828 */ 2829 upper_limit = (1LL << 48) - 1; 2830 2831 } 2832 2833 /* indirect blocks */ 2834 meta_blocks = 1; 2835 /* double indirect blocks */ 2836 meta_blocks += 1 + (1LL << (bits-2)); 2837 /* tripple indirect blocks */ 2838 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2839 2840 upper_limit -= meta_blocks; 2841 upper_limit <<= bits; 2842 2843 res += 1LL << (bits-2); 2844 res += 1LL << (2*(bits-2)); 2845 res += 1LL << (3*(bits-2)); 2846 res <<= bits; 2847 if (res > upper_limit) 2848 res = upper_limit; 2849 2850 if (res > MAX_LFS_FILESIZE) 2851 res = MAX_LFS_FILESIZE; 2852 2853 return res; 2854 } 2855 2856 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2857 ext4_fsblk_t logical_sb_block, int nr) 2858 { 2859 struct ext4_sb_info *sbi = EXT4_SB(sb); 2860 ext4_group_t bg, first_meta_bg; 2861 int has_super = 0; 2862 2863 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2864 2865 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 2866 return logical_sb_block + nr + 1; 2867 bg = sbi->s_desc_per_block * nr; 2868 if (ext4_bg_has_super(sb, bg)) 2869 has_super = 1; 2870 2871 /* 2872 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2873 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2874 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2875 * compensate. 2876 */ 2877 if (sb->s_blocksize == 1024 && nr == 0 && 2878 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 2879 has_super++; 2880 2881 return (has_super + ext4_group_first_block_no(sb, bg)); 2882 } 2883 2884 /** 2885 * ext4_get_stripe_size: Get the stripe size. 2886 * @sbi: In memory super block info 2887 * 2888 * If we have specified it via mount option, then 2889 * use the mount option value. If the value specified at mount time is 2890 * greater than the blocks per group use the super block value. 2891 * If the super block value is greater than blocks per group return 0. 2892 * Allocator needs it be less than blocks per group. 2893 * 2894 */ 2895 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2896 { 2897 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2898 unsigned long stripe_width = 2899 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2900 int ret; 2901 2902 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2903 ret = sbi->s_stripe; 2904 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 2905 ret = stripe_width; 2906 else if (stride && stride <= sbi->s_blocks_per_group) 2907 ret = stride; 2908 else 2909 ret = 0; 2910 2911 /* 2912 * If the stripe width is 1, this makes no sense and 2913 * we set it to 0 to turn off stripe handling code. 2914 */ 2915 if (ret <= 1) 2916 ret = 0; 2917 2918 return ret; 2919 } 2920 2921 /* 2922 * Check whether this filesystem can be mounted based on 2923 * the features present and the RDONLY/RDWR mount requested. 2924 * Returns 1 if this filesystem can be mounted as requested, 2925 * 0 if it cannot be. 2926 */ 2927 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2928 { 2929 if (ext4_has_unknown_ext4_incompat_features(sb)) { 2930 ext4_msg(sb, KERN_ERR, 2931 "Couldn't mount because of " 2932 "unsupported optional features (%x)", 2933 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2934 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2935 return 0; 2936 } 2937 2938 #ifndef CONFIG_UNICODE 2939 if (ext4_has_feature_casefold(sb)) { 2940 ext4_msg(sb, KERN_ERR, 2941 "Filesystem with casefold feature cannot be " 2942 "mounted without CONFIG_UNICODE"); 2943 return 0; 2944 } 2945 #endif 2946 2947 if (readonly) 2948 return 1; 2949 2950 if (ext4_has_feature_readonly(sb)) { 2951 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 2952 sb->s_flags |= SB_RDONLY; 2953 return 1; 2954 } 2955 2956 /* Check that feature set is OK for a read-write mount */ 2957 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2958 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2959 "unsupported optional features (%x)", 2960 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2961 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2962 return 0; 2963 } 2964 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2965 ext4_msg(sb, KERN_ERR, 2966 "Can't support bigalloc feature without " 2967 "extents feature\n"); 2968 return 0; 2969 } 2970 2971 #ifndef CONFIG_QUOTA 2972 if (ext4_has_feature_quota(sb) && !readonly) { 2973 ext4_msg(sb, KERN_ERR, 2974 "Filesystem with quota feature cannot be mounted RDWR " 2975 "without CONFIG_QUOTA"); 2976 return 0; 2977 } 2978 if (ext4_has_feature_project(sb) && !readonly) { 2979 ext4_msg(sb, KERN_ERR, 2980 "Filesystem with project quota feature cannot be mounted RDWR " 2981 "without CONFIG_QUOTA"); 2982 return 0; 2983 } 2984 #endif /* CONFIG_QUOTA */ 2985 return 1; 2986 } 2987 2988 /* 2989 * This function is called once a day if we have errors logged 2990 * on the file system 2991 */ 2992 static void print_daily_error_info(struct timer_list *t) 2993 { 2994 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 2995 struct super_block *sb = sbi->s_sb; 2996 struct ext4_super_block *es = sbi->s_es; 2997 2998 if (es->s_error_count) 2999 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3000 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3001 le32_to_cpu(es->s_error_count)); 3002 if (es->s_first_error_time) { 3003 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3004 sb->s_id, 3005 ext4_get_tstamp(es, s_first_error_time), 3006 (int) sizeof(es->s_first_error_func), 3007 es->s_first_error_func, 3008 le32_to_cpu(es->s_first_error_line)); 3009 if (es->s_first_error_ino) 3010 printk(KERN_CONT ": inode %u", 3011 le32_to_cpu(es->s_first_error_ino)); 3012 if (es->s_first_error_block) 3013 printk(KERN_CONT ": block %llu", (unsigned long long) 3014 le64_to_cpu(es->s_first_error_block)); 3015 printk(KERN_CONT "\n"); 3016 } 3017 if (es->s_last_error_time) { 3018 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3019 sb->s_id, 3020 ext4_get_tstamp(es, s_last_error_time), 3021 (int) sizeof(es->s_last_error_func), 3022 es->s_last_error_func, 3023 le32_to_cpu(es->s_last_error_line)); 3024 if (es->s_last_error_ino) 3025 printk(KERN_CONT ": inode %u", 3026 le32_to_cpu(es->s_last_error_ino)); 3027 if (es->s_last_error_block) 3028 printk(KERN_CONT ": block %llu", (unsigned long long) 3029 le64_to_cpu(es->s_last_error_block)); 3030 printk(KERN_CONT "\n"); 3031 } 3032 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3033 } 3034 3035 /* Find next suitable group and run ext4_init_inode_table */ 3036 static int ext4_run_li_request(struct ext4_li_request *elr) 3037 { 3038 struct ext4_group_desc *gdp = NULL; 3039 ext4_group_t group, ngroups; 3040 struct super_block *sb; 3041 unsigned long timeout = 0; 3042 int ret = 0; 3043 3044 sb = elr->lr_super; 3045 ngroups = EXT4_SB(sb)->s_groups_count; 3046 3047 for (group = elr->lr_next_group; group < ngroups; group++) { 3048 gdp = ext4_get_group_desc(sb, group, NULL); 3049 if (!gdp) { 3050 ret = 1; 3051 break; 3052 } 3053 3054 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3055 break; 3056 } 3057 3058 if (group >= ngroups) 3059 ret = 1; 3060 3061 if (!ret) { 3062 timeout = jiffies; 3063 ret = ext4_init_inode_table(sb, group, 3064 elr->lr_timeout ? 0 : 1); 3065 if (elr->lr_timeout == 0) { 3066 timeout = (jiffies - timeout) * 3067 elr->lr_sbi->s_li_wait_mult; 3068 elr->lr_timeout = timeout; 3069 } 3070 elr->lr_next_sched = jiffies + elr->lr_timeout; 3071 elr->lr_next_group = group + 1; 3072 } 3073 return ret; 3074 } 3075 3076 /* 3077 * Remove lr_request from the list_request and free the 3078 * request structure. Should be called with li_list_mtx held 3079 */ 3080 static void ext4_remove_li_request(struct ext4_li_request *elr) 3081 { 3082 struct ext4_sb_info *sbi; 3083 3084 if (!elr) 3085 return; 3086 3087 sbi = elr->lr_sbi; 3088 3089 list_del(&elr->lr_request); 3090 sbi->s_li_request = NULL; 3091 kfree(elr); 3092 } 3093 3094 static void ext4_unregister_li_request(struct super_block *sb) 3095 { 3096 mutex_lock(&ext4_li_mtx); 3097 if (!ext4_li_info) { 3098 mutex_unlock(&ext4_li_mtx); 3099 return; 3100 } 3101 3102 mutex_lock(&ext4_li_info->li_list_mtx); 3103 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3104 mutex_unlock(&ext4_li_info->li_list_mtx); 3105 mutex_unlock(&ext4_li_mtx); 3106 } 3107 3108 static struct task_struct *ext4_lazyinit_task; 3109 3110 /* 3111 * This is the function where ext4lazyinit thread lives. It walks 3112 * through the request list searching for next scheduled filesystem. 3113 * When such a fs is found, run the lazy initialization request 3114 * (ext4_rn_li_request) and keep track of the time spend in this 3115 * function. Based on that time we compute next schedule time of 3116 * the request. When walking through the list is complete, compute 3117 * next waking time and put itself into sleep. 3118 */ 3119 static int ext4_lazyinit_thread(void *arg) 3120 { 3121 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3122 struct list_head *pos, *n; 3123 struct ext4_li_request *elr; 3124 unsigned long next_wakeup, cur; 3125 3126 BUG_ON(NULL == eli); 3127 3128 cont_thread: 3129 while (true) { 3130 next_wakeup = MAX_JIFFY_OFFSET; 3131 3132 mutex_lock(&eli->li_list_mtx); 3133 if (list_empty(&eli->li_request_list)) { 3134 mutex_unlock(&eli->li_list_mtx); 3135 goto exit_thread; 3136 } 3137 list_for_each_safe(pos, n, &eli->li_request_list) { 3138 int err = 0; 3139 int progress = 0; 3140 elr = list_entry(pos, struct ext4_li_request, 3141 lr_request); 3142 3143 if (time_before(jiffies, elr->lr_next_sched)) { 3144 if (time_before(elr->lr_next_sched, next_wakeup)) 3145 next_wakeup = elr->lr_next_sched; 3146 continue; 3147 } 3148 if (down_read_trylock(&elr->lr_super->s_umount)) { 3149 if (sb_start_write_trylock(elr->lr_super)) { 3150 progress = 1; 3151 /* 3152 * We hold sb->s_umount, sb can not 3153 * be removed from the list, it is 3154 * now safe to drop li_list_mtx 3155 */ 3156 mutex_unlock(&eli->li_list_mtx); 3157 err = ext4_run_li_request(elr); 3158 sb_end_write(elr->lr_super); 3159 mutex_lock(&eli->li_list_mtx); 3160 n = pos->next; 3161 } 3162 up_read((&elr->lr_super->s_umount)); 3163 } 3164 /* error, remove the lazy_init job */ 3165 if (err) { 3166 ext4_remove_li_request(elr); 3167 continue; 3168 } 3169 if (!progress) { 3170 elr->lr_next_sched = jiffies + 3171 (prandom_u32() 3172 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3173 } 3174 if (time_before(elr->lr_next_sched, next_wakeup)) 3175 next_wakeup = elr->lr_next_sched; 3176 } 3177 mutex_unlock(&eli->li_list_mtx); 3178 3179 try_to_freeze(); 3180 3181 cur = jiffies; 3182 if ((time_after_eq(cur, next_wakeup)) || 3183 (MAX_JIFFY_OFFSET == next_wakeup)) { 3184 cond_resched(); 3185 continue; 3186 } 3187 3188 schedule_timeout_interruptible(next_wakeup - cur); 3189 3190 if (kthread_should_stop()) { 3191 ext4_clear_request_list(); 3192 goto exit_thread; 3193 } 3194 } 3195 3196 exit_thread: 3197 /* 3198 * It looks like the request list is empty, but we need 3199 * to check it under the li_list_mtx lock, to prevent any 3200 * additions into it, and of course we should lock ext4_li_mtx 3201 * to atomically free the list and ext4_li_info, because at 3202 * this point another ext4 filesystem could be registering 3203 * new one. 3204 */ 3205 mutex_lock(&ext4_li_mtx); 3206 mutex_lock(&eli->li_list_mtx); 3207 if (!list_empty(&eli->li_request_list)) { 3208 mutex_unlock(&eli->li_list_mtx); 3209 mutex_unlock(&ext4_li_mtx); 3210 goto cont_thread; 3211 } 3212 mutex_unlock(&eli->li_list_mtx); 3213 kfree(ext4_li_info); 3214 ext4_li_info = NULL; 3215 mutex_unlock(&ext4_li_mtx); 3216 3217 return 0; 3218 } 3219 3220 static void ext4_clear_request_list(void) 3221 { 3222 struct list_head *pos, *n; 3223 struct ext4_li_request *elr; 3224 3225 mutex_lock(&ext4_li_info->li_list_mtx); 3226 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3227 elr = list_entry(pos, struct ext4_li_request, 3228 lr_request); 3229 ext4_remove_li_request(elr); 3230 } 3231 mutex_unlock(&ext4_li_info->li_list_mtx); 3232 } 3233 3234 static int ext4_run_lazyinit_thread(void) 3235 { 3236 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3237 ext4_li_info, "ext4lazyinit"); 3238 if (IS_ERR(ext4_lazyinit_task)) { 3239 int err = PTR_ERR(ext4_lazyinit_task); 3240 ext4_clear_request_list(); 3241 kfree(ext4_li_info); 3242 ext4_li_info = NULL; 3243 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3244 "initialization thread\n", 3245 err); 3246 return err; 3247 } 3248 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3249 return 0; 3250 } 3251 3252 /* 3253 * Check whether it make sense to run itable init. thread or not. 3254 * If there is at least one uninitialized inode table, return 3255 * corresponding group number, else the loop goes through all 3256 * groups and return total number of groups. 3257 */ 3258 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3259 { 3260 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3261 struct ext4_group_desc *gdp = NULL; 3262 3263 if (!ext4_has_group_desc_csum(sb)) 3264 return ngroups; 3265 3266 for (group = 0; group < ngroups; group++) { 3267 gdp = ext4_get_group_desc(sb, group, NULL); 3268 if (!gdp) 3269 continue; 3270 3271 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3272 break; 3273 } 3274 3275 return group; 3276 } 3277 3278 static int ext4_li_info_new(void) 3279 { 3280 struct ext4_lazy_init *eli = NULL; 3281 3282 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3283 if (!eli) 3284 return -ENOMEM; 3285 3286 INIT_LIST_HEAD(&eli->li_request_list); 3287 mutex_init(&eli->li_list_mtx); 3288 3289 eli->li_state |= EXT4_LAZYINIT_QUIT; 3290 3291 ext4_li_info = eli; 3292 3293 return 0; 3294 } 3295 3296 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3297 ext4_group_t start) 3298 { 3299 struct ext4_sb_info *sbi = EXT4_SB(sb); 3300 struct ext4_li_request *elr; 3301 3302 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3303 if (!elr) 3304 return NULL; 3305 3306 elr->lr_super = sb; 3307 elr->lr_sbi = sbi; 3308 elr->lr_next_group = start; 3309 3310 /* 3311 * Randomize first schedule time of the request to 3312 * spread the inode table initialization requests 3313 * better. 3314 */ 3315 elr->lr_next_sched = jiffies + (prandom_u32() % 3316 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3317 return elr; 3318 } 3319 3320 int ext4_register_li_request(struct super_block *sb, 3321 ext4_group_t first_not_zeroed) 3322 { 3323 struct ext4_sb_info *sbi = EXT4_SB(sb); 3324 struct ext4_li_request *elr = NULL; 3325 ext4_group_t ngroups = sbi->s_groups_count; 3326 int ret = 0; 3327 3328 mutex_lock(&ext4_li_mtx); 3329 if (sbi->s_li_request != NULL) { 3330 /* 3331 * Reset timeout so it can be computed again, because 3332 * s_li_wait_mult might have changed. 3333 */ 3334 sbi->s_li_request->lr_timeout = 0; 3335 goto out; 3336 } 3337 3338 if (first_not_zeroed == ngroups || sb_rdonly(sb) || 3339 !test_opt(sb, INIT_INODE_TABLE)) 3340 goto out; 3341 3342 elr = ext4_li_request_new(sb, first_not_zeroed); 3343 if (!elr) { 3344 ret = -ENOMEM; 3345 goto out; 3346 } 3347 3348 if (NULL == ext4_li_info) { 3349 ret = ext4_li_info_new(); 3350 if (ret) 3351 goto out; 3352 } 3353 3354 mutex_lock(&ext4_li_info->li_list_mtx); 3355 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3356 mutex_unlock(&ext4_li_info->li_list_mtx); 3357 3358 sbi->s_li_request = elr; 3359 /* 3360 * set elr to NULL here since it has been inserted to 3361 * the request_list and the removal and free of it is 3362 * handled by ext4_clear_request_list from now on. 3363 */ 3364 elr = NULL; 3365 3366 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3367 ret = ext4_run_lazyinit_thread(); 3368 if (ret) 3369 goto out; 3370 } 3371 out: 3372 mutex_unlock(&ext4_li_mtx); 3373 if (ret) 3374 kfree(elr); 3375 return ret; 3376 } 3377 3378 /* 3379 * We do not need to lock anything since this is called on 3380 * module unload. 3381 */ 3382 static void ext4_destroy_lazyinit_thread(void) 3383 { 3384 /* 3385 * If thread exited earlier 3386 * there's nothing to be done. 3387 */ 3388 if (!ext4_li_info || !ext4_lazyinit_task) 3389 return; 3390 3391 kthread_stop(ext4_lazyinit_task); 3392 } 3393 3394 static int set_journal_csum_feature_set(struct super_block *sb) 3395 { 3396 int ret = 1; 3397 int compat, incompat; 3398 struct ext4_sb_info *sbi = EXT4_SB(sb); 3399 3400 if (ext4_has_metadata_csum(sb)) { 3401 /* journal checksum v3 */ 3402 compat = 0; 3403 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3404 } else { 3405 /* journal checksum v1 */ 3406 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3407 incompat = 0; 3408 } 3409 3410 jbd2_journal_clear_features(sbi->s_journal, 3411 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3412 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3413 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3414 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3415 ret = jbd2_journal_set_features(sbi->s_journal, 3416 compat, 0, 3417 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3418 incompat); 3419 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3420 ret = jbd2_journal_set_features(sbi->s_journal, 3421 compat, 0, 3422 incompat); 3423 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3424 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3425 } else { 3426 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3427 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3428 } 3429 3430 return ret; 3431 } 3432 3433 /* 3434 * Note: calculating the overhead so we can be compatible with 3435 * historical BSD practice is quite difficult in the face of 3436 * clusters/bigalloc. This is because multiple metadata blocks from 3437 * different block group can end up in the same allocation cluster. 3438 * Calculating the exact overhead in the face of clustered allocation 3439 * requires either O(all block bitmaps) in memory or O(number of block 3440 * groups**2) in time. We will still calculate the superblock for 3441 * older file systems --- and if we come across with a bigalloc file 3442 * system with zero in s_overhead_clusters the estimate will be close to 3443 * correct especially for very large cluster sizes --- but for newer 3444 * file systems, it's better to calculate this figure once at mkfs 3445 * time, and store it in the superblock. If the superblock value is 3446 * present (even for non-bigalloc file systems), we will use it. 3447 */ 3448 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3449 char *buf) 3450 { 3451 struct ext4_sb_info *sbi = EXT4_SB(sb); 3452 struct ext4_group_desc *gdp; 3453 ext4_fsblk_t first_block, last_block, b; 3454 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3455 int s, j, count = 0; 3456 3457 if (!ext4_has_feature_bigalloc(sb)) 3458 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3459 sbi->s_itb_per_group + 2); 3460 3461 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3462 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3463 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3464 for (i = 0; i < ngroups; i++) { 3465 gdp = ext4_get_group_desc(sb, i, NULL); 3466 b = ext4_block_bitmap(sb, gdp); 3467 if (b >= first_block && b <= last_block) { 3468 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3469 count++; 3470 } 3471 b = ext4_inode_bitmap(sb, gdp); 3472 if (b >= first_block && b <= last_block) { 3473 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3474 count++; 3475 } 3476 b = ext4_inode_table(sb, gdp); 3477 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3478 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3479 int c = EXT4_B2C(sbi, b - first_block); 3480 ext4_set_bit(c, buf); 3481 count++; 3482 } 3483 if (i != grp) 3484 continue; 3485 s = 0; 3486 if (ext4_bg_has_super(sb, grp)) { 3487 ext4_set_bit(s++, buf); 3488 count++; 3489 } 3490 j = ext4_bg_num_gdb(sb, grp); 3491 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3492 ext4_error(sb, "Invalid number of block group " 3493 "descriptor blocks: %d", j); 3494 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3495 } 3496 count += j; 3497 for (; j > 0; j--) 3498 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3499 } 3500 if (!count) 3501 return 0; 3502 return EXT4_CLUSTERS_PER_GROUP(sb) - 3503 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3504 } 3505 3506 /* 3507 * Compute the overhead and stash it in sbi->s_overhead 3508 */ 3509 int ext4_calculate_overhead(struct super_block *sb) 3510 { 3511 struct ext4_sb_info *sbi = EXT4_SB(sb); 3512 struct ext4_super_block *es = sbi->s_es; 3513 struct inode *j_inode; 3514 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3515 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3516 ext4_fsblk_t overhead = 0; 3517 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3518 3519 if (!buf) 3520 return -ENOMEM; 3521 3522 /* 3523 * Compute the overhead (FS structures). This is constant 3524 * for a given filesystem unless the number of block groups 3525 * changes so we cache the previous value until it does. 3526 */ 3527 3528 /* 3529 * All of the blocks before first_data_block are overhead 3530 */ 3531 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3532 3533 /* 3534 * Add the overhead found in each block group 3535 */ 3536 for (i = 0; i < ngroups; i++) { 3537 int blks; 3538 3539 blks = count_overhead(sb, i, buf); 3540 overhead += blks; 3541 if (blks) 3542 memset(buf, 0, PAGE_SIZE); 3543 cond_resched(); 3544 } 3545 3546 /* 3547 * Add the internal journal blocks whether the journal has been 3548 * loaded or not 3549 */ 3550 if (sbi->s_journal && !sbi->journal_bdev) 3551 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3552 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { 3553 j_inode = ext4_get_journal_inode(sb, j_inum); 3554 if (j_inode) { 3555 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3556 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3557 iput(j_inode); 3558 } else { 3559 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3560 } 3561 } 3562 sbi->s_overhead = overhead; 3563 smp_wmb(); 3564 free_page((unsigned long) buf); 3565 return 0; 3566 } 3567 3568 static void ext4_clamp_want_extra_isize(struct super_block *sb) 3569 { 3570 struct ext4_sb_info *sbi = EXT4_SB(sb); 3571 struct ext4_super_block *es = sbi->s_es; 3572 3573 /* determine the minimum size of new large inodes, if present */ 3574 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE && 3575 sbi->s_want_extra_isize == 0) { 3576 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3577 EXT4_GOOD_OLD_INODE_SIZE; 3578 if (ext4_has_feature_extra_isize(sb)) { 3579 if (sbi->s_want_extra_isize < 3580 le16_to_cpu(es->s_want_extra_isize)) 3581 sbi->s_want_extra_isize = 3582 le16_to_cpu(es->s_want_extra_isize); 3583 if (sbi->s_want_extra_isize < 3584 le16_to_cpu(es->s_min_extra_isize)) 3585 sbi->s_want_extra_isize = 3586 le16_to_cpu(es->s_min_extra_isize); 3587 } 3588 } 3589 /* Check if enough inode space is available */ 3590 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 3591 sbi->s_inode_size) { 3592 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3593 EXT4_GOOD_OLD_INODE_SIZE; 3594 ext4_msg(sb, KERN_INFO, 3595 "required extra inode space not available"); 3596 } 3597 } 3598 3599 static void ext4_set_resv_clusters(struct super_block *sb) 3600 { 3601 ext4_fsblk_t resv_clusters; 3602 struct ext4_sb_info *sbi = EXT4_SB(sb); 3603 3604 /* 3605 * There's no need to reserve anything when we aren't using extents. 3606 * The space estimates are exact, there are no unwritten extents, 3607 * hole punching doesn't need new metadata... This is needed especially 3608 * to keep ext2/3 backward compatibility. 3609 */ 3610 if (!ext4_has_feature_extents(sb)) 3611 return; 3612 /* 3613 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3614 * This should cover the situations where we can not afford to run 3615 * out of space like for example punch hole, or converting 3616 * unwritten extents in delalloc path. In most cases such 3617 * allocation would require 1, or 2 blocks, higher numbers are 3618 * very rare. 3619 */ 3620 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3621 sbi->s_cluster_bits); 3622 3623 do_div(resv_clusters, 50); 3624 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3625 3626 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3627 } 3628 3629 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3630 { 3631 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3632 char *orig_data = kstrdup(data, GFP_KERNEL); 3633 struct buffer_head *bh; 3634 struct ext4_super_block *es = NULL; 3635 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3636 ext4_fsblk_t block; 3637 ext4_fsblk_t sb_block = get_sb_block(&data); 3638 ext4_fsblk_t logical_sb_block; 3639 unsigned long offset = 0; 3640 unsigned long journal_devnum = 0; 3641 unsigned long def_mount_opts; 3642 struct inode *root; 3643 const char *descr; 3644 int ret = -ENOMEM; 3645 int blocksize, clustersize; 3646 unsigned int db_count; 3647 unsigned int i; 3648 int needs_recovery, has_huge_files, has_bigalloc; 3649 __u64 blocks_count; 3650 int err = 0; 3651 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3652 ext4_group_t first_not_zeroed; 3653 3654 if ((data && !orig_data) || !sbi) 3655 goto out_free_base; 3656 3657 sbi->s_daxdev = dax_dev; 3658 sbi->s_blockgroup_lock = 3659 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3660 if (!sbi->s_blockgroup_lock) 3661 goto out_free_base; 3662 3663 sb->s_fs_info = sbi; 3664 sbi->s_sb = sb; 3665 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3666 sbi->s_sb_block = sb_block; 3667 if (sb->s_bdev->bd_part) 3668 sbi->s_sectors_written_start = 3669 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3670 3671 /* Cleanup superblock name */ 3672 strreplace(sb->s_id, '/', '!'); 3673 3674 /* -EINVAL is default */ 3675 ret = -EINVAL; 3676 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3677 if (!blocksize) { 3678 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3679 goto out_fail; 3680 } 3681 3682 /* 3683 * The ext4 superblock will not be buffer aligned for other than 1kB 3684 * block sizes. We need to calculate the offset from buffer start. 3685 */ 3686 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3687 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3688 offset = do_div(logical_sb_block, blocksize); 3689 } else { 3690 logical_sb_block = sb_block; 3691 } 3692 3693 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3694 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3695 goto out_fail; 3696 } 3697 /* 3698 * Note: s_es must be initialized as soon as possible because 3699 * some ext4 macro-instructions depend on its value 3700 */ 3701 es = (struct ext4_super_block *) (bh->b_data + offset); 3702 sbi->s_es = es; 3703 sb->s_magic = le16_to_cpu(es->s_magic); 3704 if (sb->s_magic != EXT4_SUPER_MAGIC) 3705 goto cantfind_ext4; 3706 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3707 3708 /* Warn if metadata_csum and gdt_csum are both set. */ 3709 if (ext4_has_feature_metadata_csum(sb) && 3710 ext4_has_feature_gdt_csum(sb)) 3711 ext4_warning(sb, "metadata_csum and uninit_bg are " 3712 "redundant flags; please run fsck."); 3713 3714 /* Check for a known checksum algorithm */ 3715 if (!ext4_verify_csum_type(sb, es)) { 3716 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3717 "unknown checksum algorithm."); 3718 silent = 1; 3719 goto cantfind_ext4; 3720 } 3721 3722 /* Load the checksum driver */ 3723 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3724 if (IS_ERR(sbi->s_chksum_driver)) { 3725 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3726 ret = PTR_ERR(sbi->s_chksum_driver); 3727 sbi->s_chksum_driver = NULL; 3728 goto failed_mount; 3729 } 3730 3731 /* Check superblock checksum */ 3732 if (!ext4_superblock_csum_verify(sb, es)) { 3733 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3734 "invalid superblock checksum. Run e2fsck?"); 3735 silent = 1; 3736 ret = -EFSBADCRC; 3737 goto cantfind_ext4; 3738 } 3739 3740 /* Precompute checksum seed for all metadata */ 3741 if (ext4_has_feature_csum_seed(sb)) 3742 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3743 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3744 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3745 sizeof(es->s_uuid)); 3746 3747 /* Set defaults before we parse the mount options */ 3748 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3749 set_opt(sb, INIT_INODE_TABLE); 3750 if (def_mount_opts & EXT4_DEFM_DEBUG) 3751 set_opt(sb, DEBUG); 3752 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3753 set_opt(sb, GRPID); 3754 if (def_mount_opts & EXT4_DEFM_UID16) 3755 set_opt(sb, NO_UID32); 3756 /* xattr user namespace & acls are now defaulted on */ 3757 set_opt(sb, XATTR_USER); 3758 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3759 set_opt(sb, POSIX_ACL); 3760 #endif 3761 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 3762 if (ext4_has_metadata_csum(sb)) 3763 set_opt(sb, JOURNAL_CHECKSUM); 3764 3765 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3766 set_opt(sb, JOURNAL_DATA); 3767 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3768 set_opt(sb, ORDERED_DATA); 3769 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3770 set_opt(sb, WRITEBACK_DATA); 3771 3772 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3773 set_opt(sb, ERRORS_PANIC); 3774 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3775 set_opt(sb, ERRORS_CONT); 3776 else 3777 set_opt(sb, ERRORS_RO); 3778 /* block_validity enabled by default; disable with noblock_validity */ 3779 set_opt(sb, BLOCK_VALIDITY); 3780 if (def_mount_opts & EXT4_DEFM_DISCARD) 3781 set_opt(sb, DISCARD); 3782 3783 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3784 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3785 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3786 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3787 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3788 3789 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3790 set_opt(sb, BARRIER); 3791 3792 /* 3793 * enable delayed allocation by default 3794 * Use -o nodelalloc to turn it off 3795 */ 3796 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3797 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3798 set_opt(sb, DELALLOC); 3799 3800 /* 3801 * set default s_li_wait_mult for lazyinit, for the case there is 3802 * no mount option specified. 3803 */ 3804 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3805 3806 if (sbi->s_es->s_mount_opts[0]) { 3807 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 3808 sizeof(sbi->s_es->s_mount_opts), 3809 GFP_KERNEL); 3810 if (!s_mount_opts) 3811 goto failed_mount; 3812 if (!parse_options(s_mount_opts, sb, &journal_devnum, 3813 &journal_ioprio, 0)) { 3814 ext4_msg(sb, KERN_WARNING, 3815 "failed to parse options in superblock: %s", 3816 s_mount_opts); 3817 } 3818 kfree(s_mount_opts); 3819 } 3820 sbi->s_def_mount_opt = sbi->s_mount_opt; 3821 if (!parse_options((char *) data, sb, &journal_devnum, 3822 &journal_ioprio, 0)) 3823 goto failed_mount; 3824 3825 #ifdef CONFIG_UNICODE 3826 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 3827 const struct ext4_sb_encodings *encoding_info; 3828 struct unicode_map *encoding; 3829 __u16 encoding_flags; 3830 3831 if (ext4_has_feature_encrypt(sb)) { 3832 ext4_msg(sb, KERN_ERR, 3833 "Can't mount with encoding and encryption"); 3834 goto failed_mount; 3835 } 3836 3837 if (ext4_sb_read_encoding(es, &encoding_info, 3838 &encoding_flags)) { 3839 ext4_msg(sb, KERN_ERR, 3840 "Encoding requested by superblock is unknown"); 3841 goto failed_mount; 3842 } 3843 3844 encoding = utf8_load(encoding_info->version); 3845 if (IS_ERR(encoding)) { 3846 ext4_msg(sb, KERN_ERR, 3847 "can't mount with superblock charset: %s-%s " 3848 "not supported by the kernel. flags: 0x%x.", 3849 encoding_info->name, encoding_info->version, 3850 encoding_flags); 3851 goto failed_mount; 3852 } 3853 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 3854 "%s-%s with flags 0x%hx", encoding_info->name, 3855 encoding_info->version?:"\b", encoding_flags); 3856 3857 sbi->s_encoding = encoding; 3858 sbi->s_encoding_flags = encoding_flags; 3859 } 3860 #endif 3861 3862 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3863 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3864 "with data=journal disables delayed " 3865 "allocation and O_DIRECT support!\n"); 3866 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3867 ext4_msg(sb, KERN_ERR, "can't mount with " 3868 "both data=journal and delalloc"); 3869 goto failed_mount; 3870 } 3871 if (test_opt(sb, DIOREAD_NOLOCK)) { 3872 ext4_msg(sb, KERN_ERR, "can't mount with " 3873 "both data=journal and dioread_nolock"); 3874 goto failed_mount; 3875 } 3876 if (test_opt(sb, DAX)) { 3877 ext4_msg(sb, KERN_ERR, "can't mount with " 3878 "both data=journal and dax"); 3879 goto failed_mount; 3880 } 3881 if (ext4_has_feature_encrypt(sb)) { 3882 ext4_msg(sb, KERN_WARNING, 3883 "encrypted files will use data=ordered " 3884 "instead of data journaling mode"); 3885 } 3886 if (test_opt(sb, DELALLOC)) 3887 clear_opt(sb, DELALLOC); 3888 } else { 3889 sb->s_iflags |= SB_I_CGROUPWB; 3890 } 3891 3892 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3893 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 3894 3895 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3896 (ext4_has_compat_features(sb) || 3897 ext4_has_ro_compat_features(sb) || 3898 ext4_has_incompat_features(sb))) 3899 ext4_msg(sb, KERN_WARNING, 3900 "feature flags set on rev 0 fs, " 3901 "running e2fsck is recommended"); 3902 3903 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3904 set_opt2(sb, HURD_COMPAT); 3905 if (ext4_has_feature_64bit(sb)) { 3906 ext4_msg(sb, KERN_ERR, 3907 "The Hurd can't support 64-bit file systems"); 3908 goto failed_mount; 3909 } 3910 3911 /* 3912 * ea_inode feature uses l_i_version field which is not 3913 * available in HURD_COMPAT mode. 3914 */ 3915 if (ext4_has_feature_ea_inode(sb)) { 3916 ext4_msg(sb, KERN_ERR, 3917 "ea_inode feature is not supported for Hurd"); 3918 goto failed_mount; 3919 } 3920 } 3921 3922 if (IS_EXT2_SB(sb)) { 3923 if (ext2_feature_set_ok(sb)) 3924 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3925 "using the ext4 subsystem"); 3926 else { 3927 /* 3928 * If we're probing be silent, if this looks like 3929 * it's actually an ext[34] filesystem. 3930 */ 3931 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3932 goto failed_mount; 3933 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3934 "to feature incompatibilities"); 3935 goto failed_mount; 3936 } 3937 } 3938 3939 if (IS_EXT3_SB(sb)) { 3940 if (ext3_feature_set_ok(sb)) 3941 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3942 "using the ext4 subsystem"); 3943 else { 3944 /* 3945 * If we're probing be silent, if this looks like 3946 * it's actually an ext4 filesystem. 3947 */ 3948 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3949 goto failed_mount; 3950 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3951 "to feature incompatibilities"); 3952 goto failed_mount; 3953 } 3954 } 3955 3956 /* 3957 * Check feature flags regardless of the revision level, since we 3958 * previously didn't change the revision level when setting the flags, 3959 * so there is a chance incompat flags are set on a rev 0 filesystem. 3960 */ 3961 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 3962 goto failed_mount; 3963 3964 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3965 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3966 blocksize > EXT4_MAX_BLOCK_SIZE) { 3967 ext4_msg(sb, KERN_ERR, 3968 "Unsupported filesystem blocksize %d (%d log_block_size)", 3969 blocksize, le32_to_cpu(es->s_log_block_size)); 3970 goto failed_mount; 3971 } 3972 if (le32_to_cpu(es->s_log_block_size) > 3973 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3974 ext4_msg(sb, KERN_ERR, 3975 "Invalid log block size: %u", 3976 le32_to_cpu(es->s_log_block_size)); 3977 goto failed_mount; 3978 } 3979 if (le32_to_cpu(es->s_log_cluster_size) > 3980 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3981 ext4_msg(sb, KERN_ERR, 3982 "Invalid log cluster size: %u", 3983 le32_to_cpu(es->s_log_cluster_size)); 3984 goto failed_mount; 3985 } 3986 3987 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 3988 ext4_msg(sb, KERN_ERR, 3989 "Number of reserved GDT blocks insanely large: %d", 3990 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 3991 goto failed_mount; 3992 } 3993 3994 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 3995 if (ext4_has_feature_inline_data(sb)) { 3996 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 3997 " that may contain inline data"); 3998 goto failed_mount; 3999 } 4000 if (!bdev_dax_supported(sb->s_bdev, blocksize)) { 4001 ext4_msg(sb, KERN_ERR, 4002 "DAX unsupported by block device."); 4003 goto failed_mount; 4004 } 4005 } 4006 4007 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4008 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4009 es->s_encryption_level); 4010 goto failed_mount; 4011 } 4012 4013 if (sb->s_blocksize != blocksize) { 4014 /* Validate the filesystem blocksize */ 4015 if (!sb_set_blocksize(sb, blocksize)) { 4016 ext4_msg(sb, KERN_ERR, "bad block size %d", 4017 blocksize); 4018 goto failed_mount; 4019 } 4020 4021 brelse(bh); 4022 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 4023 offset = do_div(logical_sb_block, blocksize); 4024 bh = sb_bread_unmovable(sb, logical_sb_block); 4025 if (!bh) { 4026 ext4_msg(sb, KERN_ERR, 4027 "Can't read superblock on 2nd try"); 4028 goto failed_mount; 4029 } 4030 es = (struct ext4_super_block *)(bh->b_data + offset); 4031 sbi->s_es = es; 4032 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4033 ext4_msg(sb, KERN_ERR, 4034 "Magic mismatch, very weird!"); 4035 goto failed_mount; 4036 } 4037 } 4038 4039 has_huge_files = ext4_has_feature_huge_file(sb); 4040 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4041 has_huge_files); 4042 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4043 4044 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4045 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4046 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4047 } else { 4048 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4049 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4050 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4051 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4052 sbi->s_first_ino); 4053 goto failed_mount; 4054 } 4055 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4056 (!is_power_of_2(sbi->s_inode_size)) || 4057 (sbi->s_inode_size > blocksize)) { 4058 ext4_msg(sb, KERN_ERR, 4059 "unsupported inode size: %d", 4060 sbi->s_inode_size); 4061 goto failed_mount; 4062 } 4063 /* 4064 * i_atime_extra is the last extra field available for [acm]times in 4065 * struct ext4_inode. Checking for that field should suffice to ensure 4066 * we have extra space for all three. 4067 */ 4068 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4069 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4070 sb->s_time_gran = 1; 4071 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4072 } else { 4073 sb->s_time_gran = NSEC_PER_SEC; 4074 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4075 } 4076 4077 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4078 } 4079 4080 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4081 if (ext4_has_feature_64bit(sb)) { 4082 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4083 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4084 !is_power_of_2(sbi->s_desc_size)) { 4085 ext4_msg(sb, KERN_ERR, 4086 "unsupported descriptor size %lu", 4087 sbi->s_desc_size); 4088 goto failed_mount; 4089 } 4090 } else 4091 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4092 4093 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4094 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4095 4096 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4097 if (sbi->s_inodes_per_block == 0) 4098 goto cantfind_ext4; 4099 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4100 sbi->s_inodes_per_group > blocksize * 8) { 4101 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4102 sbi->s_blocks_per_group); 4103 goto failed_mount; 4104 } 4105 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4106 sbi->s_inodes_per_block; 4107 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4108 sbi->s_sbh = bh; 4109 sbi->s_mount_state = le16_to_cpu(es->s_state); 4110 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4111 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4112 4113 for (i = 0; i < 4; i++) 4114 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4115 sbi->s_def_hash_version = es->s_def_hash_version; 4116 if (ext4_has_feature_dir_index(sb)) { 4117 i = le32_to_cpu(es->s_flags); 4118 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4119 sbi->s_hash_unsigned = 3; 4120 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4121 #ifdef __CHAR_UNSIGNED__ 4122 if (!sb_rdonly(sb)) 4123 es->s_flags |= 4124 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4125 sbi->s_hash_unsigned = 3; 4126 #else 4127 if (!sb_rdonly(sb)) 4128 es->s_flags |= 4129 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4130 #endif 4131 } 4132 } 4133 4134 /* Handle clustersize */ 4135 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4136 has_bigalloc = ext4_has_feature_bigalloc(sb); 4137 if (has_bigalloc) { 4138 if (clustersize < blocksize) { 4139 ext4_msg(sb, KERN_ERR, 4140 "cluster size (%d) smaller than " 4141 "block size (%d)", clustersize, blocksize); 4142 goto failed_mount; 4143 } 4144 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4145 le32_to_cpu(es->s_log_block_size); 4146 sbi->s_clusters_per_group = 4147 le32_to_cpu(es->s_clusters_per_group); 4148 if (sbi->s_clusters_per_group > blocksize * 8) { 4149 ext4_msg(sb, KERN_ERR, 4150 "#clusters per group too big: %lu", 4151 sbi->s_clusters_per_group); 4152 goto failed_mount; 4153 } 4154 if (sbi->s_blocks_per_group != 4155 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4156 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4157 "clusters per group (%lu) inconsistent", 4158 sbi->s_blocks_per_group, 4159 sbi->s_clusters_per_group); 4160 goto failed_mount; 4161 } 4162 } else { 4163 if (clustersize != blocksize) { 4164 ext4_msg(sb, KERN_ERR, 4165 "fragment/cluster size (%d) != " 4166 "block size (%d)", clustersize, blocksize); 4167 goto failed_mount; 4168 } 4169 if (sbi->s_blocks_per_group > blocksize * 8) { 4170 ext4_msg(sb, KERN_ERR, 4171 "#blocks per group too big: %lu", 4172 sbi->s_blocks_per_group); 4173 goto failed_mount; 4174 } 4175 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4176 sbi->s_cluster_bits = 0; 4177 } 4178 sbi->s_cluster_ratio = clustersize / blocksize; 4179 4180 /* Do we have standard group size of clustersize * 8 blocks ? */ 4181 if (sbi->s_blocks_per_group == clustersize << 3) 4182 set_opt2(sb, STD_GROUP_SIZE); 4183 4184 /* 4185 * Test whether we have more sectors than will fit in sector_t, 4186 * and whether the max offset is addressable by the page cache. 4187 */ 4188 err = generic_check_addressable(sb->s_blocksize_bits, 4189 ext4_blocks_count(es)); 4190 if (err) { 4191 ext4_msg(sb, KERN_ERR, "filesystem" 4192 " too large to mount safely on this system"); 4193 goto failed_mount; 4194 } 4195 4196 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4197 goto cantfind_ext4; 4198 4199 /* check blocks count against device size */ 4200 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 4201 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4202 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4203 "exceeds size of device (%llu blocks)", 4204 ext4_blocks_count(es), blocks_count); 4205 goto failed_mount; 4206 } 4207 4208 /* 4209 * It makes no sense for the first data block to be beyond the end 4210 * of the filesystem. 4211 */ 4212 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4213 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4214 "block %u is beyond end of filesystem (%llu)", 4215 le32_to_cpu(es->s_first_data_block), 4216 ext4_blocks_count(es)); 4217 goto failed_mount; 4218 } 4219 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4220 (sbi->s_cluster_ratio == 1)) { 4221 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4222 "block is 0 with a 1k block and cluster size"); 4223 goto failed_mount; 4224 } 4225 4226 blocks_count = (ext4_blocks_count(es) - 4227 le32_to_cpu(es->s_first_data_block) + 4228 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4229 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4230 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4231 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 4232 "(block count %llu, first data block %u, " 4233 "blocks per group %lu)", sbi->s_groups_count, 4234 ext4_blocks_count(es), 4235 le32_to_cpu(es->s_first_data_block), 4236 EXT4_BLOCKS_PER_GROUP(sb)); 4237 goto failed_mount; 4238 } 4239 sbi->s_groups_count = blocks_count; 4240 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4241 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4242 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4243 le32_to_cpu(es->s_inodes_count)) { 4244 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4245 le32_to_cpu(es->s_inodes_count), 4246 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4247 ret = -EINVAL; 4248 goto failed_mount; 4249 } 4250 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4251 EXT4_DESC_PER_BLOCK(sb); 4252 if (ext4_has_feature_meta_bg(sb)) { 4253 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4254 ext4_msg(sb, KERN_WARNING, 4255 "first meta block group too large: %u " 4256 "(group descriptor block count %u)", 4257 le32_to_cpu(es->s_first_meta_bg), db_count); 4258 goto failed_mount; 4259 } 4260 } 4261 sbi->s_group_desc = kvmalloc_array(db_count, 4262 sizeof(struct buffer_head *), 4263 GFP_KERNEL); 4264 if (sbi->s_group_desc == NULL) { 4265 ext4_msg(sb, KERN_ERR, "not enough memory"); 4266 ret = -ENOMEM; 4267 goto failed_mount; 4268 } 4269 4270 bgl_lock_init(sbi->s_blockgroup_lock); 4271 4272 /* Pre-read the descriptors into the buffer cache */ 4273 for (i = 0; i < db_count; i++) { 4274 block = descriptor_loc(sb, logical_sb_block, i); 4275 sb_breadahead(sb, block); 4276 } 4277 4278 for (i = 0; i < db_count; i++) { 4279 block = descriptor_loc(sb, logical_sb_block, i); 4280 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 4281 if (!sbi->s_group_desc[i]) { 4282 ext4_msg(sb, KERN_ERR, 4283 "can't read group descriptor %d", i); 4284 db_count = i; 4285 goto failed_mount2; 4286 } 4287 } 4288 sbi->s_gdb_count = db_count; 4289 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4290 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4291 ret = -EFSCORRUPTED; 4292 goto failed_mount2; 4293 } 4294 4295 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4296 4297 /* Register extent status tree shrinker */ 4298 if (ext4_es_register_shrinker(sbi)) 4299 goto failed_mount3; 4300 4301 sbi->s_stripe = ext4_get_stripe_size(sbi); 4302 sbi->s_extent_max_zeroout_kb = 32; 4303 4304 /* 4305 * set up enough so that it can read an inode 4306 */ 4307 sb->s_op = &ext4_sops; 4308 sb->s_export_op = &ext4_export_ops; 4309 sb->s_xattr = ext4_xattr_handlers; 4310 #ifdef CONFIG_FS_ENCRYPTION 4311 sb->s_cop = &ext4_cryptops; 4312 #endif 4313 #ifdef CONFIG_FS_VERITY 4314 sb->s_vop = &ext4_verityops; 4315 #endif 4316 #ifdef CONFIG_QUOTA 4317 sb->dq_op = &ext4_quota_operations; 4318 if (ext4_has_feature_quota(sb)) 4319 sb->s_qcop = &dquot_quotactl_sysfile_ops; 4320 else 4321 sb->s_qcop = &ext4_qctl_operations; 4322 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4323 #endif 4324 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4325 4326 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 4327 mutex_init(&sbi->s_orphan_lock); 4328 4329 sb->s_root = NULL; 4330 4331 needs_recovery = (es->s_last_orphan != 0 || 4332 ext4_has_feature_journal_needs_recovery(sb)); 4333 4334 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4335 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 4336 goto failed_mount3a; 4337 4338 /* 4339 * The first inode we look at is the journal inode. Don't try 4340 * root first: it may be modified in the journal! 4341 */ 4342 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 4343 err = ext4_load_journal(sb, es, journal_devnum); 4344 if (err) 4345 goto failed_mount3a; 4346 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4347 ext4_has_feature_journal_needs_recovery(sb)) { 4348 ext4_msg(sb, KERN_ERR, "required journal recovery " 4349 "suppressed and not mounted read-only"); 4350 goto failed_mount_wq; 4351 } else { 4352 /* Nojournal mode, all journal mount options are illegal */ 4353 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 4354 ext4_msg(sb, KERN_ERR, "can't mount with " 4355 "journal_checksum, fs mounted w/o journal"); 4356 goto failed_mount_wq; 4357 } 4358 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4359 ext4_msg(sb, KERN_ERR, "can't mount with " 4360 "journal_async_commit, fs mounted w/o journal"); 4361 goto failed_mount_wq; 4362 } 4363 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 4364 ext4_msg(sb, KERN_ERR, "can't mount with " 4365 "commit=%lu, fs mounted w/o journal", 4366 sbi->s_commit_interval / HZ); 4367 goto failed_mount_wq; 4368 } 4369 if (EXT4_MOUNT_DATA_FLAGS & 4370 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 4371 ext4_msg(sb, KERN_ERR, "can't mount with " 4372 "data=, fs mounted w/o journal"); 4373 goto failed_mount_wq; 4374 } 4375 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 4376 clear_opt(sb, JOURNAL_CHECKSUM); 4377 clear_opt(sb, DATA_FLAGS); 4378 sbi->s_journal = NULL; 4379 needs_recovery = 0; 4380 goto no_journal; 4381 } 4382 4383 if (ext4_has_feature_64bit(sb) && 4384 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4385 JBD2_FEATURE_INCOMPAT_64BIT)) { 4386 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4387 goto failed_mount_wq; 4388 } 4389 4390 if (!set_journal_csum_feature_set(sb)) { 4391 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4392 "feature set"); 4393 goto failed_mount_wq; 4394 } 4395 4396 /* We have now updated the journal if required, so we can 4397 * validate the data journaling mode. */ 4398 switch (test_opt(sb, DATA_FLAGS)) { 4399 case 0: 4400 /* No mode set, assume a default based on the journal 4401 * capabilities: ORDERED_DATA if the journal can 4402 * cope, else JOURNAL_DATA 4403 */ 4404 if (jbd2_journal_check_available_features 4405 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4406 set_opt(sb, ORDERED_DATA); 4407 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4408 } else { 4409 set_opt(sb, JOURNAL_DATA); 4410 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4411 } 4412 break; 4413 4414 case EXT4_MOUNT_ORDERED_DATA: 4415 case EXT4_MOUNT_WRITEBACK_DATA: 4416 if (!jbd2_journal_check_available_features 4417 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4418 ext4_msg(sb, KERN_ERR, "Journal does not support " 4419 "requested data journaling mode"); 4420 goto failed_mount_wq; 4421 } 4422 default: 4423 break; 4424 } 4425 4426 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4427 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4428 ext4_msg(sb, KERN_ERR, "can't mount with " 4429 "journal_async_commit in data=ordered mode"); 4430 goto failed_mount_wq; 4431 } 4432 4433 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4434 4435 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4436 4437 no_journal: 4438 if (!test_opt(sb, NO_MBCACHE)) { 4439 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 4440 if (!sbi->s_ea_block_cache) { 4441 ext4_msg(sb, KERN_ERR, 4442 "Failed to create ea_block_cache"); 4443 goto failed_mount_wq; 4444 } 4445 4446 if (ext4_has_feature_ea_inode(sb)) { 4447 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4448 if (!sbi->s_ea_inode_cache) { 4449 ext4_msg(sb, KERN_ERR, 4450 "Failed to create ea_inode_cache"); 4451 goto failed_mount_wq; 4452 } 4453 } 4454 } 4455 4456 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && 4457 (blocksize != PAGE_SIZE)) { 4458 ext4_msg(sb, KERN_ERR, 4459 "Unsupported blocksize for fs encryption"); 4460 goto failed_mount_wq; 4461 } 4462 4463 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4464 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4465 goto failed_mount_wq; 4466 } 4467 4468 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4469 !ext4_has_feature_encrypt(sb)) { 4470 ext4_set_feature_encrypt(sb); 4471 ext4_commit_super(sb, 1); 4472 } 4473 4474 /* 4475 * Get the # of file system overhead blocks from the 4476 * superblock if present. 4477 */ 4478 if (es->s_overhead_clusters) 4479 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4480 else { 4481 err = ext4_calculate_overhead(sb); 4482 if (err) 4483 goto failed_mount_wq; 4484 } 4485 4486 /* 4487 * The maximum number of concurrent works can be high and 4488 * concurrency isn't really necessary. Limit it to 1. 4489 */ 4490 EXT4_SB(sb)->rsv_conversion_wq = 4491 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4492 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4493 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4494 ret = -ENOMEM; 4495 goto failed_mount4; 4496 } 4497 4498 /* 4499 * The jbd2_journal_load will have done any necessary log recovery, 4500 * so we can safely mount the rest of the filesystem now. 4501 */ 4502 4503 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 4504 if (IS_ERR(root)) { 4505 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4506 ret = PTR_ERR(root); 4507 root = NULL; 4508 goto failed_mount4; 4509 } 4510 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4511 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4512 iput(root); 4513 goto failed_mount4; 4514 } 4515 4516 #ifdef CONFIG_UNICODE 4517 if (sbi->s_encoding) 4518 sb->s_d_op = &ext4_dentry_ops; 4519 #endif 4520 4521 sb->s_root = d_make_root(root); 4522 if (!sb->s_root) { 4523 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4524 ret = -ENOMEM; 4525 goto failed_mount4; 4526 } 4527 4528 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4529 if (ret == -EROFS) { 4530 sb->s_flags |= SB_RDONLY; 4531 ret = 0; 4532 } else if (ret) 4533 goto failed_mount4a; 4534 4535 ext4_clamp_want_extra_isize(sb); 4536 4537 ext4_set_resv_clusters(sb); 4538 4539 err = ext4_setup_system_zone(sb); 4540 if (err) { 4541 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4542 "zone (%d)", err); 4543 goto failed_mount4a; 4544 } 4545 4546 ext4_ext_init(sb); 4547 err = ext4_mb_init(sb); 4548 if (err) { 4549 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4550 err); 4551 goto failed_mount5; 4552 } 4553 4554 block = ext4_count_free_clusters(sb); 4555 ext4_free_blocks_count_set(sbi->s_es, 4556 EXT4_C2B(sbi, block)); 4557 ext4_superblock_csum_set(sb); 4558 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4559 GFP_KERNEL); 4560 if (!err) { 4561 unsigned long freei = ext4_count_free_inodes(sb); 4562 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4563 ext4_superblock_csum_set(sb); 4564 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4565 GFP_KERNEL); 4566 } 4567 if (!err) 4568 err = percpu_counter_init(&sbi->s_dirs_counter, 4569 ext4_count_dirs(sb), GFP_KERNEL); 4570 if (!err) 4571 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4572 GFP_KERNEL); 4573 if (!err) 4574 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 4575 4576 if (err) { 4577 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4578 goto failed_mount6; 4579 } 4580 4581 if (ext4_has_feature_flex_bg(sb)) 4582 if (!ext4_fill_flex_info(sb)) { 4583 ext4_msg(sb, KERN_ERR, 4584 "unable to initialize " 4585 "flex_bg meta info!"); 4586 goto failed_mount6; 4587 } 4588 4589 err = ext4_register_li_request(sb, first_not_zeroed); 4590 if (err) 4591 goto failed_mount6; 4592 4593 err = ext4_register_sysfs(sb); 4594 if (err) 4595 goto failed_mount7; 4596 4597 #ifdef CONFIG_QUOTA 4598 /* Enable quota usage during mount. */ 4599 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 4600 err = ext4_enable_quotas(sb); 4601 if (err) 4602 goto failed_mount8; 4603 } 4604 #endif /* CONFIG_QUOTA */ 4605 4606 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4607 ext4_orphan_cleanup(sb, es); 4608 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4609 if (needs_recovery) { 4610 ext4_msg(sb, KERN_INFO, "recovery complete"); 4611 ext4_mark_recovery_complete(sb, es); 4612 } 4613 if (EXT4_SB(sb)->s_journal) { 4614 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4615 descr = " journalled data mode"; 4616 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4617 descr = " ordered data mode"; 4618 else 4619 descr = " writeback data mode"; 4620 } else 4621 descr = "out journal"; 4622 4623 if (test_opt(sb, DISCARD)) { 4624 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4625 if (!blk_queue_discard(q)) 4626 ext4_msg(sb, KERN_WARNING, 4627 "mounting with \"discard\" option, but " 4628 "the device does not support discard"); 4629 } 4630 4631 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4632 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4633 "Opts: %.*s%s%s", descr, 4634 (int) sizeof(sbi->s_es->s_mount_opts), 4635 sbi->s_es->s_mount_opts, 4636 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4637 4638 if (es->s_error_count) 4639 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4640 4641 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4642 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4643 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4644 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4645 4646 kfree(orig_data); 4647 return 0; 4648 4649 cantfind_ext4: 4650 if (!silent) 4651 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4652 goto failed_mount; 4653 4654 #ifdef CONFIG_QUOTA 4655 failed_mount8: 4656 ext4_unregister_sysfs(sb); 4657 #endif 4658 failed_mount7: 4659 ext4_unregister_li_request(sb); 4660 failed_mount6: 4661 ext4_mb_release(sb); 4662 if (sbi->s_flex_groups) 4663 kvfree(sbi->s_flex_groups); 4664 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4665 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4666 percpu_counter_destroy(&sbi->s_dirs_counter); 4667 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4668 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 4669 failed_mount5: 4670 ext4_ext_release(sb); 4671 ext4_release_system_zone(sb); 4672 failed_mount4a: 4673 dput(sb->s_root); 4674 sb->s_root = NULL; 4675 failed_mount4: 4676 ext4_msg(sb, KERN_ERR, "mount failed"); 4677 if (EXT4_SB(sb)->rsv_conversion_wq) 4678 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4679 failed_mount_wq: 4680 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4681 sbi->s_ea_inode_cache = NULL; 4682 4683 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 4684 sbi->s_ea_block_cache = NULL; 4685 4686 if (sbi->s_journal) { 4687 jbd2_journal_destroy(sbi->s_journal); 4688 sbi->s_journal = NULL; 4689 } 4690 failed_mount3a: 4691 ext4_es_unregister_shrinker(sbi); 4692 failed_mount3: 4693 del_timer_sync(&sbi->s_err_report); 4694 if (sbi->s_mmp_tsk) 4695 kthread_stop(sbi->s_mmp_tsk); 4696 failed_mount2: 4697 for (i = 0; i < db_count; i++) 4698 brelse(sbi->s_group_desc[i]); 4699 kvfree(sbi->s_group_desc); 4700 failed_mount: 4701 if (sbi->s_chksum_driver) 4702 crypto_free_shash(sbi->s_chksum_driver); 4703 4704 #ifdef CONFIG_UNICODE 4705 utf8_unload(sbi->s_encoding); 4706 #endif 4707 4708 #ifdef CONFIG_QUOTA 4709 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4710 kfree(get_qf_name(sb, sbi, i)); 4711 #endif 4712 ext4_blkdev_remove(sbi); 4713 brelse(bh); 4714 out_fail: 4715 sb->s_fs_info = NULL; 4716 kfree(sbi->s_blockgroup_lock); 4717 out_free_base: 4718 kfree(sbi); 4719 kfree(orig_data); 4720 fs_put_dax(dax_dev); 4721 return err ? err : ret; 4722 } 4723 4724 /* 4725 * Setup any per-fs journal parameters now. We'll do this both on 4726 * initial mount, once the journal has been initialised but before we've 4727 * done any recovery; and again on any subsequent remount. 4728 */ 4729 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4730 { 4731 struct ext4_sb_info *sbi = EXT4_SB(sb); 4732 4733 journal->j_commit_interval = sbi->s_commit_interval; 4734 journal->j_min_batch_time = sbi->s_min_batch_time; 4735 journal->j_max_batch_time = sbi->s_max_batch_time; 4736 4737 write_lock(&journal->j_state_lock); 4738 if (test_opt(sb, BARRIER)) 4739 journal->j_flags |= JBD2_BARRIER; 4740 else 4741 journal->j_flags &= ~JBD2_BARRIER; 4742 if (test_opt(sb, DATA_ERR_ABORT)) 4743 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4744 else 4745 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4746 write_unlock(&journal->j_state_lock); 4747 } 4748 4749 static struct inode *ext4_get_journal_inode(struct super_block *sb, 4750 unsigned int journal_inum) 4751 { 4752 struct inode *journal_inode; 4753 4754 /* 4755 * Test for the existence of a valid inode on disk. Bad things 4756 * happen if we iget() an unused inode, as the subsequent iput() 4757 * will try to delete it. 4758 */ 4759 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 4760 if (IS_ERR(journal_inode)) { 4761 ext4_msg(sb, KERN_ERR, "no journal found"); 4762 return NULL; 4763 } 4764 if (!journal_inode->i_nlink) { 4765 make_bad_inode(journal_inode); 4766 iput(journal_inode); 4767 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4768 return NULL; 4769 } 4770 4771 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4772 journal_inode, journal_inode->i_size); 4773 if (!S_ISREG(journal_inode->i_mode)) { 4774 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4775 iput(journal_inode); 4776 return NULL; 4777 } 4778 return journal_inode; 4779 } 4780 4781 static journal_t *ext4_get_journal(struct super_block *sb, 4782 unsigned int journal_inum) 4783 { 4784 struct inode *journal_inode; 4785 journal_t *journal; 4786 4787 BUG_ON(!ext4_has_feature_journal(sb)); 4788 4789 journal_inode = ext4_get_journal_inode(sb, journal_inum); 4790 if (!journal_inode) 4791 return NULL; 4792 4793 journal = jbd2_journal_init_inode(journal_inode); 4794 if (!journal) { 4795 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4796 iput(journal_inode); 4797 return NULL; 4798 } 4799 journal->j_private = sb; 4800 ext4_init_journal_params(sb, journal); 4801 return journal; 4802 } 4803 4804 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4805 dev_t j_dev) 4806 { 4807 struct buffer_head *bh; 4808 journal_t *journal; 4809 ext4_fsblk_t start; 4810 ext4_fsblk_t len; 4811 int hblock, blocksize; 4812 ext4_fsblk_t sb_block; 4813 unsigned long offset; 4814 struct ext4_super_block *es; 4815 struct block_device *bdev; 4816 4817 BUG_ON(!ext4_has_feature_journal(sb)); 4818 4819 bdev = ext4_blkdev_get(j_dev, sb); 4820 if (bdev == NULL) 4821 return NULL; 4822 4823 blocksize = sb->s_blocksize; 4824 hblock = bdev_logical_block_size(bdev); 4825 if (blocksize < hblock) { 4826 ext4_msg(sb, KERN_ERR, 4827 "blocksize too small for journal device"); 4828 goto out_bdev; 4829 } 4830 4831 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4832 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4833 set_blocksize(bdev, blocksize); 4834 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4835 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4836 "external journal"); 4837 goto out_bdev; 4838 } 4839 4840 es = (struct ext4_super_block *) (bh->b_data + offset); 4841 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4842 !(le32_to_cpu(es->s_feature_incompat) & 4843 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4844 ext4_msg(sb, KERN_ERR, "external journal has " 4845 "bad superblock"); 4846 brelse(bh); 4847 goto out_bdev; 4848 } 4849 4850 if ((le32_to_cpu(es->s_feature_ro_compat) & 4851 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4852 es->s_checksum != ext4_superblock_csum(sb, es)) { 4853 ext4_msg(sb, KERN_ERR, "external journal has " 4854 "corrupt superblock"); 4855 brelse(bh); 4856 goto out_bdev; 4857 } 4858 4859 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4860 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4861 brelse(bh); 4862 goto out_bdev; 4863 } 4864 4865 len = ext4_blocks_count(es); 4866 start = sb_block + 1; 4867 brelse(bh); /* we're done with the superblock */ 4868 4869 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4870 start, len, blocksize); 4871 if (!journal) { 4872 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4873 goto out_bdev; 4874 } 4875 journal->j_private = sb; 4876 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4877 wait_on_buffer(journal->j_sb_buffer); 4878 if (!buffer_uptodate(journal->j_sb_buffer)) { 4879 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4880 goto out_journal; 4881 } 4882 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4883 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4884 "user (unsupported) - %d", 4885 be32_to_cpu(journal->j_superblock->s_nr_users)); 4886 goto out_journal; 4887 } 4888 EXT4_SB(sb)->journal_bdev = bdev; 4889 ext4_init_journal_params(sb, journal); 4890 return journal; 4891 4892 out_journal: 4893 jbd2_journal_destroy(journal); 4894 out_bdev: 4895 ext4_blkdev_put(bdev); 4896 return NULL; 4897 } 4898 4899 static int ext4_load_journal(struct super_block *sb, 4900 struct ext4_super_block *es, 4901 unsigned long journal_devnum) 4902 { 4903 journal_t *journal; 4904 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4905 dev_t journal_dev; 4906 int err = 0; 4907 int really_read_only; 4908 4909 BUG_ON(!ext4_has_feature_journal(sb)); 4910 4911 if (journal_devnum && 4912 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4913 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4914 "numbers have changed"); 4915 journal_dev = new_decode_dev(journal_devnum); 4916 } else 4917 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4918 4919 really_read_only = bdev_read_only(sb->s_bdev); 4920 4921 /* 4922 * Are we loading a blank journal or performing recovery after a 4923 * crash? For recovery, we need to check in advance whether we 4924 * can get read-write access to the device. 4925 */ 4926 if (ext4_has_feature_journal_needs_recovery(sb)) { 4927 if (sb_rdonly(sb)) { 4928 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4929 "required on readonly filesystem"); 4930 if (really_read_only) { 4931 ext4_msg(sb, KERN_ERR, "write access " 4932 "unavailable, cannot proceed " 4933 "(try mounting with noload)"); 4934 return -EROFS; 4935 } 4936 ext4_msg(sb, KERN_INFO, "write access will " 4937 "be enabled during recovery"); 4938 } 4939 } 4940 4941 if (journal_inum && journal_dev) { 4942 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4943 "and inode journals!"); 4944 return -EINVAL; 4945 } 4946 4947 if (journal_inum) { 4948 if (!(journal = ext4_get_journal(sb, journal_inum))) 4949 return -EINVAL; 4950 } else { 4951 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4952 return -EINVAL; 4953 } 4954 4955 if (!(journal->j_flags & JBD2_BARRIER)) 4956 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4957 4958 if (!ext4_has_feature_journal_needs_recovery(sb)) 4959 err = jbd2_journal_wipe(journal, !really_read_only); 4960 if (!err) { 4961 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4962 if (save) 4963 memcpy(save, ((char *) es) + 4964 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4965 err = jbd2_journal_load(journal); 4966 if (save) 4967 memcpy(((char *) es) + EXT4_S_ERR_START, 4968 save, EXT4_S_ERR_LEN); 4969 kfree(save); 4970 } 4971 4972 if (err) { 4973 ext4_msg(sb, KERN_ERR, "error loading journal"); 4974 jbd2_journal_destroy(journal); 4975 return err; 4976 } 4977 4978 EXT4_SB(sb)->s_journal = journal; 4979 ext4_clear_journal_err(sb, es); 4980 4981 if (!really_read_only && journal_devnum && 4982 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4983 es->s_journal_dev = cpu_to_le32(journal_devnum); 4984 4985 /* Make sure we flush the recovery flag to disk. */ 4986 ext4_commit_super(sb, 1); 4987 } 4988 4989 return 0; 4990 } 4991 4992 static int ext4_commit_super(struct super_block *sb, int sync) 4993 { 4994 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4995 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4996 int error = 0; 4997 4998 if (!sbh || block_device_ejected(sb)) 4999 return error; 5000 5001 /* 5002 * The superblock bh should be mapped, but it might not be if the 5003 * device was hot-removed. Not much we can do but fail the I/O. 5004 */ 5005 if (!buffer_mapped(sbh)) 5006 return error; 5007 5008 /* 5009 * If the file system is mounted read-only, don't update the 5010 * superblock write time. This avoids updating the superblock 5011 * write time when we are mounting the root file system 5012 * read/only but we need to replay the journal; at that point, 5013 * for people who are east of GMT and who make their clock 5014 * tick in localtime for Windows bug-for-bug compatibility, 5015 * the clock is set in the future, and this will cause e2fsck 5016 * to complain and force a full file system check. 5017 */ 5018 if (!(sb->s_flags & SB_RDONLY)) 5019 ext4_update_tstamp(es, s_wtime); 5020 if (sb->s_bdev->bd_part) 5021 es->s_kbytes_written = 5022 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 5023 ((part_stat_read(sb->s_bdev->bd_part, 5024 sectors[STAT_WRITE]) - 5025 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 5026 else 5027 es->s_kbytes_written = 5028 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 5029 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 5030 ext4_free_blocks_count_set(es, 5031 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 5032 &EXT4_SB(sb)->s_freeclusters_counter))); 5033 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 5034 es->s_free_inodes_count = 5035 cpu_to_le32(percpu_counter_sum_positive( 5036 &EXT4_SB(sb)->s_freeinodes_counter)); 5037 BUFFER_TRACE(sbh, "marking dirty"); 5038 ext4_superblock_csum_set(sb); 5039 if (sync) 5040 lock_buffer(sbh); 5041 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 5042 /* 5043 * Oh, dear. A previous attempt to write the 5044 * superblock failed. This could happen because the 5045 * USB device was yanked out. Or it could happen to 5046 * be a transient write error and maybe the block will 5047 * be remapped. Nothing we can do but to retry the 5048 * write and hope for the best. 5049 */ 5050 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5051 "superblock detected"); 5052 clear_buffer_write_io_error(sbh); 5053 set_buffer_uptodate(sbh); 5054 } 5055 mark_buffer_dirty(sbh); 5056 if (sync) { 5057 unlock_buffer(sbh); 5058 error = __sync_dirty_buffer(sbh, 5059 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5060 if (buffer_write_io_error(sbh)) { 5061 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5062 "superblock"); 5063 clear_buffer_write_io_error(sbh); 5064 set_buffer_uptodate(sbh); 5065 } 5066 } 5067 return error; 5068 } 5069 5070 /* 5071 * Have we just finished recovery? If so, and if we are mounting (or 5072 * remounting) the filesystem readonly, then we will end up with a 5073 * consistent fs on disk. Record that fact. 5074 */ 5075 static void ext4_mark_recovery_complete(struct super_block *sb, 5076 struct ext4_super_block *es) 5077 { 5078 journal_t *journal = EXT4_SB(sb)->s_journal; 5079 5080 if (!ext4_has_feature_journal(sb)) { 5081 BUG_ON(journal != NULL); 5082 return; 5083 } 5084 jbd2_journal_lock_updates(journal); 5085 if (jbd2_journal_flush(journal) < 0) 5086 goto out; 5087 5088 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5089 ext4_clear_feature_journal_needs_recovery(sb); 5090 ext4_commit_super(sb, 1); 5091 } 5092 5093 out: 5094 jbd2_journal_unlock_updates(journal); 5095 } 5096 5097 /* 5098 * If we are mounting (or read-write remounting) a filesystem whose journal 5099 * has recorded an error from a previous lifetime, move that error to the 5100 * main filesystem now. 5101 */ 5102 static void ext4_clear_journal_err(struct super_block *sb, 5103 struct ext4_super_block *es) 5104 { 5105 journal_t *journal; 5106 int j_errno; 5107 const char *errstr; 5108 5109 BUG_ON(!ext4_has_feature_journal(sb)); 5110 5111 journal = EXT4_SB(sb)->s_journal; 5112 5113 /* 5114 * Now check for any error status which may have been recorded in the 5115 * journal by a prior ext4_error() or ext4_abort() 5116 */ 5117 5118 j_errno = jbd2_journal_errno(journal); 5119 if (j_errno) { 5120 char nbuf[16]; 5121 5122 errstr = ext4_decode_error(sb, j_errno, nbuf); 5123 ext4_warning(sb, "Filesystem error recorded " 5124 "from previous mount: %s", errstr); 5125 ext4_warning(sb, "Marking fs in need of filesystem check."); 5126 5127 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5128 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5129 ext4_commit_super(sb, 1); 5130 5131 jbd2_journal_clear_err(journal); 5132 jbd2_journal_update_sb_errno(journal); 5133 } 5134 } 5135 5136 /* 5137 * Force the running and committing transactions to commit, 5138 * and wait on the commit. 5139 */ 5140 int ext4_force_commit(struct super_block *sb) 5141 { 5142 journal_t *journal; 5143 5144 if (sb_rdonly(sb)) 5145 return 0; 5146 5147 journal = EXT4_SB(sb)->s_journal; 5148 return ext4_journal_force_commit(journal); 5149 } 5150 5151 static int ext4_sync_fs(struct super_block *sb, int wait) 5152 { 5153 int ret = 0; 5154 tid_t target; 5155 bool needs_barrier = false; 5156 struct ext4_sb_info *sbi = EXT4_SB(sb); 5157 5158 if (unlikely(ext4_forced_shutdown(sbi))) 5159 return 0; 5160 5161 trace_ext4_sync_fs(sb, wait); 5162 flush_workqueue(sbi->rsv_conversion_wq); 5163 /* 5164 * Writeback quota in non-journalled quota case - journalled quota has 5165 * no dirty dquots 5166 */ 5167 dquot_writeback_dquots(sb, -1); 5168 /* 5169 * Data writeback is possible w/o journal transaction, so barrier must 5170 * being sent at the end of the function. But we can skip it if 5171 * transaction_commit will do it for us. 5172 */ 5173 if (sbi->s_journal) { 5174 target = jbd2_get_latest_transaction(sbi->s_journal); 5175 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 5176 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 5177 needs_barrier = true; 5178 5179 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5180 if (wait) 5181 ret = jbd2_log_wait_commit(sbi->s_journal, 5182 target); 5183 } 5184 } else if (wait && test_opt(sb, BARRIER)) 5185 needs_barrier = true; 5186 if (needs_barrier) { 5187 int err; 5188 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 5189 if (!ret) 5190 ret = err; 5191 } 5192 5193 return ret; 5194 } 5195 5196 /* 5197 * LVM calls this function before a (read-only) snapshot is created. This 5198 * gives us a chance to flush the journal completely and mark the fs clean. 5199 * 5200 * Note that only this function cannot bring a filesystem to be in a clean 5201 * state independently. It relies on upper layer to stop all data & metadata 5202 * modifications. 5203 */ 5204 static int ext4_freeze(struct super_block *sb) 5205 { 5206 int error = 0; 5207 journal_t *journal; 5208 5209 if (sb_rdonly(sb)) 5210 return 0; 5211 5212 journal = EXT4_SB(sb)->s_journal; 5213 5214 if (journal) { 5215 /* Now we set up the journal barrier. */ 5216 jbd2_journal_lock_updates(journal); 5217 5218 /* 5219 * Don't clear the needs_recovery flag if we failed to 5220 * flush the journal. 5221 */ 5222 error = jbd2_journal_flush(journal); 5223 if (error < 0) 5224 goto out; 5225 5226 /* Journal blocked and flushed, clear needs_recovery flag. */ 5227 ext4_clear_feature_journal_needs_recovery(sb); 5228 } 5229 5230 error = ext4_commit_super(sb, 1); 5231 out: 5232 if (journal) 5233 /* we rely on upper layer to stop further updates */ 5234 jbd2_journal_unlock_updates(journal); 5235 return error; 5236 } 5237 5238 /* 5239 * Called by LVM after the snapshot is done. We need to reset the RECOVER 5240 * flag here, even though the filesystem is not technically dirty yet. 5241 */ 5242 static int ext4_unfreeze(struct super_block *sb) 5243 { 5244 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 5245 return 0; 5246 5247 if (EXT4_SB(sb)->s_journal) { 5248 /* Reset the needs_recovery flag before the fs is unlocked. */ 5249 ext4_set_feature_journal_needs_recovery(sb); 5250 } 5251 5252 ext4_commit_super(sb, 1); 5253 return 0; 5254 } 5255 5256 /* 5257 * Structure to save mount options for ext4_remount's benefit 5258 */ 5259 struct ext4_mount_options { 5260 unsigned long s_mount_opt; 5261 unsigned long s_mount_opt2; 5262 kuid_t s_resuid; 5263 kgid_t s_resgid; 5264 unsigned long s_commit_interval; 5265 u32 s_min_batch_time, s_max_batch_time; 5266 #ifdef CONFIG_QUOTA 5267 int s_jquota_fmt; 5268 char *s_qf_names[EXT4_MAXQUOTAS]; 5269 #endif 5270 }; 5271 5272 static int ext4_remount(struct super_block *sb, int *flags, char *data) 5273 { 5274 struct ext4_super_block *es; 5275 struct ext4_sb_info *sbi = EXT4_SB(sb); 5276 unsigned long old_sb_flags; 5277 struct ext4_mount_options old_opts; 5278 int enable_quota = 0; 5279 ext4_group_t g; 5280 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5281 int err = 0; 5282 #ifdef CONFIG_QUOTA 5283 int i, j; 5284 char *to_free[EXT4_MAXQUOTAS]; 5285 #endif 5286 char *orig_data = kstrdup(data, GFP_KERNEL); 5287 5288 if (data && !orig_data) 5289 return -ENOMEM; 5290 5291 /* Store the original options */ 5292 old_sb_flags = sb->s_flags; 5293 old_opts.s_mount_opt = sbi->s_mount_opt; 5294 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5295 old_opts.s_resuid = sbi->s_resuid; 5296 old_opts.s_resgid = sbi->s_resgid; 5297 old_opts.s_commit_interval = sbi->s_commit_interval; 5298 old_opts.s_min_batch_time = sbi->s_min_batch_time; 5299 old_opts.s_max_batch_time = sbi->s_max_batch_time; 5300 #ifdef CONFIG_QUOTA 5301 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5302 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5303 if (sbi->s_qf_names[i]) { 5304 char *qf_name = get_qf_name(sb, sbi, i); 5305 5306 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 5307 if (!old_opts.s_qf_names[i]) { 5308 for (j = 0; j < i; j++) 5309 kfree(old_opts.s_qf_names[j]); 5310 kfree(orig_data); 5311 return -ENOMEM; 5312 } 5313 } else 5314 old_opts.s_qf_names[i] = NULL; 5315 #endif 5316 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5317 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5318 5319 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5320 err = -EINVAL; 5321 goto restore_opts; 5322 } 5323 5324 ext4_clamp_want_extra_isize(sb); 5325 5326 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 5327 test_opt(sb, JOURNAL_CHECKSUM)) { 5328 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 5329 "during remount not supported; ignoring"); 5330 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5331 } 5332 5333 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5334 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5335 ext4_msg(sb, KERN_ERR, "can't mount with " 5336 "both data=journal and delalloc"); 5337 err = -EINVAL; 5338 goto restore_opts; 5339 } 5340 if (test_opt(sb, DIOREAD_NOLOCK)) { 5341 ext4_msg(sb, KERN_ERR, "can't mount with " 5342 "both data=journal and dioread_nolock"); 5343 err = -EINVAL; 5344 goto restore_opts; 5345 } 5346 if (test_opt(sb, DAX)) { 5347 ext4_msg(sb, KERN_ERR, "can't mount with " 5348 "both data=journal and dax"); 5349 err = -EINVAL; 5350 goto restore_opts; 5351 } 5352 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5353 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5354 ext4_msg(sb, KERN_ERR, "can't mount with " 5355 "journal_async_commit in data=ordered mode"); 5356 err = -EINVAL; 5357 goto restore_opts; 5358 } 5359 } 5360 5361 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5362 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5363 err = -EINVAL; 5364 goto restore_opts; 5365 } 5366 5367 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 5368 ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 5369 "dax flag with busy inodes while remounting"); 5370 sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 5371 } 5372 5373 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 5374 ext4_abort(sb, "Abort forced by user"); 5375 5376 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5377 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5378 5379 es = sbi->s_es; 5380 5381 if (sbi->s_journal) { 5382 ext4_init_journal_params(sb, sbi->s_journal); 5383 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5384 } 5385 5386 if (*flags & SB_LAZYTIME) 5387 sb->s_flags |= SB_LAZYTIME; 5388 5389 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 5390 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5391 err = -EROFS; 5392 goto restore_opts; 5393 } 5394 5395 if (*flags & SB_RDONLY) { 5396 err = sync_filesystem(sb); 5397 if (err < 0) 5398 goto restore_opts; 5399 err = dquot_suspend(sb, -1); 5400 if (err < 0) 5401 goto restore_opts; 5402 5403 /* 5404 * First of all, the unconditional stuff we have to do 5405 * to disable replay of the journal when we next remount 5406 */ 5407 sb->s_flags |= SB_RDONLY; 5408 5409 /* 5410 * OK, test if we are remounting a valid rw partition 5411 * readonly, and if so set the rdonly flag and then 5412 * mark the partition as valid again. 5413 */ 5414 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5415 (sbi->s_mount_state & EXT4_VALID_FS)) 5416 es->s_state = cpu_to_le16(sbi->s_mount_state); 5417 5418 if (sbi->s_journal) 5419 ext4_mark_recovery_complete(sb, es); 5420 if (sbi->s_mmp_tsk) 5421 kthread_stop(sbi->s_mmp_tsk); 5422 } else { 5423 /* Make sure we can mount this feature set readwrite */ 5424 if (ext4_has_feature_readonly(sb) || 5425 !ext4_feature_set_ok(sb, 0)) { 5426 err = -EROFS; 5427 goto restore_opts; 5428 } 5429 /* 5430 * Make sure the group descriptor checksums 5431 * are sane. If they aren't, refuse to remount r/w. 5432 */ 5433 for (g = 0; g < sbi->s_groups_count; g++) { 5434 struct ext4_group_desc *gdp = 5435 ext4_get_group_desc(sb, g, NULL); 5436 5437 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5438 ext4_msg(sb, KERN_ERR, 5439 "ext4_remount: Checksum for group %u failed (%u!=%u)", 5440 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 5441 le16_to_cpu(gdp->bg_checksum)); 5442 err = -EFSBADCRC; 5443 goto restore_opts; 5444 } 5445 } 5446 5447 /* 5448 * If we have an unprocessed orphan list hanging 5449 * around from a previously readonly bdev mount, 5450 * require a full umount/remount for now. 5451 */ 5452 if (es->s_last_orphan) { 5453 ext4_msg(sb, KERN_WARNING, "Couldn't " 5454 "remount RDWR because of unprocessed " 5455 "orphan inode list. Please " 5456 "umount/remount instead"); 5457 err = -EINVAL; 5458 goto restore_opts; 5459 } 5460 5461 /* 5462 * Mounting a RDONLY partition read-write, so reread 5463 * and store the current valid flag. (It may have 5464 * been changed by e2fsck since we originally mounted 5465 * the partition.) 5466 */ 5467 if (sbi->s_journal) 5468 ext4_clear_journal_err(sb, es); 5469 sbi->s_mount_state = le16_to_cpu(es->s_state); 5470 5471 err = ext4_setup_super(sb, es, 0); 5472 if (err) 5473 goto restore_opts; 5474 5475 sb->s_flags &= ~SB_RDONLY; 5476 if (ext4_has_feature_mmp(sb)) 5477 if (ext4_multi_mount_protect(sb, 5478 le64_to_cpu(es->s_mmp_block))) { 5479 err = -EROFS; 5480 goto restore_opts; 5481 } 5482 enable_quota = 1; 5483 } 5484 } 5485 5486 /* 5487 * Reinitialize lazy itable initialization thread based on 5488 * current settings 5489 */ 5490 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5491 ext4_unregister_li_request(sb); 5492 else { 5493 ext4_group_t first_not_zeroed; 5494 first_not_zeroed = ext4_has_uninit_itable(sb); 5495 ext4_register_li_request(sb, first_not_zeroed); 5496 } 5497 5498 ext4_setup_system_zone(sb); 5499 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5500 err = ext4_commit_super(sb, 1); 5501 if (err) 5502 goto restore_opts; 5503 } 5504 5505 #ifdef CONFIG_QUOTA 5506 /* Release old quota file names */ 5507 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5508 kfree(old_opts.s_qf_names[i]); 5509 if (enable_quota) { 5510 if (sb_any_quota_suspended(sb)) 5511 dquot_resume(sb, -1); 5512 else if (ext4_has_feature_quota(sb)) { 5513 err = ext4_enable_quotas(sb); 5514 if (err) 5515 goto restore_opts; 5516 } 5517 } 5518 #endif 5519 5520 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 5521 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5522 kfree(orig_data); 5523 return 0; 5524 5525 restore_opts: 5526 sb->s_flags = old_sb_flags; 5527 sbi->s_mount_opt = old_opts.s_mount_opt; 5528 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5529 sbi->s_resuid = old_opts.s_resuid; 5530 sbi->s_resgid = old_opts.s_resgid; 5531 sbi->s_commit_interval = old_opts.s_commit_interval; 5532 sbi->s_min_batch_time = old_opts.s_min_batch_time; 5533 sbi->s_max_batch_time = old_opts.s_max_batch_time; 5534 #ifdef CONFIG_QUOTA 5535 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5536 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 5537 to_free[i] = get_qf_name(sb, sbi, i); 5538 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5539 } 5540 synchronize_rcu(); 5541 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5542 kfree(to_free[i]); 5543 #endif 5544 kfree(orig_data); 5545 return err; 5546 } 5547 5548 #ifdef CONFIG_QUOTA 5549 static int ext4_statfs_project(struct super_block *sb, 5550 kprojid_t projid, struct kstatfs *buf) 5551 { 5552 struct kqid qid; 5553 struct dquot *dquot; 5554 u64 limit; 5555 u64 curblock; 5556 5557 qid = make_kqid_projid(projid); 5558 dquot = dqget(sb, qid); 5559 if (IS_ERR(dquot)) 5560 return PTR_ERR(dquot); 5561 spin_lock(&dquot->dq_dqb_lock); 5562 5563 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 5564 dquot->dq_dqb.dqb_bsoftlimit : 5565 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 5566 if (limit && buf->f_blocks > limit) { 5567 curblock = (dquot->dq_dqb.dqb_curspace + 5568 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5569 buf->f_blocks = limit; 5570 buf->f_bfree = buf->f_bavail = 5571 (buf->f_blocks > curblock) ? 5572 (buf->f_blocks - curblock) : 0; 5573 } 5574 5575 limit = dquot->dq_dqb.dqb_isoftlimit ? 5576 dquot->dq_dqb.dqb_isoftlimit : 5577 dquot->dq_dqb.dqb_ihardlimit; 5578 if (limit && buf->f_files > limit) { 5579 buf->f_files = limit; 5580 buf->f_ffree = 5581 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5582 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5583 } 5584 5585 spin_unlock(&dquot->dq_dqb_lock); 5586 dqput(dquot); 5587 return 0; 5588 } 5589 #endif 5590 5591 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5592 { 5593 struct super_block *sb = dentry->d_sb; 5594 struct ext4_sb_info *sbi = EXT4_SB(sb); 5595 struct ext4_super_block *es = sbi->s_es; 5596 ext4_fsblk_t overhead = 0, resv_blocks; 5597 u64 fsid; 5598 s64 bfree; 5599 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5600 5601 if (!test_opt(sb, MINIX_DF)) 5602 overhead = sbi->s_overhead; 5603 5604 buf->f_type = EXT4_SUPER_MAGIC; 5605 buf->f_bsize = sb->s_blocksize; 5606 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5607 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5608 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5609 /* prevent underflow in case that few free space is available */ 5610 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5611 buf->f_bavail = buf->f_bfree - 5612 (ext4_r_blocks_count(es) + resv_blocks); 5613 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5614 buf->f_bavail = 0; 5615 buf->f_files = le32_to_cpu(es->s_inodes_count); 5616 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5617 buf->f_namelen = EXT4_NAME_LEN; 5618 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5619 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5620 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5621 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5622 5623 #ifdef CONFIG_QUOTA 5624 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5625 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5626 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5627 #endif 5628 return 0; 5629 } 5630 5631 5632 #ifdef CONFIG_QUOTA 5633 5634 /* 5635 * Helper functions so that transaction is started before we acquire dqio_sem 5636 * to keep correct lock ordering of transaction > dqio_sem 5637 */ 5638 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5639 { 5640 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5641 } 5642 5643 static int ext4_write_dquot(struct dquot *dquot) 5644 { 5645 int ret, err; 5646 handle_t *handle; 5647 struct inode *inode; 5648 5649 inode = dquot_to_inode(dquot); 5650 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5651 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5652 if (IS_ERR(handle)) 5653 return PTR_ERR(handle); 5654 ret = dquot_commit(dquot); 5655 err = ext4_journal_stop(handle); 5656 if (!ret) 5657 ret = err; 5658 return ret; 5659 } 5660 5661 static int ext4_acquire_dquot(struct dquot *dquot) 5662 { 5663 int ret, err; 5664 handle_t *handle; 5665 5666 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5667 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5668 if (IS_ERR(handle)) 5669 return PTR_ERR(handle); 5670 ret = dquot_acquire(dquot); 5671 err = ext4_journal_stop(handle); 5672 if (!ret) 5673 ret = err; 5674 return ret; 5675 } 5676 5677 static int ext4_release_dquot(struct dquot *dquot) 5678 { 5679 int ret, err; 5680 handle_t *handle; 5681 5682 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5683 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5684 if (IS_ERR(handle)) { 5685 /* Release dquot anyway to avoid endless cycle in dqput() */ 5686 dquot_release(dquot); 5687 return PTR_ERR(handle); 5688 } 5689 ret = dquot_release(dquot); 5690 err = ext4_journal_stop(handle); 5691 if (!ret) 5692 ret = err; 5693 return ret; 5694 } 5695 5696 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5697 { 5698 struct super_block *sb = dquot->dq_sb; 5699 struct ext4_sb_info *sbi = EXT4_SB(sb); 5700 5701 /* Are we journaling quotas? */ 5702 if (ext4_has_feature_quota(sb) || 5703 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5704 dquot_mark_dquot_dirty(dquot); 5705 return ext4_write_dquot(dquot); 5706 } else { 5707 return dquot_mark_dquot_dirty(dquot); 5708 } 5709 } 5710 5711 static int ext4_write_info(struct super_block *sb, int type) 5712 { 5713 int ret, err; 5714 handle_t *handle; 5715 5716 /* Data block + inode block */ 5717 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5718 if (IS_ERR(handle)) 5719 return PTR_ERR(handle); 5720 ret = dquot_commit_info(sb, type); 5721 err = ext4_journal_stop(handle); 5722 if (!ret) 5723 ret = err; 5724 return ret; 5725 } 5726 5727 /* 5728 * Turn on quotas during mount time - we need to find 5729 * the quota file and such... 5730 */ 5731 static int ext4_quota_on_mount(struct super_block *sb, int type) 5732 { 5733 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 5734 EXT4_SB(sb)->s_jquota_fmt, type); 5735 } 5736 5737 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5738 { 5739 struct ext4_inode_info *ei = EXT4_I(inode); 5740 5741 /* The first argument of lockdep_set_subclass has to be 5742 * *exactly* the same as the argument to init_rwsem() --- in 5743 * this case, in init_once() --- or lockdep gets unhappy 5744 * because the name of the lock is set using the 5745 * stringification of the argument to init_rwsem(). 5746 */ 5747 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5748 lockdep_set_subclass(&ei->i_data_sem, subclass); 5749 } 5750 5751 /* 5752 * Standard function to be called on quota_on 5753 */ 5754 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5755 const struct path *path) 5756 { 5757 int err; 5758 5759 if (!test_opt(sb, QUOTA)) 5760 return -EINVAL; 5761 5762 /* Quotafile not on the same filesystem? */ 5763 if (path->dentry->d_sb != sb) 5764 return -EXDEV; 5765 /* Journaling quota? */ 5766 if (EXT4_SB(sb)->s_qf_names[type]) { 5767 /* Quotafile not in fs root? */ 5768 if (path->dentry->d_parent != sb->s_root) 5769 ext4_msg(sb, KERN_WARNING, 5770 "Quota file not on filesystem root. " 5771 "Journaled quota will not work"); 5772 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 5773 } else { 5774 /* 5775 * Clear the flag just in case mount options changed since 5776 * last time. 5777 */ 5778 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 5779 } 5780 5781 /* 5782 * When we journal data on quota file, we have to flush journal to see 5783 * all updates to the file when we bypass pagecache... 5784 */ 5785 if (EXT4_SB(sb)->s_journal && 5786 ext4_should_journal_data(d_inode(path->dentry))) { 5787 /* 5788 * We don't need to lock updates but journal_flush() could 5789 * otherwise be livelocked... 5790 */ 5791 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5792 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5793 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5794 if (err) 5795 return err; 5796 } 5797 5798 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5799 err = dquot_quota_on(sb, type, format_id, path); 5800 if (err) { 5801 lockdep_set_quota_inode(path->dentry->d_inode, 5802 I_DATA_SEM_NORMAL); 5803 } else { 5804 struct inode *inode = d_inode(path->dentry); 5805 handle_t *handle; 5806 5807 /* 5808 * Set inode flags to prevent userspace from messing with quota 5809 * files. If this fails, we return success anyway since quotas 5810 * are already enabled and this is not a hard failure. 5811 */ 5812 inode_lock(inode); 5813 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5814 if (IS_ERR(handle)) 5815 goto unlock_inode; 5816 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 5817 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 5818 S_NOATIME | S_IMMUTABLE); 5819 ext4_mark_inode_dirty(handle, inode); 5820 ext4_journal_stop(handle); 5821 unlock_inode: 5822 inode_unlock(inode); 5823 } 5824 return err; 5825 } 5826 5827 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5828 unsigned int flags) 5829 { 5830 int err; 5831 struct inode *qf_inode; 5832 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5833 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5834 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5835 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5836 }; 5837 5838 BUG_ON(!ext4_has_feature_quota(sb)); 5839 5840 if (!qf_inums[type]) 5841 return -EPERM; 5842 5843 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 5844 if (IS_ERR(qf_inode)) { 5845 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5846 return PTR_ERR(qf_inode); 5847 } 5848 5849 /* Don't account quota for quota files to avoid recursion */ 5850 qf_inode->i_flags |= S_NOQUOTA; 5851 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 5852 err = dquot_enable(qf_inode, type, format_id, flags); 5853 if (err) 5854 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 5855 iput(qf_inode); 5856 5857 return err; 5858 } 5859 5860 /* Enable usage tracking for all quota types. */ 5861 static int ext4_enable_quotas(struct super_block *sb) 5862 { 5863 int type, err = 0; 5864 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5865 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5866 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5867 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5868 }; 5869 bool quota_mopt[EXT4_MAXQUOTAS] = { 5870 test_opt(sb, USRQUOTA), 5871 test_opt(sb, GRPQUOTA), 5872 test_opt(sb, PRJQUOTA), 5873 }; 5874 5875 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 5876 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 5877 if (qf_inums[type]) { 5878 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5879 DQUOT_USAGE_ENABLED | 5880 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 5881 if (err) { 5882 ext4_warning(sb, 5883 "Failed to enable quota tracking " 5884 "(type=%d, err=%d). Please run " 5885 "e2fsck to fix.", type, err); 5886 for (type--; type >= 0; type--) 5887 dquot_quota_off(sb, type); 5888 5889 return err; 5890 } 5891 } 5892 } 5893 return 0; 5894 } 5895 5896 static int ext4_quota_off(struct super_block *sb, int type) 5897 { 5898 struct inode *inode = sb_dqopt(sb)->files[type]; 5899 handle_t *handle; 5900 int err; 5901 5902 /* Force all delayed allocation blocks to be allocated. 5903 * Caller already holds s_umount sem */ 5904 if (test_opt(sb, DELALLOC)) 5905 sync_filesystem(sb); 5906 5907 if (!inode || !igrab(inode)) 5908 goto out; 5909 5910 err = dquot_quota_off(sb, type); 5911 if (err || ext4_has_feature_quota(sb)) 5912 goto out_put; 5913 5914 inode_lock(inode); 5915 /* 5916 * Update modification times of quota files when userspace can 5917 * start looking at them. If we fail, we return success anyway since 5918 * this is not a hard failure and quotas are already disabled. 5919 */ 5920 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5921 if (IS_ERR(handle)) 5922 goto out_unlock; 5923 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 5924 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 5925 inode->i_mtime = inode->i_ctime = current_time(inode); 5926 ext4_mark_inode_dirty(handle, inode); 5927 ext4_journal_stop(handle); 5928 out_unlock: 5929 inode_unlock(inode); 5930 out_put: 5931 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 5932 iput(inode); 5933 return err; 5934 out: 5935 return dquot_quota_off(sb, type); 5936 } 5937 5938 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5939 * acquiring the locks... As quota files are never truncated and quota code 5940 * itself serializes the operations (and no one else should touch the files) 5941 * we don't have to be afraid of races */ 5942 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5943 size_t len, loff_t off) 5944 { 5945 struct inode *inode = sb_dqopt(sb)->files[type]; 5946 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5947 int offset = off & (sb->s_blocksize - 1); 5948 int tocopy; 5949 size_t toread; 5950 struct buffer_head *bh; 5951 loff_t i_size = i_size_read(inode); 5952 5953 if (off > i_size) 5954 return 0; 5955 if (off+len > i_size) 5956 len = i_size-off; 5957 toread = len; 5958 while (toread > 0) { 5959 tocopy = sb->s_blocksize - offset < toread ? 5960 sb->s_blocksize - offset : toread; 5961 bh = ext4_bread(NULL, inode, blk, 0); 5962 if (IS_ERR(bh)) 5963 return PTR_ERR(bh); 5964 if (!bh) /* A hole? */ 5965 memset(data, 0, tocopy); 5966 else 5967 memcpy(data, bh->b_data+offset, tocopy); 5968 brelse(bh); 5969 offset = 0; 5970 toread -= tocopy; 5971 data += tocopy; 5972 blk++; 5973 } 5974 return len; 5975 } 5976 5977 /* Write to quotafile (we know the transaction is already started and has 5978 * enough credits) */ 5979 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5980 const char *data, size_t len, loff_t off) 5981 { 5982 struct inode *inode = sb_dqopt(sb)->files[type]; 5983 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5984 int err, offset = off & (sb->s_blocksize - 1); 5985 int retries = 0; 5986 struct buffer_head *bh; 5987 handle_t *handle = journal_current_handle(); 5988 5989 if (EXT4_SB(sb)->s_journal && !handle) { 5990 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5991 " cancelled because transaction is not started", 5992 (unsigned long long)off, (unsigned long long)len); 5993 return -EIO; 5994 } 5995 /* 5996 * Since we account only one data block in transaction credits, 5997 * then it is impossible to cross a block boundary. 5998 */ 5999 if (sb->s_blocksize - offset < len) { 6000 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6001 " cancelled because not block aligned", 6002 (unsigned long long)off, (unsigned long long)len); 6003 return -EIO; 6004 } 6005 6006 do { 6007 bh = ext4_bread(handle, inode, blk, 6008 EXT4_GET_BLOCKS_CREATE | 6009 EXT4_GET_BLOCKS_METADATA_NOFAIL); 6010 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 6011 ext4_should_retry_alloc(inode->i_sb, &retries)); 6012 if (IS_ERR(bh)) 6013 return PTR_ERR(bh); 6014 if (!bh) 6015 goto out; 6016 BUFFER_TRACE(bh, "get write access"); 6017 err = ext4_journal_get_write_access(handle, bh); 6018 if (err) { 6019 brelse(bh); 6020 return err; 6021 } 6022 lock_buffer(bh); 6023 memcpy(bh->b_data+offset, data, len); 6024 flush_dcache_page(bh->b_page); 6025 unlock_buffer(bh); 6026 err = ext4_handle_dirty_metadata(handle, NULL, bh); 6027 brelse(bh); 6028 out: 6029 if (inode->i_size < off + len) { 6030 i_size_write(inode, off + len); 6031 EXT4_I(inode)->i_disksize = inode->i_size; 6032 ext4_mark_inode_dirty(handle, inode); 6033 } 6034 return len; 6035 } 6036 6037 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) 6038 { 6039 const struct quota_format_ops *ops; 6040 6041 if (!sb_has_quota_loaded(sb, qid->type)) 6042 return -ESRCH; 6043 ops = sb_dqopt(sb)->ops[qid->type]; 6044 if (!ops || !ops->get_next_id) 6045 return -ENOSYS; 6046 return dquot_get_next_id(sb, qid); 6047 } 6048 #endif 6049 6050 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6051 const char *dev_name, void *data) 6052 { 6053 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6054 } 6055 6056 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 6057 static inline void register_as_ext2(void) 6058 { 6059 int err = register_filesystem(&ext2_fs_type); 6060 if (err) 6061 printk(KERN_WARNING 6062 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 6063 } 6064 6065 static inline void unregister_as_ext2(void) 6066 { 6067 unregister_filesystem(&ext2_fs_type); 6068 } 6069 6070 static inline int ext2_feature_set_ok(struct super_block *sb) 6071 { 6072 if (ext4_has_unknown_ext2_incompat_features(sb)) 6073 return 0; 6074 if (sb_rdonly(sb)) 6075 return 1; 6076 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 6077 return 0; 6078 return 1; 6079 } 6080 #else 6081 static inline void register_as_ext2(void) { } 6082 static inline void unregister_as_ext2(void) { } 6083 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 6084 #endif 6085 6086 static inline void register_as_ext3(void) 6087 { 6088 int err = register_filesystem(&ext3_fs_type); 6089 if (err) 6090 printk(KERN_WARNING 6091 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 6092 } 6093 6094 static inline void unregister_as_ext3(void) 6095 { 6096 unregister_filesystem(&ext3_fs_type); 6097 } 6098 6099 static inline int ext3_feature_set_ok(struct super_block *sb) 6100 { 6101 if (ext4_has_unknown_ext3_incompat_features(sb)) 6102 return 0; 6103 if (!ext4_has_feature_journal(sb)) 6104 return 0; 6105 if (sb_rdonly(sb)) 6106 return 1; 6107 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 6108 return 0; 6109 return 1; 6110 } 6111 6112 static struct file_system_type ext4_fs_type = { 6113 .owner = THIS_MODULE, 6114 .name = "ext4", 6115 .mount = ext4_mount, 6116 .kill_sb = kill_block_super, 6117 .fs_flags = FS_REQUIRES_DEV, 6118 }; 6119 MODULE_ALIAS_FS("ext4"); 6120 6121 /* Shared across all ext4 file systems */ 6122 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6123 6124 static int __init ext4_init_fs(void) 6125 { 6126 int i, err; 6127 6128 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 6129 ext4_li_info = NULL; 6130 mutex_init(&ext4_li_mtx); 6131 6132 /* Build-time check for flags consistency */ 6133 ext4_check_flag_values(); 6134 6135 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6136 init_waitqueue_head(&ext4__ioend_wq[i]); 6137 6138 err = ext4_init_es(); 6139 if (err) 6140 return err; 6141 6142 err = ext4_init_pending(); 6143 if (err) 6144 goto out7; 6145 6146 err = ext4_init_post_read_processing(); 6147 if (err) 6148 goto out6; 6149 6150 err = ext4_init_pageio(); 6151 if (err) 6152 goto out5; 6153 6154 err = ext4_init_system_zone(); 6155 if (err) 6156 goto out4; 6157 6158 err = ext4_init_sysfs(); 6159 if (err) 6160 goto out3; 6161 6162 err = ext4_init_mballoc(); 6163 if (err) 6164 goto out2; 6165 err = init_inodecache(); 6166 if (err) 6167 goto out1; 6168 register_as_ext3(); 6169 register_as_ext2(); 6170 err = register_filesystem(&ext4_fs_type); 6171 if (err) 6172 goto out; 6173 6174 return 0; 6175 out: 6176 unregister_as_ext2(); 6177 unregister_as_ext3(); 6178 destroy_inodecache(); 6179 out1: 6180 ext4_exit_mballoc(); 6181 out2: 6182 ext4_exit_sysfs(); 6183 out3: 6184 ext4_exit_system_zone(); 6185 out4: 6186 ext4_exit_pageio(); 6187 out5: 6188 ext4_exit_post_read_processing(); 6189 out6: 6190 ext4_exit_pending(); 6191 out7: 6192 ext4_exit_es(); 6193 6194 return err; 6195 } 6196 6197 static void __exit ext4_exit_fs(void) 6198 { 6199 ext4_destroy_lazyinit_thread(); 6200 unregister_as_ext2(); 6201 unregister_as_ext3(); 6202 unregister_filesystem(&ext4_fs_type); 6203 destroy_inodecache(); 6204 ext4_exit_mballoc(); 6205 ext4_exit_sysfs(); 6206 ext4_exit_system_zone(); 6207 ext4_exit_pageio(); 6208 ext4_exit_post_read_processing(); 6209 ext4_exit_es(); 6210 ext4_exit_pending(); 6211 } 6212 6213 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 6214 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6215 MODULE_LICENSE("GPL"); 6216 MODULE_SOFTDEP("pre: crc32c"); 6217 module_init(ext4_init_fs) 6218 module_exit(ext4_exit_fs) 6219