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 46 #include <linux/kthread.h> 47 #include <linux/freezer.h> 48 49 #include "ext4.h" 50 #include "ext4_extents.h" /* Needed for trace points definition */ 51 #include "ext4_jbd2.h" 52 #include "xattr.h" 53 #include "acl.h" 54 #include "mballoc.h" 55 #include "fsmap.h" 56 57 #define CREATE_TRACE_POINTS 58 #include <trace/events/ext4.h> 59 60 static struct ext4_lazy_init *ext4_li_info; 61 static struct mutex ext4_li_mtx; 62 static struct ratelimit_state ext4_mount_msg_ratelimit; 63 64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 65 unsigned long journal_devnum); 66 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 67 static int ext4_commit_super(struct super_block *sb, int sync); 68 static void ext4_mark_recovery_complete(struct super_block *sb, 69 struct ext4_super_block *es); 70 static void ext4_clear_journal_err(struct super_block *sb, 71 struct ext4_super_block *es); 72 static int ext4_sync_fs(struct super_block *sb, int wait); 73 static int ext4_remount(struct super_block *sb, int *flags, char *data); 74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 75 static int ext4_unfreeze(struct super_block *sb); 76 static int ext4_freeze(struct super_block *sb); 77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 78 const char *dev_name, void *data); 79 static inline int ext2_feature_set_ok(struct super_block *sb); 80 static inline int ext3_feature_set_ok(struct super_block *sb); 81 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 82 static void ext4_destroy_lazyinit_thread(void); 83 static void ext4_unregister_li_request(struct super_block *sb); 84 static void ext4_clear_request_list(void); 85 static struct inode *ext4_get_journal_inode(struct super_block *sb, 86 unsigned int journal_inum); 87 88 /* 89 * Lock ordering 90 * 91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 92 * i_mmap_rwsem (inode->i_mmap_rwsem)! 93 * 94 * page fault path: 95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 96 * page lock -> i_data_sem (rw) 97 * 98 * buffered write path: 99 * sb_start_write -> i_mutex -> mmap_sem 100 * sb_start_write -> i_mutex -> transaction start -> page lock -> 101 * i_data_sem (rw) 102 * 103 * truncate: 104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 106 * i_data_sem (rw) 107 * 108 * direct IO: 109 * sb_start_write -> i_mutex -> mmap_sem 110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 111 * 112 * writepages: 113 * transaction start -> page lock(s) -> i_data_sem (rw) 114 */ 115 116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 117 static struct file_system_type ext2_fs_type = { 118 .owner = THIS_MODULE, 119 .name = "ext2", 120 .mount = ext4_mount, 121 .kill_sb = kill_block_super, 122 .fs_flags = FS_REQUIRES_DEV, 123 }; 124 MODULE_ALIAS_FS("ext2"); 125 MODULE_ALIAS("ext2"); 126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 127 #else 128 #define IS_EXT2_SB(sb) (0) 129 #endif 130 131 132 static struct file_system_type ext3_fs_type = { 133 .owner = THIS_MODULE, 134 .name = "ext3", 135 .mount = ext4_mount, 136 .kill_sb = kill_block_super, 137 .fs_flags = FS_REQUIRES_DEV, 138 }; 139 MODULE_ALIAS_FS("ext3"); 140 MODULE_ALIAS("ext3"); 141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 142 143 /* 144 * This works like sb_bread() except it uses ERR_PTR for error 145 * returns. Currently with sb_bread it's impossible to distinguish 146 * between ENOMEM and EIO situations (since both result in a NULL 147 * return. 148 */ 149 struct buffer_head * 150 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags) 151 { 152 struct buffer_head *bh = sb_getblk(sb, block); 153 154 if (bh == NULL) 155 return ERR_PTR(-ENOMEM); 156 if (buffer_uptodate(bh)) 157 return bh; 158 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh); 159 wait_on_buffer(bh); 160 if (buffer_uptodate(bh)) 161 return bh; 162 put_bh(bh); 163 return ERR_PTR(-EIO); 164 } 165 166 static int ext4_verify_csum_type(struct super_block *sb, 167 struct ext4_super_block *es) 168 { 169 if (!ext4_has_feature_metadata_csum(sb)) 170 return 1; 171 172 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 173 } 174 175 static __le32 ext4_superblock_csum(struct super_block *sb, 176 struct ext4_super_block *es) 177 { 178 struct ext4_sb_info *sbi = EXT4_SB(sb); 179 int offset = offsetof(struct ext4_super_block, s_checksum); 180 __u32 csum; 181 182 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 183 184 return cpu_to_le32(csum); 185 } 186 187 static int ext4_superblock_csum_verify(struct super_block *sb, 188 struct ext4_super_block *es) 189 { 190 if (!ext4_has_metadata_csum(sb)) 191 return 1; 192 193 return es->s_checksum == ext4_superblock_csum(sb, es); 194 } 195 196 void ext4_superblock_csum_set(struct super_block *sb) 197 { 198 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 199 200 if (!ext4_has_metadata_csum(sb)) 201 return; 202 203 es->s_checksum = ext4_superblock_csum(sb, es); 204 } 205 206 void *ext4_kvmalloc(size_t size, gfp_t flags) 207 { 208 void *ret; 209 210 ret = kmalloc(size, flags | __GFP_NOWARN); 211 if (!ret) 212 ret = __vmalloc(size, flags, PAGE_KERNEL); 213 return ret; 214 } 215 216 void *ext4_kvzalloc(size_t size, gfp_t flags) 217 { 218 void *ret; 219 220 ret = kzalloc(size, flags | __GFP_NOWARN); 221 if (!ret) 222 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 223 return ret; 224 } 225 226 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 227 struct ext4_group_desc *bg) 228 { 229 return le32_to_cpu(bg->bg_block_bitmap_lo) | 230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 231 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 232 } 233 234 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 235 struct ext4_group_desc *bg) 236 { 237 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 239 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 240 } 241 242 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 243 struct ext4_group_desc *bg) 244 { 245 return le32_to_cpu(bg->bg_inode_table_lo) | 246 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 247 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 248 } 249 250 __u32 ext4_free_group_clusters(struct super_block *sb, 251 struct ext4_group_desc *bg) 252 { 253 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 254 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 255 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 256 } 257 258 __u32 ext4_free_inodes_count(struct super_block *sb, 259 struct ext4_group_desc *bg) 260 { 261 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 262 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 263 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 264 } 265 266 __u32 ext4_used_dirs_count(struct super_block *sb, 267 struct ext4_group_desc *bg) 268 { 269 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 270 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 271 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 272 } 273 274 __u32 ext4_itable_unused_count(struct super_block *sb, 275 struct ext4_group_desc *bg) 276 { 277 return le16_to_cpu(bg->bg_itable_unused_lo) | 278 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 279 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 280 } 281 282 void ext4_block_bitmap_set(struct super_block *sb, 283 struct ext4_group_desc *bg, ext4_fsblk_t blk) 284 { 285 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 287 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 288 } 289 290 void ext4_inode_bitmap_set(struct super_block *sb, 291 struct ext4_group_desc *bg, ext4_fsblk_t blk) 292 { 293 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 295 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 296 } 297 298 void ext4_inode_table_set(struct super_block *sb, 299 struct ext4_group_desc *bg, ext4_fsblk_t blk) 300 { 301 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 302 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 303 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 304 } 305 306 void ext4_free_group_clusters_set(struct super_block *sb, 307 struct ext4_group_desc *bg, __u32 count) 308 { 309 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 310 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 311 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 312 } 313 314 void ext4_free_inodes_set(struct super_block *sb, 315 struct ext4_group_desc *bg, __u32 count) 316 { 317 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 318 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 319 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 320 } 321 322 void ext4_used_dirs_set(struct super_block *sb, 323 struct ext4_group_desc *bg, __u32 count) 324 { 325 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 326 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 327 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 328 } 329 330 void ext4_itable_unused_set(struct super_block *sb, 331 struct ext4_group_desc *bg, __u32 count) 332 { 333 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 334 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 335 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 336 } 337 338 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi) 339 { 340 time64_t now = ktime_get_real_seconds(); 341 342 now = clamp_val(now, 0, (1ull << 40) - 1); 343 344 *lo = cpu_to_le32(lower_32_bits(now)); 345 *hi = upper_32_bits(now); 346 } 347 348 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 349 { 350 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 351 } 352 #define ext4_update_tstamp(es, tstamp) \ 353 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 354 #define ext4_get_tstamp(es, tstamp) \ 355 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 356 357 static void __save_error_info(struct super_block *sb, const char *func, 358 unsigned int line) 359 { 360 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 361 362 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 363 if (bdev_read_only(sb->s_bdev)) 364 return; 365 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 366 ext4_update_tstamp(es, s_last_error_time); 367 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 368 es->s_last_error_line = cpu_to_le32(line); 369 if (!es->s_first_error_time) { 370 es->s_first_error_time = es->s_last_error_time; 371 es->s_first_error_time_hi = es->s_last_error_time_hi; 372 strncpy(es->s_first_error_func, func, 373 sizeof(es->s_first_error_func)); 374 es->s_first_error_line = cpu_to_le32(line); 375 es->s_first_error_ino = es->s_last_error_ino; 376 es->s_first_error_block = es->s_last_error_block; 377 } 378 /* 379 * Start the daily error reporting function if it hasn't been 380 * started already 381 */ 382 if (!es->s_error_count) 383 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 384 le32_add_cpu(&es->s_error_count, 1); 385 } 386 387 static void save_error_info(struct super_block *sb, const char *func, 388 unsigned int line) 389 { 390 __save_error_info(sb, func, line); 391 ext4_commit_super(sb, 1); 392 } 393 394 /* 395 * The del_gendisk() function uninitializes the disk-specific data 396 * structures, including the bdi structure, without telling anyone 397 * else. Once this happens, any attempt to call mark_buffer_dirty() 398 * (for example, by ext4_commit_super), will cause a kernel OOPS. 399 * This is a kludge to prevent these oops until we can put in a proper 400 * hook in del_gendisk() to inform the VFS and file system layers. 401 */ 402 static int block_device_ejected(struct super_block *sb) 403 { 404 struct inode *bd_inode = sb->s_bdev->bd_inode; 405 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 406 407 return bdi->dev == NULL; 408 } 409 410 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 411 { 412 struct super_block *sb = journal->j_private; 413 struct ext4_sb_info *sbi = EXT4_SB(sb); 414 int error = is_journal_aborted(journal); 415 struct ext4_journal_cb_entry *jce; 416 417 BUG_ON(txn->t_state == T_FINISHED); 418 419 ext4_process_freed_data(sb, txn->t_tid); 420 421 spin_lock(&sbi->s_md_lock); 422 while (!list_empty(&txn->t_private_list)) { 423 jce = list_entry(txn->t_private_list.next, 424 struct ext4_journal_cb_entry, jce_list); 425 list_del_init(&jce->jce_list); 426 spin_unlock(&sbi->s_md_lock); 427 jce->jce_func(sb, jce, error); 428 spin_lock(&sbi->s_md_lock); 429 } 430 spin_unlock(&sbi->s_md_lock); 431 } 432 433 /* Deal with the reporting of failure conditions on a filesystem such as 434 * inconsistencies detected or read IO failures. 435 * 436 * On ext2, we can store the error state of the filesystem in the 437 * superblock. That is not possible on ext4, because we may have other 438 * write ordering constraints on the superblock which prevent us from 439 * writing it out straight away; and given that the journal is about to 440 * be aborted, we can't rely on the current, or future, transactions to 441 * write out the superblock safely. 442 * 443 * We'll just use the jbd2_journal_abort() error code to record an error in 444 * the journal instead. On recovery, the journal will complain about 445 * that error until we've noted it down and cleared it. 446 */ 447 448 static void ext4_handle_error(struct super_block *sb) 449 { 450 if (test_opt(sb, WARN_ON_ERROR)) 451 WARN_ON_ONCE(1); 452 453 if (sb_rdonly(sb)) 454 return; 455 456 if (!test_opt(sb, ERRORS_CONT)) { 457 journal_t *journal = EXT4_SB(sb)->s_journal; 458 459 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 460 if (journal) 461 jbd2_journal_abort(journal, -EIO); 462 } 463 if (test_opt(sb, ERRORS_RO)) { 464 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 465 /* 466 * Make sure updated value of ->s_mount_flags will be visible 467 * before ->s_flags update 468 */ 469 smp_wmb(); 470 sb->s_flags |= SB_RDONLY; 471 } 472 if (test_opt(sb, ERRORS_PANIC)) { 473 if (EXT4_SB(sb)->s_journal && 474 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 475 return; 476 panic("EXT4-fs (device %s): panic forced after error\n", 477 sb->s_id); 478 } 479 } 480 481 #define ext4_error_ratelimit(sb) \ 482 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 483 "EXT4-fs error") 484 485 void __ext4_error(struct super_block *sb, const char *function, 486 unsigned int line, const char *fmt, ...) 487 { 488 struct va_format vaf; 489 va_list args; 490 491 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 492 return; 493 494 trace_ext4_error(sb, function, line); 495 if (ext4_error_ratelimit(sb)) { 496 va_start(args, fmt); 497 vaf.fmt = fmt; 498 vaf.va = &args; 499 printk(KERN_CRIT 500 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 501 sb->s_id, function, line, current->comm, &vaf); 502 va_end(args); 503 } 504 save_error_info(sb, function, line); 505 ext4_handle_error(sb); 506 } 507 508 void __ext4_error_inode(struct inode *inode, const char *function, 509 unsigned int line, ext4_fsblk_t block, 510 const char *fmt, ...) 511 { 512 va_list args; 513 struct va_format vaf; 514 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 515 516 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 517 return; 518 519 trace_ext4_error(inode->i_sb, function, line); 520 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 521 es->s_last_error_block = cpu_to_le64(block); 522 if (ext4_error_ratelimit(inode->i_sb)) { 523 va_start(args, fmt); 524 vaf.fmt = fmt; 525 vaf.va = &args; 526 if (block) 527 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 528 "inode #%lu: block %llu: comm %s: %pV\n", 529 inode->i_sb->s_id, function, line, inode->i_ino, 530 block, current->comm, &vaf); 531 else 532 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 533 "inode #%lu: comm %s: %pV\n", 534 inode->i_sb->s_id, function, line, inode->i_ino, 535 current->comm, &vaf); 536 va_end(args); 537 } 538 save_error_info(inode->i_sb, function, line); 539 ext4_handle_error(inode->i_sb); 540 } 541 542 void __ext4_error_file(struct file *file, const char *function, 543 unsigned int line, ext4_fsblk_t block, 544 const char *fmt, ...) 545 { 546 va_list args; 547 struct va_format vaf; 548 struct ext4_super_block *es; 549 struct inode *inode = file_inode(file); 550 char pathname[80], *path; 551 552 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 553 return; 554 555 trace_ext4_error(inode->i_sb, function, line); 556 es = EXT4_SB(inode->i_sb)->s_es; 557 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 558 if (ext4_error_ratelimit(inode->i_sb)) { 559 path = file_path(file, pathname, sizeof(pathname)); 560 if (IS_ERR(path)) 561 path = "(unknown)"; 562 va_start(args, fmt); 563 vaf.fmt = fmt; 564 vaf.va = &args; 565 if (block) 566 printk(KERN_CRIT 567 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 568 "block %llu: comm %s: path %s: %pV\n", 569 inode->i_sb->s_id, function, line, inode->i_ino, 570 block, current->comm, path, &vaf); 571 else 572 printk(KERN_CRIT 573 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 574 "comm %s: path %s: %pV\n", 575 inode->i_sb->s_id, function, line, inode->i_ino, 576 current->comm, path, &vaf); 577 va_end(args); 578 } 579 save_error_info(inode->i_sb, function, line); 580 ext4_handle_error(inode->i_sb); 581 } 582 583 const char *ext4_decode_error(struct super_block *sb, int errno, 584 char nbuf[16]) 585 { 586 char *errstr = NULL; 587 588 switch (errno) { 589 case -EFSCORRUPTED: 590 errstr = "Corrupt filesystem"; 591 break; 592 case -EFSBADCRC: 593 errstr = "Filesystem failed CRC"; 594 break; 595 case -EIO: 596 errstr = "IO failure"; 597 break; 598 case -ENOMEM: 599 errstr = "Out of memory"; 600 break; 601 case -EROFS: 602 if (!sb || (EXT4_SB(sb)->s_journal && 603 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 604 errstr = "Journal has aborted"; 605 else 606 errstr = "Readonly filesystem"; 607 break; 608 default: 609 /* If the caller passed in an extra buffer for unknown 610 * errors, textualise them now. Else we just return 611 * NULL. */ 612 if (nbuf) { 613 /* Check for truncated error codes... */ 614 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 615 errstr = nbuf; 616 } 617 break; 618 } 619 620 return errstr; 621 } 622 623 /* __ext4_std_error decodes expected errors from journaling functions 624 * automatically and invokes the appropriate error response. */ 625 626 void __ext4_std_error(struct super_block *sb, const char *function, 627 unsigned int line, int errno) 628 { 629 char nbuf[16]; 630 const char *errstr; 631 632 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 633 return; 634 635 /* Special case: if the error is EROFS, and we're not already 636 * inside a transaction, then there's really no point in logging 637 * an error. */ 638 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 639 return; 640 641 if (ext4_error_ratelimit(sb)) { 642 errstr = ext4_decode_error(sb, errno, nbuf); 643 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 644 sb->s_id, function, line, errstr); 645 } 646 647 save_error_info(sb, function, line); 648 ext4_handle_error(sb); 649 } 650 651 /* 652 * ext4_abort is a much stronger failure handler than ext4_error. The 653 * abort function may be used to deal with unrecoverable failures such 654 * as journal IO errors or ENOMEM at a critical moment in log management. 655 * 656 * We unconditionally force the filesystem into an ABORT|READONLY state, 657 * unless the error response on the fs has been set to panic in which 658 * case we take the easy way out and panic immediately. 659 */ 660 661 void __ext4_abort(struct super_block *sb, const char *function, 662 unsigned int line, const char *fmt, ...) 663 { 664 struct va_format vaf; 665 va_list args; 666 667 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 668 return; 669 670 save_error_info(sb, function, line); 671 va_start(args, fmt); 672 vaf.fmt = fmt; 673 vaf.va = &args; 674 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 675 sb->s_id, function, line, &vaf); 676 va_end(args); 677 678 if (sb_rdonly(sb) == 0) { 679 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 680 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 681 /* 682 * Make sure updated value of ->s_mount_flags will be visible 683 * before ->s_flags update 684 */ 685 smp_wmb(); 686 sb->s_flags |= SB_RDONLY; 687 if (EXT4_SB(sb)->s_journal) 688 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 689 save_error_info(sb, function, line); 690 } 691 if (test_opt(sb, ERRORS_PANIC)) { 692 if (EXT4_SB(sb)->s_journal && 693 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 694 return; 695 panic("EXT4-fs panic from previous error\n"); 696 } 697 } 698 699 void __ext4_msg(struct super_block *sb, 700 const char *prefix, const char *fmt, ...) 701 { 702 struct va_format vaf; 703 va_list args; 704 705 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 706 return; 707 708 va_start(args, fmt); 709 vaf.fmt = fmt; 710 vaf.va = &args; 711 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 712 va_end(args); 713 } 714 715 #define ext4_warning_ratelimit(sb) \ 716 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ 717 "EXT4-fs warning") 718 719 void __ext4_warning(struct super_block *sb, const char *function, 720 unsigned int line, const char *fmt, ...) 721 { 722 struct va_format vaf; 723 va_list args; 724 725 if (!ext4_warning_ratelimit(sb)) 726 return; 727 728 va_start(args, fmt); 729 vaf.fmt = fmt; 730 vaf.va = &args; 731 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 732 sb->s_id, function, line, &vaf); 733 va_end(args); 734 } 735 736 void __ext4_warning_inode(const struct inode *inode, const char *function, 737 unsigned int line, const char *fmt, ...) 738 { 739 struct va_format vaf; 740 va_list args; 741 742 if (!ext4_warning_ratelimit(inode->i_sb)) 743 return; 744 745 va_start(args, fmt); 746 vaf.fmt = fmt; 747 vaf.va = &args; 748 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 749 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 750 function, line, inode->i_ino, current->comm, &vaf); 751 va_end(args); 752 } 753 754 void __ext4_grp_locked_error(const char *function, unsigned int line, 755 struct super_block *sb, ext4_group_t grp, 756 unsigned long ino, ext4_fsblk_t block, 757 const char *fmt, ...) 758 __releases(bitlock) 759 __acquires(bitlock) 760 { 761 struct va_format vaf; 762 va_list args; 763 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 764 765 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 766 return; 767 768 trace_ext4_error(sb, function, line); 769 es->s_last_error_ino = cpu_to_le32(ino); 770 es->s_last_error_block = cpu_to_le64(block); 771 __save_error_info(sb, function, line); 772 773 if (ext4_error_ratelimit(sb)) { 774 va_start(args, fmt); 775 vaf.fmt = fmt; 776 vaf.va = &args; 777 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 778 sb->s_id, function, line, grp); 779 if (ino) 780 printk(KERN_CONT "inode %lu: ", ino); 781 if (block) 782 printk(KERN_CONT "block %llu:", 783 (unsigned long long) block); 784 printk(KERN_CONT "%pV\n", &vaf); 785 va_end(args); 786 } 787 788 if (test_opt(sb, WARN_ON_ERROR)) 789 WARN_ON_ONCE(1); 790 791 if (test_opt(sb, ERRORS_CONT)) { 792 ext4_commit_super(sb, 0); 793 return; 794 } 795 796 ext4_unlock_group(sb, grp); 797 ext4_commit_super(sb, 1); 798 ext4_handle_error(sb); 799 /* 800 * We only get here in the ERRORS_RO case; relocking the group 801 * may be dangerous, but nothing bad will happen since the 802 * filesystem will have already been marked read/only and the 803 * journal has been aborted. We return 1 as a hint to callers 804 * who might what to use the return value from 805 * ext4_grp_locked_error() to distinguish between the 806 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 807 * aggressively from the ext4 function in question, with a 808 * more appropriate error code. 809 */ 810 ext4_lock_group(sb, grp); 811 return; 812 } 813 814 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 815 ext4_group_t group, 816 unsigned int flags) 817 { 818 struct ext4_sb_info *sbi = EXT4_SB(sb); 819 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 820 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 821 int ret; 822 823 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 824 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 825 &grp->bb_state); 826 if (!ret) 827 percpu_counter_sub(&sbi->s_freeclusters_counter, 828 grp->bb_free); 829 } 830 831 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 832 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 833 &grp->bb_state); 834 if (!ret && gdp) { 835 int count; 836 837 count = ext4_free_inodes_count(sb, gdp); 838 percpu_counter_sub(&sbi->s_freeinodes_counter, 839 count); 840 } 841 } 842 } 843 844 void ext4_update_dynamic_rev(struct super_block *sb) 845 { 846 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 847 848 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 849 return; 850 851 ext4_warning(sb, 852 "updating to rev %d because of new feature flag, " 853 "running e2fsck is recommended", 854 EXT4_DYNAMIC_REV); 855 856 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 857 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 858 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 859 /* leave es->s_feature_*compat flags alone */ 860 /* es->s_uuid will be set by e2fsck if empty */ 861 862 /* 863 * The rest of the superblock fields should be zero, and if not it 864 * means they are likely already in use, so leave them alone. We 865 * can leave it up to e2fsck to clean up any inconsistencies there. 866 */ 867 } 868 869 /* 870 * Open the external journal device 871 */ 872 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 873 { 874 struct block_device *bdev; 875 char b[BDEVNAME_SIZE]; 876 877 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 878 if (IS_ERR(bdev)) 879 goto fail; 880 return bdev; 881 882 fail: 883 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 884 __bdevname(dev, b), PTR_ERR(bdev)); 885 return NULL; 886 } 887 888 /* 889 * Release the journal device 890 */ 891 static void ext4_blkdev_put(struct block_device *bdev) 892 { 893 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 894 } 895 896 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 897 { 898 struct block_device *bdev; 899 bdev = sbi->journal_bdev; 900 if (bdev) { 901 ext4_blkdev_put(bdev); 902 sbi->journal_bdev = NULL; 903 } 904 } 905 906 static inline struct inode *orphan_list_entry(struct list_head *l) 907 { 908 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 909 } 910 911 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 912 { 913 struct list_head *l; 914 915 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 916 le32_to_cpu(sbi->s_es->s_last_orphan)); 917 918 printk(KERN_ERR "sb_info orphan list:\n"); 919 list_for_each(l, &sbi->s_orphan) { 920 struct inode *inode = orphan_list_entry(l); 921 printk(KERN_ERR " " 922 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 923 inode->i_sb->s_id, inode->i_ino, inode, 924 inode->i_mode, inode->i_nlink, 925 NEXT_ORPHAN(inode)); 926 } 927 } 928 929 #ifdef CONFIG_QUOTA 930 static int ext4_quota_off(struct super_block *sb, int type); 931 932 static inline void ext4_quota_off_umount(struct super_block *sb) 933 { 934 int type; 935 936 /* Use our quota_off function to clear inode flags etc. */ 937 for (type = 0; type < EXT4_MAXQUOTAS; type++) 938 ext4_quota_off(sb, type); 939 } 940 941 /* 942 * This is a helper function which is used in the mount/remount 943 * codepaths (which holds s_umount) to fetch the quota file name. 944 */ 945 static inline char *get_qf_name(struct super_block *sb, 946 struct ext4_sb_info *sbi, 947 int type) 948 { 949 return rcu_dereference_protected(sbi->s_qf_names[type], 950 lockdep_is_held(&sb->s_umount)); 951 } 952 #else 953 static inline void ext4_quota_off_umount(struct super_block *sb) 954 { 955 } 956 #endif 957 958 static void ext4_put_super(struct super_block *sb) 959 { 960 struct ext4_sb_info *sbi = EXT4_SB(sb); 961 struct ext4_super_block *es = sbi->s_es; 962 int aborted = 0; 963 int i, err; 964 965 ext4_unregister_li_request(sb); 966 ext4_quota_off_umount(sb); 967 968 destroy_workqueue(sbi->rsv_conversion_wq); 969 970 if (sbi->s_journal) { 971 aborted = is_journal_aborted(sbi->s_journal); 972 err = jbd2_journal_destroy(sbi->s_journal); 973 sbi->s_journal = NULL; 974 if ((err < 0) && !aborted) 975 ext4_abort(sb, "Couldn't clean up the journal"); 976 } 977 978 ext4_unregister_sysfs(sb); 979 ext4_es_unregister_shrinker(sbi); 980 del_timer_sync(&sbi->s_err_report); 981 ext4_release_system_zone(sb); 982 ext4_mb_release(sb); 983 ext4_ext_release(sb); 984 985 if (!sb_rdonly(sb) && !aborted) { 986 ext4_clear_feature_journal_needs_recovery(sb); 987 es->s_state = cpu_to_le16(sbi->s_mount_state); 988 } 989 if (!sb_rdonly(sb)) 990 ext4_commit_super(sb, 1); 991 992 for (i = 0; i < sbi->s_gdb_count; i++) 993 brelse(sbi->s_group_desc[i]); 994 kvfree(sbi->s_group_desc); 995 kvfree(sbi->s_flex_groups); 996 percpu_counter_destroy(&sbi->s_freeclusters_counter); 997 percpu_counter_destroy(&sbi->s_freeinodes_counter); 998 percpu_counter_destroy(&sbi->s_dirs_counter); 999 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1000 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 1001 #ifdef CONFIG_QUOTA 1002 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1003 kfree(get_qf_name(sb, sbi, i)); 1004 #endif 1005 1006 /* Debugging code just in case the in-memory inode orphan list 1007 * isn't empty. The on-disk one can be non-empty if we've 1008 * detected an error and taken the fs readonly, but the 1009 * in-memory list had better be clean by this point. */ 1010 if (!list_empty(&sbi->s_orphan)) 1011 dump_orphan_list(sb, sbi); 1012 J_ASSERT(list_empty(&sbi->s_orphan)); 1013 1014 sync_blockdev(sb->s_bdev); 1015 invalidate_bdev(sb->s_bdev); 1016 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 1017 /* 1018 * Invalidate the journal device's buffers. We don't want them 1019 * floating about in memory - the physical journal device may 1020 * hotswapped, and it breaks the `ro-after' testing code. 1021 */ 1022 sync_blockdev(sbi->journal_bdev); 1023 invalidate_bdev(sbi->journal_bdev); 1024 ext4_blkdev_remove(sbi); 1025 } 1026 1027 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1028 sbi->s_ea_inode_cache = NULL; 1029 1030 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1031 sbi->s_ea_block_cache = NULL; 1032 1033 if (sbi->s_mmp_tsk) 1034 kthread_stop(sbi->s_mmp_tsk); 1035 brelse(sbi->s_sbh); 1036 sb->s_fs_info = NULL; 1037 /* 1038 * Now that we are completely done shutting down the 1039 * superblock, we need to actually destroy the kobject. 1040 */ 1041 kobject_put(&sbi->s_kobj); 1042 wait_for_completion(&sbi->s_kobj_unregister); 1043 if (sbi->s_chksum_driver) 1044 crypto_free_shash(sbi->s_chksum_driver); 1045 kfree(sbi->s_blockgroup_lock); 1046 fs_put_dax(sbi->s_daxdev); 1047 kfree(sbi); 1048 } 1049 1050 static struct kmem_cache *ext4_inode_cachep; 1051 1052 /* 1053 * Called inside transaction, so use GFP_NOFS 1054 */ 1055 static struct inode *ext4_alloc_inode(struct super_block *sb) 1056 { 1057 struct ext4_inode_info *ei; 1058 1059 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1060 if (!ei) 1061 return NULL; 1062 1063 inode_set_iversion(&ei->vfs_inode, 1); 1064 spin_lock_init(&ei->i_raw_lock); 1065 INIT_LIST_HEAD(&ei->i_prealloc_list); 1066 spin_lock_init(&ei->i_prealloc_lock); 1067 ext4_es_init_tree(&ei->i_es_tree); 1068 rwlock_init(&ei->i_es_lock); 1069 INIT_LIST_HEAD(&ei->i_es_list); 1070 ei->i_es_all_nr = 0; 1071 ei->i_es_shk_nr = 0; 1072 ei->i_es_shrink_lblk = 0; 1073 ei->i_reserved_data_blocks = 0; 1074 ei->i_da_metadata_calc_len = 0; 1075 ei->i_da_metadata_calc_last_lblock = 0; 1076 spin_lock_init(&(ei->i_block_reservation_lock)); 1077 ext4_init_pending_tree(&ei->i_pending_tree); 1078 #ifdef CONFIG_QUOTA 1079 ei->i_reserved_quota = 0; 1080 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1081 #endif 1082 ei->jinode = NULL; 1083 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1084 spin_lock_init(&ei->i_completed_io_lock); 1085 ei->i_sync_tid = 0; 1086 ei->i_datasync_tid = 0; 1087 atomic_set(&ei->i_unwritten, 0); 1088 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1089 return &ei->vfs_inode; 1090 } 1091 1092 static int ext4_drop_inode(struct inode *inode) 1093 { 1094 int drop = generic_drop_inode(inode); 1095 1096 trace_ext4_drop_inode(inode, drop); 1097 return drop; 1098 } 1099 1100 static void ext4_i_callback(struct rcu_head *head) 1101 { 1102 struct inode *inode = container_of(head, struct inode, i_rcu); 1103 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1104 } 1105 1106 static void ext4_destroy_inode(struct inode *inode) 1107 { 1108 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1109 ext4_msg(inode->i_sb, KERN_ERR, 1110 "Inode %lu (%p): orphan list check failed!", 1111 inode->i_ino, EXT4_I(inode)); 1112 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1113 EXT4_I(inode), sizeof(struct ext4_inode_info), 1114 true); 1115 dump_stack(); 1116 } 1117 call_rcu(&inode->i_rcu, ext4_i_callback); 1118 } 1119 1120 static void init_once(void *foo) 1121 { 1122 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1123 1124 INIT_LIST_HEAD(&ei->i_orphan); 1125 init_rwsem(&ei->xattr_sem); 1126 init_rwsem(&ei->i_data_sem); 1127 init_rwsem(&ei->i_mmap_sem); 1128 inode_init_once(&ei->vfs_inode); 1129 } 1130 1131 static int __init init_inodecache(void) 1132 { 1133 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1134 sizeof(struct ext4_inode_info), 0, 1135 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1136 SLAB_ACCOUNT), 1137 offsetof(struct ext4_inode_info, i_data), 1138 sizeof_field(struct ext4_inode_info, i_data), 1139 init_once); 1140 if (ext4_inode_cachep == NULL) 1141 return -ENOMEM; 1142 return 0; 1143 } 1144 1145 static void destroy_inodecache(void) 1146 { 1147 /* 1148 * Make sure all delayed rcu free inodes are flushed before we 1149 * destroy cache. 1150 */ 1151 rcu_barrier(); 1152 kmem_cache_destroy(ext4_inode_cachep); 1153 } 1154 1155 void ext4_clear_inode(struct inode *inode) 1156 { 1157 invalidate_inode_buffers(inode); 1158 clear_inode(inode); 1159 dquot_drop(inode); 1160 ext4_discard_preallocations(inode); 1161 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1162 if (EXT4_I(inode)->jinode) { 1163 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1164 EXT4_I(inode)->jinode); 1165 jbd2_free_inode(EXT4_I(inode)->jinode); 1166 EXT4_I(inode)->jinode = NULL; 1167 } 1168 fscrypt_put_encryption_info(inode); 1169 } 1170 1171 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1172 u64 ino, u32 generation) 1173 { 1174 struct inode *inode; 1175 1176 /* 1177 * Currently we don't know the generation for parent directory, so 1178 * a generation of 0 means "accept any" 1179 */ 1180 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1181 if (IS_ERR(inode)) 1182 return ERR_CAST(inode); 1183 if (generation && inode->i_generation != generation) { 1184 iput(inode); 1185 return ERR_PTR(-ESTALE); 1186 } 1187 1188 return inode; 1189 } 1190 1191 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1192 int fh_len, int fh_type) 1193 { 1194 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1195 ext4_nfs_get_inode); 1196 } 1197 1198 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1199 int fh_len, int fh_type) 1200 { 1201 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1202 ext4_nfs_get_inode); 1203 } 1204 1205 /* 1206 * Try to release metadata pages (indirect blocks, directories) which are 1207 * mapped via the block device. Since these pages could have journal heads 1208 * which would prevent try_to_free_buffers() from freeing them, we must use 1209 * jbd2 layer's try_to_free_buffers() function to release them. 1210 */ 1211 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1212 gfp_t wait) 1213 { 1214 journal_t *journal = EXT4_SB(sb)->s_journal; 1215 1216 WARN_ON(PageChecked(page)); 1217 if (!page_has_buffers(page)) 1218 return 0; 1219 if (journal) 1220 return jbd2_journal_try_to_free_buffers(journal, page, 1221 wait & ~__GFP_DIRECT_RECLAIM); 1222 return try_to_free_buffers(page); 1223 } 1224 1225 #ifdef CONFIG_EXT4_FS_ENCRYPTION 1226 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1227 { 1228 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1229 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1230 } 1231 1232 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1233 void *fs_data) 1234 { 1235 handle_t *handle = fs_data; 1236 int res, res2, credits, retries = 0; 1237 1238 /* 1239 * Encrypting the root directory is not allowed because e2fsck expects 1240 * lost+found to exist and be unencrypted, and encrypting the root 1241 * directory would imply encrypting the lost+found directory as well as 1242 * the filename "lost+found" itself. 1243 */ 1244 if (inode->i_ino == EXT4_ROOT_INO) 1245 return -EPERM; 1246 1247 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 1248 return -EINVAL; 1249 1250 res = ext4_convert_inline_data(inode); 1251 if (res) 1252 return res; 1253 1254 /* 1255 * If a journal handle was specified, then the encryption context is 1256 * being set on a new inode via inheritance and is part of a larger 1257 * transaction to create the inode. Otherwise the encryption context is 1258 * being set on an existing inode in its own transaction. Only in the 1259 * latter case should the "retry on ENOSPC" logic be used. 1260 */ 1261 1262 if (handle) { 1263 res = ext4_xattr_set_handle(handle, inode, 1264 EXT4_XATTR_INDEX_ENCRYPTION, 1265 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1266 ctx, len, 0); 1267 if (!res) { 1268 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1269 ext4_clear_inode_state(inode, 1270 EXT4_STATE_MAY_INLINE_DATA); 1271 /* 1272 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1273 * S_DAX may be disabled 1274 */ 1275 ext4_set_inode_flags(inode); 1276 } 1277 return res; 1278 } 1279 1280 res = dquot_initialize(inode); 1281 if (res) 1282 return res; 1283 retry: 1284 res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1285 &credits); 1286 if (res) 1287 return res; 1288 1289 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1290 if (IS_ERR(handle)) 1291 return PTR_ERR(handle); 1292 1293 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 1294 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1295 ctx, len, 0); 1296 if (!res) { 1297 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 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 res = ext4_mark_inode_dirty(handle, inode); 1304 if (res) 1305 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1306 } 1307 res2 = ext4_journal_stop(handle); 1308 1309 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 1310 goto retry; 1311 if (!res) 1312 res = res2; 1313 return res; 1314 } 1315 1316 static bool ext4_dummy_context(struct inode *inode) 1317 { 1318 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); 1319 } 1320 1321 static const struct fscrypt_operations ext4_cryptops = { 1322 .key_prefix = "ext4:", 1323 .get_context = ext4_get_context, 1324 .set_context = ext4_set_context, 1325 .dummy_context = ext4_dummy_context, 1326 .empty_dir = ext4_empty_dir, 1327 .max_namelen = EXT4_NAME_LEN, 1328 }; 1329 #endif 1330 1331 #ifdef CONFIG_QUOTA 1332 static const char * const quotatypes[] = INITQFNAMES; 1333 #define QTYPE2NAME(t) (quotatypes[t]) 1334 1335 static int ext4_write_dquot(struct dquot *dquot); 1336 static int ext4_acquire_dquot(struct dquot *dquot); 1337 static int ext4_release_dquot(struct dquot *dquot); 1338 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1339 static int ext4_write_info(struct super_block *sb, int type); 1340 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1341 const struct path *path); 1342 static int ext4_quota_on_mount(struct super_block *sb, int type); 1343 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1344 size_t len, loff_t off); 1345 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1346 const char *data, size_t len, loff_t off); 1347 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1348 unsigned int flags); 1349 static int ext4_enable_quotas(struct super_block *sb); 1350 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); 1351 1352 static struct dquot **ext4_get_dquots(struct inode *inode) 1353 { 1354 return EXT4_I(inode)->i_dquot; 1355 } 1356 1357 static const struct dquot_operations ext4_quota_operations = { 1358 .get_reserved_space = ext4_get_reserved_space, 1359 .write_dquot = ext4_write_dquot, 1360 .acquire_dquot = ext4_acquire_dquot, 1361 .release_dquot = ext4_release_dquot, 1362 .mark_dirty = ext4_mark_dquot_dirty, 1363 .write_info = ext4_write_info, 1364 .alloc_dquot = dquot_alloc, 1365 .destroy_dquot = dquot_destroy, 1366 .get_projid = ext4_get_projid, 1367 .get_inode_usage = ext4_get_inode_usage, 1368 .get_next_id = ext4_get_next_id, 1369 }; 1370 1371 static const struct quotactl_ops ext4_qctl_operations = { 1372 .quota_on = ext4_quota_on, 1373 .quota_off = ext4_quota_off, 1374 .quota_sync = dquot_quota_sync, 1375 .get_state = dquot_get_state, 1376 .set_info = dquot_set_dqinfo, 1377 .get_dqblk = dquot_get_dqblk, 1378 .set_dqblk = dquot_set_dqblk, 1379 .get_nextdqblk = dquot_get_next_dqblk, 1380 }; 1381 #endif 1382 1383 static const struct super_operations ext4_sops = { 1384 .alloc_inode = ext4_alloc_inode, 1385 .destroy_inode = ext4_destroy_inode, 1386 .write_inode = ext4_write_inode, 1387 .dirty_inode = ext4_dirty_inode, 1388 .drop_inode = ext4_drop_inode, 1389 .evict_inode = ext4_evict_inode, 1390 .put_super = ext4_put_super, 1391 .sync_fs = ext4_sync_fs, 1392 .freeze_fs = ext4_freeze, 1393 .unfreeze_fs = ext4_unfreeze, 1394 .statfs = ext4_statfs, 1395 .remount_fs = ext4_remount, 1396 .show_options = ext4_show_options, 1397 #ifdef CONFIG_QUOTA 1398 .quota_read = ext4_quota_read, 1399 .quota_write = ext4_quota_write, 1400 .get_dquots = ext4_get_dquots, 1401 #endif 1402 .bdev_try_to_free_page = bdev_try_to_free_page, 1403 }; 1404 1405 static const struct export_operations ext4_export_ops = { 1406 .fh_to_dentry = ext4_fh_to_dentry, 1407 .fh_to_parent = ext4_fh_to_parent, 1408 .get_parent = ext4_get_parent, 1409 }; 1410 1411 enum { 1412 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1413 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1414 Opt_nouid32, Opt_debug, Opt_removed, 1415 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1416 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1417 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1418 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1419 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1420 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1421 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1422 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1423 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1424 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, 1425 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1426 Opt_nowarn_on_error, Opt_mblk_io_submit, 1427 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 1428 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1429 Opt_inode_readahead_blks, Opt_journal_ioprio, 1430 Opt_dioread_nolock, Opt_dioread_lock, 1431 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1432 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1433 }; 1434 1435 static const match_table_t tokens = { 1436 {Opt_bsd_df, "bsddf"}, 1437 {Opt_minix_df, "minixdf"}, 1438 {Opt_grpid, "grpid"}, 1439 {Opt_grpid, "bsdgroups"}, 1440 {Opt_nogrpid, "nogrpid"}, 1441 {Opt_nogrpid, "sysvgroups"}, 1442 {Opt_resgid, "resgid=%u"}, 1443 {Opt_resuid, "resuid=%u"}, 1444 {Opt_sb, "sb=%u"}, 1445 {Opt_err_cont, "errors=continue"}, 1446 {Opt_err_panic, "errors=panic"}, 1447 {Opt_err_ro, "errors=remount-ro"}, 1448 {Opt_nouid32, "nouid32"}, 1449 {Opt_debug, "debug"}, 1450 {Opt_removed, "oldalloc"}, 1451 {Opt_removed, "orlov"}, 1452 {Opt_user_xattr, "user_xattr"}, 1453 {Opt_nouser_xattr, "nouser_xattr"}, 1454 {Opt_acl, "acl"}, 1455 {Opt_noacl, "noacl"}, 1456 {Opt_noload, "norecovery"}, 1457 {Opt_noload, "noload"}, 1458 {Opt_removed, "nobh"}, 1459 {Opt_removed, "bh"}, 1460 {Opt_commit, "commit=%u"}, 1461 {Opt_min_batch_time, "min_batch_time=%u"}, 1462 {Opt_max_batch_time, "max_batch_time=%u"}, 1463 {Opt_journal_dev, "journal_dev=%u"}, 1464 {Opt_journal_path, "journal_path=%s"}, 1465 {Opt_journal_checksum, "journal_checksum"}, 1466 {Opt_nojournal_checksum, "nojournal_checksum"}, 1467 {Opt_journal_async_commit, "journal_async_commit"}, 1468 {Opt_abort, "abort"}, 1469 {Opt_data_journal, "data=journal"}, 1470 {Opt_data_ordered, "data=ordered"}, 1471 {Opt_data_writeback, "data=writeback"}, 1472 {Opt_data_err_abort, "data_err=abort"}, 1473 {Opt_data_err_ignore, "data_err=ignore"}, 1474 {Opt_offusrjquota, "usrjquota="}, 1475 {Opt_usrjquota, "usrjquota=%s"}, 1476 {Opt_offgrpjquota, "grpjquota="}, 1477 {Opt_grpjquota, "grpjquota=%s"}, 1478 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1479 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1480 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1481 {Opt_grpquota, "grpquota"}, 1482 {Opt_noquota, "noquota"}, 1483 {Opt_quota, "quota"}, 1484 {Opt_usrquota, "usrquota"}, 1485 {Opt_prjquota, "prjquota"}, 1486 {Opt_barrier, "barrier=%u"}, 1487 {Opt_barrier, "barrier"}, 1488 {Opt_nobarrier, "nobarrier"}, 1489 {Opt_i_version, "i_version"}, 1490 {Opt_dax, "dax"}, 1491 {Opt_stripe, "stripe=%u"}, 1492 {Opt_delalloc, "delalloc"}, 1493 {Opt_warn_on_error, "warn_on_error"}, 1494 {Opt_nowarn_on_error, "nowarn_on_error"}, 1495 {Opt_lazytime, "lazytime"}, 1496 {Opt_nolazytime, "nolazytime"}, 1497 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1498 {Opt_nodelalloc, "nodelalloc"}, 1499 {Opt_removed, "mblk_io_submit"}, 1500 {Opt_removed, "nomblk_io_submit"}, 1501 {Opt_block_validity, "block_validity"}, 1502 {Opt_noblock_validity, "noblock_validity"}, 1503 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1504 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1505 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1506 {Opt_auto_da_alloc, "auto_da_alloc"}, 1507 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1508 {Opt_dioread_nolock, "dioread_nolock"}, 1509 {Opt_dioread_lock, "dioread_lock"}, 1510 {Opt_discard, "discard"}, 1511 {Opt_nodiscard, "nodiscard"}, 1512 {Opt_init_itable, "init_itable=%u"}, 1513 {Opt_init_itable, "init_itable"}, 1514 {Opt_noinit_itable, "noinit_itable"}, 1515 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1516 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1517 {Opt_nombcache, "nombcache"}, 1518 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1519 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1520 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1521 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1522 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1523 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1524 {Opt_err, NULL}, 1525 }; 1526 1527 static ext4_fsblk_t get_sb_block(void **data) 1528 { 1529 ext4_fsblk_t sb_block; 1530 char *options = (char *) *data; 1531 1532 if (!options || strncmp(options, "sb=", 3) != 0) 1533 return 1; /* Default location */ 1534 1535 options += 3; 1536 /* TODO: use simple_strtoll with >32bit ext4 */ 1537 sb_block = simple_strtoul(options, &options, 0); 1538 if (*options && *options != ',') { 1539 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1540 (char *) *data); 1541 return 1; 1542 } 1543 if (*options == ',') 1544 options++; 1545 *data = (void *) options; 1546 1547 return sb_block; 1548 } 1549 1550 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1551 static const char deprecated_msg[] = 1552 "Mount option \"%s\" will be removed by %s\n" 1553 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1554 1555 #ifdef CONFIG_QUOTA 1556 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1557 { 1558 struct ext4_sb_info *sbi = EXT4_SB(sb); 1559 char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 1560 int ret = -1; 1561 1562 if (sb_any_quota_loaded(sb) && !old_qname) { 1563 ext4_msg(sb, KERN_ERR, 1564 "Cannot change journaled " 1565 "quota options when quota turned on"); 1566 return -1; 1567 } 1568 if (ext4_has_feature_quota(sb)) { 1569 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1570 "ignored when QUOTA feature is enabled"); 1571 return 1; 1572 } 1573 qname = match_strdup(args); 1574 if (!qname) { 1575 ext4_msg(sb, KERN_ERR, 1576 "Not enough memory for storing quotafile name"); 1577 return -1; 1578 } 1579 if (old_qname) { 1580 if (strcmp(old_qname, qname) == 0) 1581 ret = 1; 1582 else 1583 ext4_msg(sb, KERN_ERR, 1584 "%s quota file already specified", 1585 QTYPE2NAME(qtype)); 1586 goto errout; 1587 } 1588 if (strchr(qname, '/')) { 1589 ext4_msg(sb, KERN_ERR, 1590 "quotafile must be on filesystem root"); 1591 goto errout; 1592 } 1593 rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1594 set_opt(sb, QUOTA); 1595 return 1; 1596 errout: 1597 kfree(qname); 1598 return ret; 1599 } 1600 1601 static int clear_qf_name(struct super_block *sb, int qtype) 1602 { 1603 1604 struct ext4_sb_info *sbi = EXT4_SB(sb); 1605 char *old_qname = get_qf_name(sb, sbi, qtype); 1606 1607 if (sb_any_quota_loaded(sb) && old_qname) { 1608 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1609 " when quota turned on"); 1610 return -1; 1611 } 1612 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 1613 synchronize_rcu(); 1614 kfree(old_qname); 1615 return 1; 1616 } 1617 #endif 1618 1619 #define MOPT_SET 0x0001 1620 #define MOPT_CLEAR 0x0002 1621 #define MOPT_NOSUPPORT 0x0004 1622 #define MOPT_EXPLICIT 0x0008 1623 #define MOPT_CLEAR_ERR 0x0010 1624 #define MOPT_GTE0 0x0020 1625 #ifdef CONFIG_QUOTA 1626 #define MOPT_Q 0 1627 #define MOPT_QFMT 0x0040 1628 #else 1629 #define MOPT_Q MOPT_NOSUPPORT 1630 #define MOPT_QFMT MOPT_NOSUPPORT 1631 #endif 1632 #define MOPT_DATAJ 0x0080 1633 #define MOPT_NO_EXT2 0x0100 1634 #define MOPT_NO_EXT3 0x0200 1635 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1636 #define MOPT_STRING 0x0400 1637 1638 static const struct mount_opts { 1639 int token; 1640 int mount_opt; 1641 int flags; 1642 } ext4_mount_opts[] = { 1643 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1644 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1645 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1646 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1647 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1648 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1649 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1650 MOPT_EXT4_ONLY | MOPT_SET}, 1651 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1652 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1653 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1654 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1655 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1656 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1657 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1658 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1659 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1660 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1661 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1662 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1663 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1664 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1665 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1666 EXT4_MOUNT_JOURNAL_CHECKSUM), 1667 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1668 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1669 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1670 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1671 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1672 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1673 MOPT_NO_EXT2}, 1674 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1675 MOPT_NO_EXT2}, 1676 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1677 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1678 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1679 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1680 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1681 {Opt_commit, 0, MOPT_GTE0}, 1682 {Opt_max_batch_time, 0, MOPT_GTE0}, 1683 {Opt_min_batch_time, 0, MOPT_GTE0}, 1684 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1685 {Opt_init_itable, 0, MOPT_GTE0}, 1686 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 1687 {Opt_stripe, 0, MOPT_GTE0}, 1688 {Opt_resuid, 0, MOPT_GTE0}, 1689 {Opt_resgid, 0, MOPT_GTE0}, 1690 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1691 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1692 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1693 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1694 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1695 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1696 MOPT_NO_EXT2 | MOPT_DATAJ}, 1697 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1698 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1699 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1700 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1701 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1702 #else 1703 {Opt_acl, 0, MOPT_NOSUPPORT}, 1704 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1705 #endif 1706 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1707 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1708 {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 1709 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1710 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1711 MOPT_SET | MOPT_Q}, 1712 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1713 MOPT_SET | MOPT_Q}, 1714 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1715 MOPT_SET | MOPT_Q}, 1716 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1717 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1718 MOPT_CLEAR | MOPT_Q}, 1719 {Opt_usrjquota, 0, MOPT_Q}, 1720 {Opt_grpjquota, 0, MOPT_Q}, 1721 {Opt_offusrjquota, 0, MOPT_Q}, 1722 {Opt_offgrpjquota, 0, MOPT_Q}, 1723 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1724 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1725 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1726 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1727 {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1728 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1729 {Opt_err, 0, 0} 1730 }; 1731 1732 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1733 substring_t *args, unsigned long *journal_devnum, 1734 unsigned int *journal_ioprio, int is_remount) 1735 { 1736 struct ext4_sb_info *sbi = EXT4_SB(sb); 1737 const struct mount_opts *m; 1738 kuid_t uid; 1739 kgid_t gid; 1740 int arg = 0; 1741 1742 #ifdef CONFIG_QUOTA 1743 if (token == Opt_usrjquota) 1744 return set_qf_name(sb, USRQUOTA, &args[0]); 1745 else if (token == Opt_grpjquota) 1746 return set_qf_name(sb, GRPQUOTA, &args[0]); 1747 else if (token == Opt_offusrjquota) 1748 return clear_qf_name(sb, USRQUOTA); 1749 else if (token == Opt_offgrpjquota) 1750 return clear_qf_name(sb, GRPQUOTA); 1751 #endif 1752 switch (token) { 1753 case Opt_noacl: 1754 case Opt_nouser_xattr: 1755 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1756 break; 1757 case Opt_sb: 1758 return 1; /* handled by get_sb_block() */ 1759 case Opt_removed: 1760 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1761 return 1; 1762 case Opt_abort: 1763 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1764 return 1; 1765 case Opt_i_version: 1766 sb->s_flags |= SB_I_VERSION; 1767 return 1; 1768 case Opt_lazytime: 1769 sb->s_flags |= SB_LAZYTIME; 1770 return 1; 1771 case Opt_nolazytime: 1772 sb->s_flags &= ~SB_LAZYTIME; 1773 return 1; 1774 } 1775 1776 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1777 if (token == m->token) 1778 break; 1779 1780 if (m->token == Opt_err) { 1781 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1782 "or missing value", opt); 1783 return -1; 1784 } 1785 1786 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1787 ext4_msg(sb, KERN_ERR, 1788 "Mount option \"%s\" incompatible with ext2", opt); 1789 return -1; 1790 } 1791 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1792 ext4_msg(sb, KERN_ERR, 1793 "Mount option \"%s\" incompatible with ext3", opt); 1794 return -1; 1795 } 1796 1797 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1798 return -1; 1799 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1800 return -1; 1801 if (m->flags & MOPT_EXPLICIT) { 1802 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 1803 set_opt2(sb, EXPLICIT_DELALLOC); 1804 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 1805 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1806 } else 1807 return -1; 1808 } 1809 if (m->flags & MOPT_CLEAR_ERR) 1810 clear_opt(sb, ERRORS_MASK); 1811 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1812 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1813 "options when quota turned on"); 1814 return -1; 1815 } 1816 1817 if (m->flags & MOPT_NOSUPPORT) { 1818 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1819 } else if (token == Opt_commit) { 1820 if (arg == 0) 1821 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1822 sbi->s_commit_interval = HZ * arg; 1823 } else if (token == Opt_debug_want_extra_isize) { 1824 sbi->s_want_extra_isize = arg; 1825 } else if (token == Opt_max_batch_time) { 1826 sbi->s_max_batch_time = arg; 1827 } else if (token == Opt_min_batch_time) { 1828 sbi->s_min_batch_time = arg; 1829 } else if (token == Opt_inode_readahead_blks) { 1830 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1831 ext4_msg(sb, KERN_ERR, 1832 "EXT4-fs: inode_readahead_blks must be " 1833 "0 or a power of 2 smaller than 2^31"); 1834 return -1; 1835 } 1836 sbi->s_inode_readahead_blks = arg; 1837 } else if (token == Opt_init_itable) { 1838 set_opt(sb, INIT_INODE_TABLE); 1839 if (!args->from) 1840 arg = EXT4_DEF_LI_WAIT_MULT; 1841 sbi->s_li_wait_mult = arg; 1842 } else if (token == Opt_max_dir_size_kb) { 1843 sbi->s_max_dir_size_kb = arg; 1844 } else if (token == Opt_stripe) { 1845 sbi->s_stripe = arg; 1846 } else if (token == Opt_resuid) { 1847 uid = make_kuid(current_user_ns(), arg); 1848 if (!uid_valid(uid)) { 1849 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1850 return -1; 1851 } 1852 sbi->s_resuid = uid; 1853 } else if (token == Opt_resgid) { 1854 gid = make_kgid(current_user_ns(), arg); 1855 if (!gid_valid(gid)) { 1856 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1857 return -1; 1858 } 1859 sbi->s_resgid = gid; 1860 } else if (token == Opt_journal_dev) { 1861 if (is_remount) { 1862 ext4_msg(sb, KERN_ERR, 1863 "Cannot specify journal on remount"); 1864 return -1; 1865 } 1866 *journal_devnum = arg; 1867 } else if (token == Opt_journal_path) { 1868 char *journal_path; 1869 struct inode *journal_inode; 1870 struct path path; 1871 int error; 1872 1873 if (is_remount) { 1874 ext4_msg(sb, KERN_ERR, 1875 "Cannot specify journal on remount"); 1876 return -1; 1877 } 1878 journal_path = match_strdup(&args[0]); 1879 if (!journal_path) { 1880 ext4_msg(sb, KERN_ERR, "error: could not dup " 1881 "journal device string"); 1882 return -1; 1883 } 1884 1885 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1886 if (error) { 1887 ext4_msg(sb, KERN_ERR, "error: could not find " 1888 "journal device path: error %d", error); 1889 kfree(journal_path); 1890 return -1; 1891 } 1892 1893 journal_inode = d_inode(path.dentry); 1894 if (!S_ISBLK(journal_inode->i_mode)) { 1895 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1896 "is not a block device", journal_path); 1897 path_put(&path); 1898 kfree(journal_path); 1899 return -1; 1900 } 1901 1902 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1903 path_put(&path); 1904 kfree(journal_path); 1905 } else if (token == Opt_journal_ioprio) { 1906 if (arg > 7) { 1907 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1908 " (must be 0-7)"); 1909 return -1; 1910 } 1911 *journal_ioprio = 1912 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1913 } else if (token == Opt_test_dummy_encryption) { 1914 #ifdef CONFIG_EXT4_FS_ENCRYPTION 1915 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 1916 ext4_msg(sb, KERN_WARNING, 1917 "Test dummy encryption mode enabled"); 1918 #else 1919 ext4_msg(sb, KERN_WARNING, 1920 "Test dummy encryption mount option ignored"); 1921 #endif 1922 } else if (m->flags & MOPT_DATAJ) { 1923 if (is_remount) { 1924 if (!sbi->s_journal) 1925 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 1926 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 1927 ext4_msg(sb, KERN_ERR, 1928 "Cannot change data mode on remount"); 1929 return -1; 1930 } 1931 } else { 1932 clear_opt(sb, DATA_FLAGS); 1933 sbi->s_mount_opt |= m->mount_opt; 1934 } 1935 #ifdef CONFIG_QUOTA 1936 } else if (m->flags & MOPT_QFMT) { 1937 if (sb_any_quota_loaded(sb) && 1938 sbi->s_jquota_fmt != m->mount_opt) { 1939 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 1940 "quota options when quota turned on"); 1941 return -1; 1942 } 1943 if (ext4_has_feature_quota(sb)) { 1944 ext4_msg(sb, KERN_INFO, 1945 "Quota format mount options ignored " 1946 "when QUOTA feature is enabled"); 1947 return 1; 1948 } 1949 sbi->s_jquota_fmt = m->mount_opt; 1950 #endif 1951 } else if (token == Opt_dax) { 1952 #ifdef CONFIG_FS_DAX 1953 ext4_msg(sb, KERN_WARNING, 1954 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 1955 sbi->s_mount_opt |= m->mount_opt; 1956 #else 1957 ext4_msg(sb, KERN_INFO, "dax option not supported"); 1958 return -1; 1959 #endif 1960 } else if (token == Opt_data_err_abort) { 1961 sbi->s_mount_opt |= m->mount_opt; 1962 } else if (token == Opt_data_err_ignore) { 1963 sbi->s_mount_opt &= ~m->mount_opt; 1964 } else { 1965 if (!args->from) 1966 arg = 1; 1967 if (m->flags & MOPT_CLEAR) 1968 arg = !arg; 1969 else if (unlikely(!(m->flags & MOPT_SET))) { 1970 ext4_msg(sb, KERN_WARNING, 1971 "buggy handling of option %s", opt); 1972 WARN_ON(1); 1973 return -1; 1974 } 1975 if (arg != 0) 1976 sbi->s_mount_opt |= m->mount_opt; 1977 else 1978 sbi->s_mount_opt &= ~m->mount_opt; 1979 } 1980 return 1; 1981 } 1982 1983 static int parse_options(char *options, struct super_block *sb, 1984 unsigned long *journal_devnum, 1985 unsigned int *journal_ioprio, 1986 int is_remount) 1987 { 1988 struct ext4_sb_info *sbi = EXT4_SB(sb); 1989 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 1990 substring_t args[MAX_OPT_ARGS]; 1991 int token; 1992 1993 if (!options) 1994 return 1; 1995 1996 while ((p = strsep(&options, ",")) != NULL) { 1997 if (!*p) 1998 continue; 1999 /* 2000 * Initialize args struct so we know whether arg was 2001 * found; some options take optional arguments. 2002 */ 2003 args[0].to = args[0].from = NULL; 2004 token = match_token(p, tokens, args); 2005 if (handle_mount_opt(sb, p, token, args, journal_devnum, 2006 journal_ioprio, is_remount) < 0) 2007 return 0; 2008 } 2009 #ifdef CONFIG_QUOTA 2010 /* 2011 * We do the test below only for project quotas. 'usrquota' and 2012 * 'grpquota' mount options are allowed even without quota feature 2013 * to support legacy quotas in quota files. 2014 */ 2015 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 2016 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 2017 "Cannot enable project quota enforcement."); 2018 return 0; 2019 } 2020 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 2021 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 2022 if (usr_qf_name || grp_qf_name) { 2023 if (test_opt(sb, USRQUOTA) && usr_qf_name) 2024 clear_opt(sb, USRQUOTA); 2025 2026 if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2027 clear_opt(sb, GRPQUOTA); 2028 2029 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2030 ext4_msg(sb, KERN_ERR, "old and new quota " 2031 "format mixing"); 2032 return 0; 2033 } 2034 2035 if (!sbi->s_jquota_fmt) { 2036 ext4_msg(sb, KERN_ERR, "journaled quota format " 2037 "not specified"); 2038 return 0; 2039 } 2040 } 2041 #endif 2042 if (test_opt(sb, DIOREAD_NOLOCK)) { 2043 int blocksize = 2044 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2045 2046 if (blocksize < PAGE_SIZE) { 2047 ext4_msg(sb, KERN_ERR, "can't mount with " 2048 "dioread_nolock if block size != PAGE_SIZE"); 2049 return 0; 2050 } 2051 } 2052 return 1; 2053 } 2054 2055 static inline void ext4_show_quota_options(struct seq_file *seq, 2056 struct super_block *sb) 2057 { 2058 #if defined(CONFIG_QUOTA) 2059 struct ext4_sb_info *sbi = EXT4_SB(sb); 2060 char *usr_qf_name, *grp_qf_name; 2061 2062 if (sbi->s_jquota_fmt) { 2063 char *fmtname = ""; 2064 2065 switch (sbi->s_jquota_fmt) { 2066 case QFMT_VFS_OLD: 2067 fmtname = "vfsold"; 2068 break; 2069 case QFMT_VFS_V0: 2070 fmtname = "vfsv0"; 2071 break; 2072 case QFMT_VFS_V1: 2073 fmtname = "vfsv1"; 2074 break; 2075 } 2076 seq_printf(seq, ",jqfmt=%s", fmtname); 2077 } 2078 2079 rcu_read_lock(); 2080 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2081 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2082 if (usr_qf_name) 2083 seq_show_option(seq, "usrjquota", usr_qf_name); 2084 if (grp_qf_name) 2085 seq_show_option(seq, "grpjquota", grp_qf_name); 2086 rcu_read_unlock(); 2087 #endif 2088 } 2089 2090 static const char *token2str(int token) 2091 { 2092 const struct match_token *t; 2093 2094 for (t = tokens; t->token != Opt_err; t++) 2095 if (t->token == token && !strchr(t->pattern, '=')) 2096 break; 2097 return t->pattern; 2098 } 2099 2100 /* 2101 * Show an option if 2102 * - it's set to a non-default value OR 2103 * - if the per-sb default is different from the global default 2104 */ 2105 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2106 int nodefs) 2107 { 2108 struct ext4_sb_info *sbi = EXT4_SB(sb); 2109 struct ext4_super_block *es = sbi->s_es; 2110 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2111 const struct mount_opts *m; 2112 char sep = nodefs ? '\n' : ','; 2113 2114 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2115 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2116 2117 if (sbi->s_sb_block != 1) 2118 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2119 2120 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2121 int want_set = m->flags & MOPT_SET; 2122 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2123 (m->flags & MOPT_CLEAR_ERR)) 2124 continue; 2125 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2126 continue; /* skip if same as the default */ 2127 if ((want_set && 2128 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2129 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2130 continue; /* select Opt_noFoo vs Opt_Foo */ 2131 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2132 } 2133 2134 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2135 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2136 SEQ_OPTS_PRINT("resuid=%u", 2137 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2138 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2139 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2140 SEQ_OPTS_PRINT("resgid=%u", 2141 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2142 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2143 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2144 SEQ_OPTS_PUTS("errors=remount-ro"); 2145 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2146 SEQ_OPTS_PUTS("errors=continue"); 2147 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2148 SEQ_OPTS_PUTS("errors=panic"); 2149 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2150 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2151 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2152 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2153 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2154 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2155 if (sb->s_flags & SB_I_VERSION) 2156 SEQ_OPTS_PUTS("i_version"); 2157 if (nodefs || sbi->s_stripe) 2158 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2159 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2160 (sbi->s_mount_opt ^ def_mount_opt)) { 2161 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2162 SEQ_OPTS_PUTS("data=journal"); 2163 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2164 SEQ_OPTS_PUTS("data=ordered"); 2165 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2166 SEQ_OPTS_PUTS("data=writeback"); 2167 } 2168 if (nodefs || 2169 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2170 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2171 sbi->s_inode_readahead_blks); 2172 2173 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2174 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2175 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2176 if (nodefs || sbi->s_max_dir_size_kb) 2177 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2178 if (test_opt(sb, DATA_ERR_ABORT)) 2179 SEQ_OPTS_PUTS("data_err=abort"); 2180 if (DUMMY_ENCRYPTION_ENABLED(sbi)) 2181 SEQ_OPTS_PUTS("test_dummy_encryption"); 2182 2183 ext4_show_quota_options(seq, sb); 2184 return 0; 2185 } 2186 2187 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2188 { 2189 return _ext4_show_options(seq, root->d_sb, 0); 2190 } 2191 2192 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2193 { 2194 struct super_block *sb = seq->private; 2195 int rc; 2196 2197 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 2198 rc = _ext4_show_options(seq, sb, 1); 2199 seq_puts(seq, "\n"); 2200 return rc; 2201 } 2202 2203 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2204 int read_only) 2205 { 2206 struct ext4_sb_info *sbi = EXT4_SB(sb); 2207 int err = 0; 2208 2209 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2210 ext4_msg(sb, KERN_ERR, "revision level too high, " 2211 "forcing read-only mode"); 2212 err = -EROFS; 2213 } 2214 if (read_only) 2215 goto done; 2216 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2217 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2218 "running e2fsck is recommended"); 2219 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2220 ext4_msg(sb, KERN_WARNING, 2221 "warning: mounting fs with errors, " 2222 "running e2fsck is recommended"); 2223 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2224 le16_to_cpu(es->s_mnt_count) >= 2225 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2226 ext4_msg(sb, KERN_WARNING, 2227 "warning: maximal mount count reached, " 2228 "running e2fsck is recommended"); 2229 else if (le32_to_cpu(es->s_checkinterval) && 2230 (ext4_get_tstamp(es, s_lastcheck) + 2231 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2232 ext4_msg(sb, KERN_WARNING, 2233 "warning: checktime reached, " 2234 "running e2fsck is recommended"); 2235 if (!sbi->s_journal) 2236 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2237 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2238 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2239 le16_add_cpu(&es->s_mnt_count, 1); 2240 ext4_update_tstamp(es, s_mtime); 2241 ext4_update_dynamic_rev(sb); 2242 if (sbi->s_journal) 2243 ext4_set_feature_journal_needs_recovery(sb); 2244 2245 err = ext4_commit_super(sb, 1); 2246 done: 2247 if (test_opt(sb, DEBUG)) 2248 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2249 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2250 sb->s_blocksize, 2251 sbi->s_groups_count, 2252 EXT4_BLOCKS_PER_GROUP(sb), 2253 EXT4_INODES_PER_GROUP(sb), 2254 sbi->s_mount_opt, sbi->s_mount_opt2); 2255 2256 cleancache_init_fs(sb); 2257 return err; 2258 } 2259 2260 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2261 { 2262 struct ext4_sb_info *sbi = EXT4_SB(sb); 2263 struct flex_groups *new_groups; 2264 int size; 2265 2266 if (!sbi->s_log_groups_per_flex) 2267 return 0; 2268 2269 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2270 if (size <= sbi->s_flex_groups_allocated) 2271 return 0; 2272 2273 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2274 new_groups = kvzalloc(size, GFP_KERNEL); 2275 if (!new_groups) { 2276 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2277 size / (int) sizeof(struct flex_groups)); 2278 return -ENOMEM; 2279 } 2280 2281 if (sbi->s_flex_groups) { 2282 memcpy(new_groups, sbi->s_flex_groups, 2283 (sbi->s_flex_groups_allocated * 2284 sizeof(struct flex_groups))); 2285 kvfree(sbi->s_flex_groups); 2286 } 2287 sbi->s_flex_groups = new_groups; 2288 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2289 return 0; 2290 } 2291 2292 static int ext4_fill_flex_info(struct super_block *sb) 2293 { 2294 struct ext4_sb_info *sbi = EXT4_SB(sb); 2295 struct ext4_group_desc *gdp = NULL; 2296 ext4_group_t flex_group; 2297 int i, err; 2298 2299 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2300 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2301 sbi->s_log_groups_per_flex = 0; 2302 return 1; 2303 } 2304 2305 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2306 if (err) 2307 goto failed; 2308 2309 for (i = 0; i < sbi->s_groups_count; i++) { 2310 gdp = ext4_get_group_desc(sb, i, NULL); 2311 2312 flex_group = ext4_flex_group(sbi, i); 2313 atomic_add(ext4_free_inodes_count(sb, gdp), 2314 &sbi->s_flex_groups[flex_group].free_inodes); 2315 atomic64_add(ext4_free_group_clusters(sb, gdp), 2316 &sbi->s_flex_groups[flex_group].free_clusters); 2317 atomic_add(ext4_used_dirs_count(sb, gdp), 2318 &sbi->s_flex_groups[flex_group].used_dirs); 2319 } 2320 2321 return 1; 2322 failed: 2323 return 0; 2324 } 2325 2326 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2327 struct ext4_group_desc *gdp) 2328 { 2329 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2330 __u16 crc = 0; 2331 __le32 le_group = cpu_to_le32(block_group); 2332 struct ext4_sb_info *sbi = EXT4_SB(sb); 2333 2334 if (ext4_has_metadata_csum(sbi->s_sb)) { 2335 /* Use new metadata_csum algorithm */ 2336 __u32 csum32; 2337 __u16 dummy_csum = 0; 2338 2339 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2340 sizeof(le_group)); 2341 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2342 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2343 sizeof(dummy_csum)); 2344 offset += sizeof(dummy_csum); 2345 if (offset < sbi->s_desc_size) 2346 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2347 sbi->s_desc_size - offset); 2348 2349 crc = csum32 & 0xFFFF; 2350 goto out; 2351 } 2352 2353 /* old crc16 code */ 2354 if (!ext4_has_feature_gdt_csum(sb)) 2355 return 0; 2356 2357 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2358 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2359 crc = crc16(crc, (__u8 *)gdp, offset); 2360 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2361 /* for checksum of struct ext4_group_desc do the rest...*/ 2362 if (ext4_has_feature_64bit(sb) && 2363 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2364 crc = crc16(crc, (__u8 *)gdp + offset, 2365 le16_to_cpu(sbi->s_es->s_desc_size) - 2366 offset); 2367 2368 out: 2369 return cpu_to_le16(crc); 2370 } 2371 2372 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2373 struct ext4_group_desc *gdp) 2374 { 2375 if (ext4_has_group_desc_csum(sb) && 2376 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2377 return 0; 2378 2379 return 1; 2380 } 2381 2382 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2383 struct ext4_group_desc *gdp) 2384 { 2385 if (!ext4_has_group_desc_csum(sb)) 2386 return; 2387 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2388 } 2389 2390 /* Called at mount-time, super-block is locked */ 2391 static int ext4_check_descriptors(struct super_block *sb, 2392 ext4_fsblk_t sb_block, 2393 ext4_group_t *first_not_zeroed) 2394 { 2395 struct ext4_sb_info *sbi = EXT4_SB(sb); 2396 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2397 ext4_fsblk_t last_block; 2398 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2399 ext4_fsblk_t block_bitmap; 2400 ext4_fsblk_t inode_bitmap; 2401 ext4_fsblk_t inode_table; 2402 int flexbg_flag = 0; 2403 ext4_group_t i, grp = sbi->s_groups_count; 2404 2405 if (ext4_has_feature_flex_bg(sb)) 2406 flexbg_flag = 1; 2407 2408 ext4_debug("Checking group descriptors"); 2409 2410 for (i = 0; i < sbi->s_groups_count; i++) { 2411 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2412 2413 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2414 last_block = ext4_blocks_count(sbi->s_es) - 1; 2415 else 2416 last_block = first_block + 2417 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2418 2419 if ((grp == sbi->s_groups_count) && 2420 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2421 grp = i; 2422 2423 block_bitmap = ext4_block_bitmap(sb, gdp); 2424 if (block_bitmap == sb_block) { 2425 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2426 "Block bitmap for group %u overlaps " 2427 "superblock", i); 2428 if (!sb_rdonly(sb)) 2429 return 0; 2430 } 2431 if (block_bitmap >= sb_block + 1 && 2432 block_bitmap <= last_bg_block) { 2433 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2434 "Block bitmap for group %u overlaps " 2435 "block group descriptors", i); 2436 if (!sb_rdonly(sb)) 2437 return 0; 2438 } 2439 if (block_bitmap < first_block || block_bitmap > last_block) { 2440 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2441 "Block bitmap for group %u not in group " 2442 "(block %llu)!", i, block_bitmap); 2443 return 0; 2444 } 2445 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2446 if (inode_bitmap == sb_block) { 2447 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2448 "Inode bitmap for group %u overlaps " 2449 "superblock", i); 2450 if (!sb_rdonly(sb)) 2451 return 0; 2452 } 2453 if (inode_bitmap >= sb_block + 1 && 2454 inode_bitmap <= last_bg_block) { 2455 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2456 "Inode bitmap for group %u overlaps " 2457 "block group descriptors", i); 2458 if (!sb_rdonly(sb)) 2459 return 0; 2460 } 2461 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2462 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2463 "Inode bitmap for group %u not in group " 2464 "(block %llu)!", i, inode_bitmap); 2465 return 0; 2466 } 2467 inode_table = ext4_inode_table(sb, gdp); 2468 if (inode_table == sb_block) { 2469 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2470 "Inode table for group %u overlaps " 2471 "superblock", i); 2472 if (!sb_rdonly(sb)) 2473 return 0; 2474 } 2475 if (inode_table >= sb_block + 1 && 2476 inode_table <= last_bg_block) { 2477 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2478 "Inode table for group %u overlaps " 2479 "block group descriptors", i); 2480 if (!sb_rdonly(sb)) 2481 return 0; 2482 } 2483 if (inode_table < first_block || 2484 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2485 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2486 "Inode table for group %u not in group " 2487 "(block %llu)!", i, inode_table); 2488 return 0; 2489 } 2490 ext4_lock_group(sb, i); 2491 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2492 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2493 "Checksum for group %u failed (%u!=%u)", 2494 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2495 gdp)), le16_to_cpu(gdp->bg_checksum)); 2496 if (!sb_rdonly(sb)) { 2497 ext4_unlock_group(sb, i); 2498 return 0; 2499 } 2500 } 2501 ext4_unlock_group(sb, i); 2502 if (!flexbg_flag) 2503 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2504 } 2505 if (NULL != first_not_zeroed) 2506 *first_not_zeroed = grp; 2507 return 1; 2508 } 2509 2510 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2511 * the superblock) which were deleted from all directories, but held open by 2512 * a process at the time of a crash. We walk the list and try to delete these 2513 * inodes at recovery time (only with a read-write filesystem). 2514 * 2515 * In order to keep the orphan inode chain consistent during traversal (in 2516 * case of crash during recovery), we link each inode into the superblock 2517 * orphan list_head and handle it the same way as an inode deletion during 2518 * normal operation (which journals the operations for us). 2519 * 2520 * We only do an iget() and an iput() on each inode, which is very safe if we 2521 * accidentally point at an in-use or already deleted inode. The worst that 2522 * can happen in this case is that we get a "bit already cleared" message from 2523 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2524 * e2fsck was run on this filesystem, and it must have already done the orphan 2525 * inode cleanup for us, so we can safely abort without any further action. 2526 */ 2527 static void ext4_orphan_cleanup(struct super_block *sb, 2528 struct ext4_super_block *es) 2529 { 2530 unsigned int s_flags = sb->s_flags; 2531 int ret, nr_orphans = 0, nr_truncates = 0; 2532 #ifdef CONFIG_QUOTA 2533 int quota_update = 0; 2534 int i; 2535 #endif 2536 if (!es->s_last_orphan) { 2537 jbd_debug(4, "no orphan inodes to clean up\n"); 2538 return; 2539 } 2540 2541 if (bdev_read_only(sb->s_bdev)) { 2542 ext4_msg(sb, KERN_ERR, "write access " 2543 "unavailable, skipping orphan cleanup"); 2544 return; 2545 } 2546 2547 /* Check if feature set would not allow a r/w mount */ 2548 if (!ext4_feature_set_ok(sb, 0)) { 2549 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2550 "unknown ROCOMPAT features"); 2551 return; 2552 } 2553 2554 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2555 /* don't clear list on RO mount w/ errors */ 2556 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 2557 ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2558 "clearing orphan list.\n"); 2559 es->s_last_orphan = 0; 2560 } 2561 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2562 return; 2563 } 2564 2565 if (s_flags & SB_RDONLY) { 2566 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2567 sb->s_flags &= ~SB_RDONLY; 2568 } 2569 #ifdef CONFIG_QUOTA 2570 /* Needed for iput() to work correctly and not trash data */ 2571 sb->s_flags |= SB_ACTIVE; 2572 2573 /* 2574 * Turn on quotas which were not enabled for read-only mounts if 2575 * filesystem has quota feature, so that they are updated correctly. 2576 */ 2577 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 2578 int ret = ext4_enable_quotas(sb); 2579 2580 if (!ret) 2581 quota_update = 1; 2582 else 2583 ext4_msg(sb, KERN_ERR, 2584 "Cannot turn on quotas: error %d", ret); 2585 } 2586 2587 /* Turn on journaled quotas used for old sytle */ 2588 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2589 if (EXT4_SB(sb)->s_qf_names[i]) { 2590 int ret = ext4_quota_on_mount(sb, i); 2591 2592 if (!ret) 2593 quota_update = 1; 2594 else 2595 ext4_msg(sb, KERN_ERR, 2596 "Cannot turn on journaled " 2597 "quota: type %d: error %d", i, ret); 2598 } 2599 } 2600 #endif 2601 2602 while (es->s_last_orphan) { 2603 struct inode *inode; 2604 2605 /* 2606 * We may have encountered an error during cleanup; if 2607 * so, skip the rest. 2608 */ 2609 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2610 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2611 es->s_last_orphan = 0; 2612 break; 2613 } 2614 2615 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2616 if (IS_ERR(inode)) { 2617 es->s_last_orphan = 0; 2618 break; 2619 } 2620 2621 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2622 dquot_initialize(inode); 2623 if (inode->i_nlink) { 2624 if (test_opt(sb, DEBUG)) 2625 ext4_msg(sb, KERN_DEBUG, 2626 "%s: truncating inode %lu to %lld bytes", 2627 __func__, inode->i_ino, inode->i_size); 2628 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2629 inode->i_ino, inode->i_size); 2630 inode_lock(inode); 2631 truncate_inode_pages(inode->i_mapping, inode->i_size); 2632 ret = ext4_truncate(inode); 2633 if (ret) 2634 ext4_std_error(inode->i_sb, ret); 2635 inode_unlock(inode); 2636 nr_truncates++; 2637 } else { 2638 if (test_opt(sb, DEBUG)) 2639 ext4_msg(sb, KERN_DEBUG, 2640 "%s: deleting unreferenced inode %lu", 2641 __func__, inode->i_ino); 2642 jbd_debug(2, "deleting unreferenced inode %lu\n", 2643 inode->i_ino); 2644 nr_orphans++; 2645 } 2646 iput(inode); /* The delete magic happens here! */ 2647 } 2648 2649 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2650 2651 if (nr_orphans) 2652 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2653 PLURAL(nr_orphans)); 2654 if (nr_truncates) 2655 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2656 PLURAL(nr_truncates)); 2657 #ifdef CONFIG_QUOTA 2658 /* Turn off quotas if they were enabled for orphan cleanup */ 2659 if (quota_update) { 2660 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2661 if (sb_dqopt(sb)->files[i]) 2662 dquot_quota_off(sb, i); 2663 } 2664 } 2665 #endif 2666 sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2667 } 2668 2669 /* 2670 * Maximal extent format file size. 2671 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2672 * extent format containers, within a sector_t, and within i_blocks 2673 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2674 * so that won't be a limiting factor. 2675 * 2676 * However there is other limiting factor. We do store extents in the form 2677 * of starting block and length, hence the resulting length of the extent 2678 * covering maximum file size must fit into on-disk format containers as 2679 * well. Given that length is always by 1 unit bigger than max unit (because 2680 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2681 * 2682 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2683 */ 2684 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2685 { 2686 loff_t res; 2687 loff_t upper_limit = MAX_LFS_FILESIZE; 2688 2689 /* small i_blocks in vfs inode? */ 2690 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2691 /* 2692 * CONFIG_LBDAF is not enabled implies the inode 2693 * i_block represent total blocks in 512 bytes 2694 * 32 == size of vfs inode i_blocks * 8 2695 */ 2696 upper_limit = (1LL << 32) - 1; 2697 2698 /* total blocks in file system block size */ 2699 upper_limit >>= (blkbits - 9); 2700 upper_limit <<= blkbits; 2701 } 2702 2703 /* 2704 * 32-bit extent-start container, ee_block. We lower the maxbytes 2705 * by one fs block, so ee_len can cover the extent of maximum file 2706 * size 2707 */ 2708 res = (1LL << 32) - 1; 2709 res <<= blkbits; 2710 2711 /* Sanity check against vm- & vfs- imposed limits */ 2712 if (res > upper_limit) 2713 res = upper_limit; 2714 2715 return res; 2716 } 2717 2718 /* 2719 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2720 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2721 * We need to be 1 filesystem block less than the 2^48 sector limit. 2722 */ 2723 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2724 { 2725 loff_t res = EXT4_NDIR_BLOCKS; 2726 int meta_blocks; 2727 loff_t upper_limit; 2728 /* This is calculated to be the largest file size for a dense, block 2729 * mapped file such that the file's total number of 512-byte sectors, 2730 * including data and all indirect blocks, does not exceed (2^48 - 1). 2731 * 2732 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2733 * number of 512-byte sectors of the file. 2734 */ 2735 2736 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2737 /* 2738 * !has_huge_files or CONFIG_LBDAF not enabled implies that 2739 * the inode i_block field represents total file blocks in 2740 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 2741 */ 2742 upper_limit = (1LL << 32) - 1; 2743 2744 /* total blocks in file system block size */ 2745 upper_limit >>= (bits - 9); 2746 2747 } else { 2748 /* 2749 * We use 48 bit ext4_inode i_blocks 2750 * With EXT4_HUGE_FILE_FL set the i_blocks 2751 * represent total number of blocks in 2752 * file system block size 2753 */ 2754 upper_limit = (1LL << 48) - 1; 2755 2756 } 2757 2758 /* indirect blocks */ 2759 meta_blocks = 1; 2760 /* double indirect blocks */ 2761 meta_blocks += 1 + (1LL << (bits-2)); 2762 /* tripple indirect blocks */ 2763 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2764 2765 upper_limit -= meta_blocks; 2766 upper_limit <<= bits; 2767 2768 res += 1LL << (bits-2); 2769 res += 1LL << (2*(bits-2)); 2770 res += 1LL << (3*(bits-2)); 2771 res <<= bits; 2772 if (res > upper_limit) 2773 res = upper_limit; 2774 2775 if (res > MAX_LFS_FILESIZE) 2776 res = MAX_LFS_FILESIZE; 2777 2778 return res; 2779 } 2780 2781 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2782 ext4_fsblk_t logical_sb_block, int nr) 2783 { 2784 struct ext4_sb_info *sbi = EXT4_SB(sb); 2785 ext4_group_t bg, first_meta_bg; 2786 int has_super = 0; 2787 2788 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2789 2790 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 2791 return logical_sb_block + nr + 1; 2792 bg = sbi->s_desc_per_block * nr; 2793 if (ext4_bg_has_super(sb, bg)) 2794 has_super = 1; 2795 2796 /* 2797 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2798 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2799 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2800 * compensate. 2801 */ 2802 if (sb->s_blocksize == 1024 && nr == 0 && 2803 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 2804 has_super++; 2805 2806 return (has_super + ext4_group_first_block_no(sb, bg)); 2807 } 2808 2809 /** 2810 * ext4_get_stripe_size: Get the stripe size. 2811 * @sbi: In memory super block info 2812 * 2813 * If we have specified it via mount option, then 2814 * use the mount option value. If the value specified at mount time is 2815 * greater than the blocks per group use the super block value. 2816 * If the super block value is greater than blocks per group return 0. 2817 * Allocator needs it be less than blocks per group. 2818 * 2819 */ 2820 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2821 { 2822 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2823 unsigned long stripe_width = 2824 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2825 int ret; 2826 2827 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2828 ret = sbi->s_stripe; 2829 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 2830 ret = stripe_width; 2831 else if (stride && stride <= sbi->s_blocks_per_group) 2832 ret = stride; 2833 else 2834 ret = 0; 2835 2836 /* 2837 * If the stripe width is 1, this makes no sense and 2838 * we set it to 0 to turn off stripe handling code. 2839 */ 2840 if (ret <= 1) 2841 ret = 0; 2842 2843 return ret; 2844 } 2845 2846 /* 2847 * Check whether this filesystem can be mounted based on 2848 * the features present and the RDONLY/RDWR mount requested. 2849 * Returns 1 if this filesystem can be mounted as requested, 2850 * 0 if it cannot be. 2851 */ 2852 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2853 { 2854 if (ext4_has_unknown_ext4_incompat_features(sb)) { 2855 ext4_msg(sb, KERN_ERR, 2856 "Couldn't mount because of " 2857 "unsupported optional features (%x)", 2858 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2859 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2860 return 0; 2861 } 2862 2863 if (readonly) 2864 return 1; 2865 2866 if (ext4_has_feature_readonly(sb)) { 2867 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 2868 sb->s_flags |= SB_RDONLY; 2869 return 1; 2870 } 2871 2872 /* Check that feature set is OK for a read-write mount */ 2873 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2874 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2875 "unsupported optional features (%x)", 2876 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2877 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2878 return 0; 2879 } 2880 /* 2881 * Large file size enabled file system can only be mounted 2882 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2883 */ 2884 if (ext4_has_feature_huge_file(sb)) { 2885 if (sizeof(blkcnt_t) < sizeof(u64)) { 2886 ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2887 "cannot be mounted RDWR without " 2888 "CONFIG_LBDAF"); 2889 return 0; 2890 } 2891 } 2892 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2893 ext4_msg(sb, KERN_ERR, 2894 "Can't support bigalloc feature without " 2895 "extents feature\n"); 2896 return 0; 2897 } 2898 2899 #ifndef CONFIG_QUOTA 2900 if (ext4_has_feature_quota(sb) && !readonly) { 2901 ext4_msg(sb, KERN_ERR, 2902 "Filesystem with quota feature cannot be mounted RDWR " 2903 "without CONFIG_QUOTA"); 2904 return 0; 2905 } 2906 if (ext4_has_feature_project(sb) && !readonly) { 2907 ext4_msg(sb, KERN_ERR, 2908 "Filesystem with project quota feature cannot be mounted RDWR " 2909 "without CONFIG_QUOTA"); 2910 return 0; 2911 } 2912 #endif /* CONFIG_QUOTA */ 2913 return 1; 2914 } 2915 2916 /* 2917 * This function is called once a day if we have errors logged 2918 * on the file system 2919 */ 2920 static void print_daily_error_info(struct timer_list *t) 2921 { 2922 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 2923 struct super_block *sb = sbi->s_sb; 2924 struct ext4_super_block *es = sbi->s_es; 2925 2926 if (es->s_error_count) 2927 /* fsck newer than v1.41.13 is needed to clean this condition. */ 2928 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 2929 le32_to_cpu(es->s_error_count)); 2930 if (es->s_first_error_time) { 2931 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 2932 sb->s_id, 2933 ext4_get_tstamp(es, s_first_error_time), 2934 (int) sizeof(es->s_first_error_func), 2935 es->s_first_error_func, 2936 le32_to_cpu(es->s_first_error_line)); 2937 if (es->s_first_error_ino) 2938 printk(KERN_CONT ": inode %u", 2939 le32_to_cpu(es->s_first_error_ino)); 2940 if (es->s_first_error_block) 2941 printk(KERN_CONT ": block %llu", (unsigned long long) 2942 le64_to_cpu(es->s_first_error_block)); 2943 printk(KERN_CONT "\n"); 2944 } 2945 if (es->s_last_error_time) { 2946 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 2947 sb->s_id, 2948 ext4_get_tstamp(es, s_last_error_time), 2949 (int) sizeof(es->s_last_error_func), 2950 es->s_last_error_func, 2951 le32_to_cpu(es->s_last_error_line)); 2952 if (es->s_last_error_ino) 2953 printk(KERN_CONT ": inode %u", 2954 le32_to_cpu(es->s_last_error_ino)); 2955 if (es->s_last_error_block) 2956 printk(KERN_CONT ": block %llu", (unsigned long long) 2957 le64_to_cpu(es->s_last_error_block)); 2958 printk(KERN_CONT "\n"); 2959 } 2960 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 2961 } 2962 2963 /* Find next suitable group and run ext4_init_inode_table */ 2964 static int ext4_run_li_request(struct ext4_li_request *elr) 2965 { 2966 struct ext4_group_desc *gdp = NULL; 2967 ext4_group_t group, ngroups; 2968 struct super_block *sb; 2969 unsigned long timeout = 0; 2970 int ret = 0; 2971 2972 sb = elr->lr_super; 2973 ngroups = EXT4_SB(sb)->s_groups_count; 2974 2975 for (group = elr->lr_next_group; group < ngroups; group++) { 2976 gdp = ext4_get_group_desc(sb, group, NULL); 2977 if (!gdp) { 2978 ret = 1; 2979 break; 2980 } 2981 2982 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2983 break; 2984 } 2985 2986 if (group >= ngroups) 2987 ret = 1; 2988 2989 if (!ret) { 2990 timeout = jiffies; 2991 ret = ext4_init_inode_table(sb, group, 2992 elr->lr_timeout ? 0 : 1); 2993 if (elr->lr_timeout == 0) { 2994 timeout = (jiffies - timeout) * 2995 elr->lr_sbi->s_li_wait_mult; 2996 elr->lr_timeout = timeout; 2997 } 2998 elr->lr_next_sched = jiffies + elr->lr_timeout; 2999 elr->lr_next_group = group + 1; 3000 } 3001 return ret; 3002 } 3003 3004 /* 3005 * Remove lr_request from the list_request and free the 3006 * request structure. Should be called with li_list_mtx held 3007 */ 3008 static void ext4_remove_li_request(struct ext4_li_request *elr) 3009 { 3010 struct ext4_sb_info *sbi; 3011 3012 if (!elr) 3013 return; 3014 3015 sbi = elr->lr_sbi; 3016 3017 list_del(&elr->lr_request); 3018 sbi->s_li_request = NULL; 3019 kfree(elr); 3020 } 3021 3022 static void ext4_unregister_li_request(struct super_block *sb) 3023 { 3024 mutex_lock(&ext4_li_mtx); 3025 if (!ext4_li_info) { 3026 mutex_unlock(&ext4_li_mtx); 3027 return; 3028 } 3029 3030 mutex_lock(&ext4_li_info->li_list_mtx); 3031 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3032 mutex_unlock(&ext4_li_info->li_list_mtx); 3033 mutex_unlock(&ext4_li_mtx); 3034 } 3035 3036 static struct task_struct *ext4_lazyinit_task; 3037 3038 /* 3039 * This is the function where ext4lazyinit thread lives. It walks 3040 * through the request list searching for next scheduled filesystem. 3041 * When such a fs is found, run the lazy initialization request 3042 * (ext4_rn_li_request) and keep track of the time spend in this 3043 * function. Based on that time we compute next schedule time of 3044 * the request. When walking through the list is complete, compute 3045 * next waking time and put itself into sleep. 3046 */ 3047 static int ext4_lazyinit_thread(void *arg) 3048 { 3049 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3050 struct list_head *pos, *n; 3051 struct ext4_li_request *elr; 3052 unsigned long next_wakeup, cur; 3053 3054 BUG_ON(NULL == eli); 3055 3056 cont_thread: 3057 while (true) { 3058 next_wakeup = MAX_JIFFY_OFFSET; 3059 3060 mutex_lock(&eli->li_list_mtx); 3061 if (list_empty(&eli->li_request_list)) { 3062 mutex_unlock(&eli->li_list_mtx); 3063 goto exit_thread; 3064 } 3065 list_for_each_safe(pos, n, &eli->li_request_list) { 3066 int err = 0; 3067 int progress = 0; 3068 elr = list_entry(pos, struct ext4_li_request, 3069 lr_request); 3070 3071 if (time_before(jiffies, elr->lr_next_sched)) { 3072 if (time_before(elr->lr_next_sched, next_wakeup)) 3073 next_wakeup = elr->lr_next_sched; 3074 continue; 3075 } 3076 if (down_read_trylock(&elr->lr_super->s_umount)) { 3077 if (sb_start_write_trylock(elr->lr_super)) { 3078 progress = 1; 3079 /* 3080 * We hold sb->s_umount, sb can not 3081 * be removed from the list, it is 3082 * now safe to drop li_list_mtx 3083 */ 3084 mutex_unlock(&eli->li_list_mtx); 3085 err = ext4_run_li_request(elr); 3086 sb_end_write(elr->lr_super); 3087 mutex_lock(&eli->li_list_mtx); 3088 n = pos->next; 3089 } 3090 up_read((&elr->lr_super->s_umount)); 3091 } 3092 /* error, remove the lazy_init job */ 3093 if (err) { 3094 ext4_remove_li_request(elr); 3095 continue; 3096 } 3097 if (!progress) { 3098 elr->lr_next_sched = jiffies + 3099 (prandom_u32() 3100 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3101 } 3102 if (time_before(elr->lr_next_sched, next_wakeup)) 3103 next_wakeup = elr->lr_next_sched; 3104 } 3105 mutex_unlock(&eli->li_list_mtx); 3106 3107 try_to_freeze(); 3108 3109 cur = jiffies; 3110 if ((time_after_eq(cur, next_wakeup)) || 3111 (MAX_JIFFY_OFFSET == next_wakeup)) { 3112 cond_resched(); 3113 continue; 3114 } 3115 3116 schedule_timeout_interruptible(next_wakeup - cur); 3117 3118 if (kthread_should_stop()) { 3119 ext4_clear_request_list(); 3120 goto exit_thread; 3121 } 3122 } 3123 3124 exit_thread: 3125 /* 3126 * It looks like the request list is empty, but we need 3127 * to check it under the li_list_mtx lock, to prevent any 3128 * additions into it, and of course we should lock ext4_li_mtx 3129 * to atomically free the list and ext4_li_info, because at 3130 * this point another ext4 filesystem could be registering 3131 * new one. 3132 */ 3133 mutex_lock(&ext4_li_mtx); 3134 mutex_lock(&eli->li_list_mtx); 3135 if (!list_empty(&eli->li_request_list)) { 3136 mutex_unlock(&eli->li_list_mtx); 3137 mutex_unlock(&ext4_li_mtx); 3138 goto cont_thread; 3139 } 3140 mutex_unlock(&eli->li_list_mtx); 3141 kfree(ext4_li_info); 3142 ext4_li_info = NULL; 3143 mutex_unlock(&ext4_li_mtx); 3144 3145 return 0; 3146 } 3147 3148 static void ext4_clear_request_list(void) 3149 { 3150 struct list_head *pos, *n; 3151 struct ext4_li_request *elr; 3152 3153 mutex_lock(&ext4_li_info->li_list_mtx); 3154 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3155 elr = list_entry(pos, struct ext4_li_request, 3156 lr_request); 3157 ext4_remove_li_request(elr); 3158 } 3159 mutex_unlock(&ext4_li_info->li_list_mtx); 3160 } 3161 3162 static int ext4_run_lazyinit_thread(void) 3163 { 3164 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3165 ext4_li_info, "ext4lazyinit"); 3166 if (IS_ERR(ext4_lazyinit_task)) { 3167 int err = PTR_ERR(ext4_lazyinit_task); 3168 ext4_clear_request_list(); 3169 kfree(ext4_li_info); 3170 ext4_li_info = NULL; 3171 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3172 "initialization thread\n", 3173 err); 3174 return err; 3175 } 3176 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3177 return 0; 3178 } 3179 3180 /* 3181 * Check whether it make sense to run itable init. thread or not. 3182 * If there is at least one uninitialized inode table, return 3183 * corresponding group number, else the loop goes through all 3184 * groups and return total number of groups. 3185 */ 3186 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3187 { 3188 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3189 struct ext4_group_desc *gdp = NULL; 3190 3191 if (!ext4_has_group_desc_csum(sb)) 3192 return ngroups; 3193 3194 for (group = 0; group < ngroups; group++) { 3195 gdp = ext4_get_group_desc(sb, group, NULL); 3196 if (!gdp) 3197 continue; 3198 3199 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3200 break; 3201 } 3202 3203 return group; 3204 } 3205 3206 static int ext4_li_info_new(void) 3207 { 3208 struct ext4_lazy_init *eli = NULL; 3209 3210 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3211 if (!eli) 3212 return -ENOMEM; 3213 3214 INIT_LIST_HEAD(&eli->li_request_list); 3215 mutex_init(&eli->li_list_mtx); 3216 3217 eli->li_state |= EXT4_LAZYINIT_QUIT; 3218 3219 ext4_li_info = eli; 3220 3221 return 0; 3222 } 3223 3224 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3225 ext4_group_t start) 3226 { 3227 struct ext4_sb_info *sbi = EXT4_SB(sb); 3228 struct ext4_li_request *elr; 3229 3230 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3231 if (!elr) 3232 return NULL; 3233 3234 elr->lr_super = sb; 3235 elr->lr_sbi = sbi; 3236 elr->lr_next_group = start; 3237 3238 /* 3239 * Randomize first schedule time of the request to 3240 * spread the inode table initialization requests 3241 * better. 3242 */ 3243 elr->lr_next_sched = jiffies + (prandom_u32() % 3244 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3245 return elr; 3246 } 3247 3248 int ext4_register_li_request(struct super_block *sb, 3249 ext4_group_t first_not_zeroed) 3250 { 3251 struct ext4_sb_info *sbi = EXT4_SB(sb); 3252 struct ext4_li_request *elr = NULL; 3253 ext4_group_t ngroups = sbi->s_groups_count; 3254 int ret = 0; 3255 3256 mutex_lock(&ext4_li_mtx); 3257 if (sbi->s_li_request != NULL) { 3258 /* 3259 * Reset timeout so it can be computed again, because 3260 * s_li_wait_mult might have changed. 3261 */ 3262 sbi->s_li_request->lr_timeout = 0; 3263 goto out; 3264 } 3265 3266 if (first_not_zeroed == ngroups || sb_rdonly(sb) || 3267 !test_opt(sb, INIT_INODE_TABLE)) 3268 goto out; 3269 3270 elr = ext4_li_request_new(sb, first_not_zeroed); 3271 if (!elr) { 3272 ret = -ENOMEM; 3273 goto out; 3274 } 3275 3276 if (NULL == ext4_li_info) { 3277 ret = ext4_li_info_new(); 3278 if (ret) 3279 goto out; 3280 } 3281 3282 mutex_lock(&ext4_li_info->li_list_mtx); 3283 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3284 mutex_unlock(&ext4_li_info->li_list_mtx); 3285 3286 sbi->s_li_request = elr; 3287 /* 3288 * set elr to NULL here since it has been inserted to 3289 * the request_list and the removal and free of it is 3290 * handled by ext4_clear_request_list from now on. 3291 */ 3292 elr = NULL; 3293 3294 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3295 ret = ext4_run_lazyinit_thread(); 3296 if (ret) 3297 goto out; 3298 } 3299 out: 3300 mutex_unlock(&ext4_li_mtx); 3301 if (ret) 3302 kfree(elr); 3303 return ret; 3304 } 3305 3306 /* 3307 * We do not need to lock anything since this is called on 3308 * module unload. 3309 */ 3310 static void ext4_destroy_lazyinit_thread(void) 3311 { 3312 /* 3313 * If thread exited earlier 3314 * there's nothing to be done. 3315 */ 3316 if (!ext4_li_info || !ext4_lazyinit_task) 3317 return; 3318 3319 kthread_stop(ext4_lazyinit_task); 3320 } 3321 3322 static int set_journal_csum_feature_set(struct super_block *sb) 3323 { 3324 int ret = 1; 3325 int compat, incompat; 3326 struct ext4_sb_info *sbi = EXT4_SB(sb); 3327 3328 if (ext4_has_metadata_csum(sb)) { 3329 /* journal checksum v3 */ 3330 compat = 0; 3331 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3332 } else { 3333 /* journal checksum v1 */ 3334 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3335 incompat = 0; 3336 } 3337 3338 jbd2_journal_clear_features(sbi->s_journal, 3339 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3340 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3341 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3342 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3343 ret = jbd2_journal_set_features(sbi->s_journal, 3344 compat, 0, 3345 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3346 incompat); 3347 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3348 ret = jbd2_journal_set_features(sbi->s_journal, 3349 compat, 0, 3350 incompat); 3351 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3352 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3353 } else { 3354 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3355 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3356 } 3357 3358 return ret; 3359 } 3360 3361 /* 3362 * Note: calculating the overhead so we can be compatible with 3363 * historical BSD practice is quite difficult in the face of 3364 * clusters/bigalloc. This is because multiple metadata blocks from 3365 * different block group can end up in the same allocation cluster. 3366 * Calculating the exact overhead in the face of clustered allocation 3367 * requires either O(all block bitmaps) in memory or O(number of block 3368 * groups**2) in time. We will still calculate the superblock for 3369 * older file systems --- and if we come across with a bigalloc file 3370 * system with zero in s_overhead_clusters the estimate will be close to 3371 * correct especially for very large cluster sizes --- but for newer 3372 * file systems, it's better to calculate this figure once at mkfs 3373 * time, and store it in the superblock. If the superblock value is 3374 * present (even for non-bigalloc file systems), we will use it. 3375 */ 3376 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3377 char *buf) 3378 { 3379 struct ext4_sb_info *sbi = EXT4_SB(sb); 3380 struct ext4_group_desc *gdp; 3381 ext4_fsblk_t first_block, last_block, b; 3382 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3383 int s, j, count = 0; 3384 3385 if (!ext4_has_feature_bigalloc(sb)) 3386 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3387 sbi->s_itb_per_group + 2); 3388 3389 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3390 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3391 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3392 for (i = 0; i < ngroups; i++) { 3393 gdp = ext4_get_group_desc(sb, i, NULL); 3394 b = ext4_block_bitmap(sb, gdp); 3395 if (b >= first_block && b <= last_block) { 3396 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3397 count++; 3398 } 3399 b = ext4_inode_bitmap(sb, gdp); 3400 if (b >= first_block && b <= last_block) { 3401 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3402 count++; 3403 } 3404 b = ext4_inode_table(sb, gdp); 3405 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3406 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3407 int c = EXT4_B2C(sbi, b - first_block); 3408 ext4_set_bit(c, buf); 3409 count++; 3410 } 3411 if (i != grp) 3412 continue; 3413 s = 0; 3414 if (ext4_bg_has_super(sb, grp)) { 3415 ext4_set_bit(s++, buf); 3416 count++; 3417 } 3418 j = ext4_bg_num_gdb(sb, grp); 3419 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3420 ext4_error(sb, "Invalid number of block group " 3421 "descriptor blocks: %d", j); 3422 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3423 } 3424 count += j; 3425 for (; j > 0; j--) 3426 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3427 } 3428 if (!count) 3429 return 0; 3430 return EXT4_CLUSTERS_PER_GROUP(sb) - 3431 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3432 } 3433 3434 /* 3435 * Compute the overhead and stash it in sbi->s_overhead 3436 */ 3437 int ext4_calculate_overhead(struct super_block *sb) 3438 { 3439 struct ext4_sb_info *sbi = EXT4_SB(sb); 3440 struct ext4_super_block *es = sbi->s_es; 3441 struct inode *j_inode; 3442 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3443 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3444 ext4_fsblk_t overhead = 0; 3445 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3446 3447 if (!buf) 3448 return -ENOMEM; 3449 3450 /* 3451 * Compute the overhead (FS structures). This is constant 3452 * for a given filesystem unless the number of block groups 3453 * changes so we cache the previous value until it does. 3454 */ 3455 3456 /* 3457 * All of the blocks before first_data_block are overhead 3458 */ 3459 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3460 3461 /* 3462 * Add the overhead found in each block group 3463 */ 3464 for (i = 0; i < ngroups; i++) { 3465 int blks; 3466 3467 blks = count_overhead(sb, i, buf); 3468 overhead += blks; 3469 if (blks) 3470 memset(buf, 0, PAGE_SIZE); 3471 cond_resched(); 3472 } 3473 3474 /* 3475 * Add the internal journal blocks whether the journal has been 3476 * loaded or not 3477 */ 3478 if (sbi->s_journal && !sbi->journal_bdev) 3479 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3480 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { 3481 j_inode = ext4_get_journal_inode(sb, j_inum); 3482 if (j_inode) { 3483 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3484 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3485 iput(j_inode); 3486 } else { 3487 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3488 } 3489 } 3490 sbi->s_overhead = overhead; 3491 smp_wmb(); 3492 free_page((unsigned long) buf); 3493 return 0; 3494 } 3495 3496 static void ext4_set_resv_clusters(struct super_block *sb) 3497 { 3498 ext4_fsblk_t resv_clusters; 3499 struct ext4_sb_info *sbi = EXT4_SB(sb); 3500 3501 /* 3502 * There's no need to reserve anything when we aren't using extents. 3503 * The space estimates are exact, there are no unwritten extents, 3504 * hole punching doesn't need new metadata... This is needed especially 3505 * to keep ext2/3 backward compatibility. 3506 */ 3507 if (!ext4_has_feature_extents(sb)) 3508 return; 3509 /* 3510 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3511 * This should cover the situations where we can not afford to run 3512 * out of space like for example punch hole, or converting 3513 * unwritten extents in delalloc path. In most cases such 3514 * allocation would require 1, or 2 blocks, higher numbers are 3515 * very rare. 3516 */ 3517 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3518 sbi->s_cluster_bits); 3519 3520 do_div(resv_clusters, 50); 3521 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3522 3523 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3524 } 3525 3526 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3527 { 3528 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3529 char *orig_data = kstrdup(data, GFP_KERNEL); 3530 struct buffer_head *bh; 3531 struct ext4_super_block *es = NULL; 3532 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3533 ext4_fsblk_t block; 3534 ext4_fsblk_t sb_block = get_sb_block(&data); 3535 ext4_fsblk_t logical_sb_block; 3536 unsigned long offset = 0; 3537 unsigned long journal_devnum = 0; 3538 unsigned long def_mount_opts; 3539 struct inode *root; 3540 const char *descr; 3541 int ret = -ENOMEM; 3542 int blocksize, clustersize; 3543 unsigned int db_count; 3544 unsigned int i; 3545 int needs_recovery, has_huge_files, has_bigalloc; 3546 __u64 blocks_count; 3547 int err = 0; 3548 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3549 ext4_group_t first_not_zeroed; 3550 3551 if ((data && !orig_data) || !sbi) 3552 goto out_free_base; 3553 3554 sbi->s_daxdev = dax_dev; 3555 sbi->s_blockgroup_lock = 3556 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3557 if (!sbi->s_blockgroup_lock) 3558 goto out_free_base; 3559 3560 sb->s_fs_info = sbi; 3561 sbi->s_sb = sb; 3562 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3563 sbi->s_sb_block = sb_block; 3564 if (sb->s_bdev->bd_part) 3565 sbi->s_sectors_written_start = 3566 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3567 3568 /* Cleanup superblock name */ 3569 strreplace(sb->s_id, '/', '!'); 3570 3571 /* -EINVAL is default */ 3572 ret = -EINVAL; 3573 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3574 if (!blocksize) { 3575 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3576 goto out_fail; 3577 } 3578 3579 /* 3580 * The ext4 superblock will not be buffer aligned for other than 1kB 3581 * block sizes. We need to calculate the offset from buffer start. 3582 */ 3583 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3584 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3585 offset = do_div(logical_sb_block, blocksize); 3586 } else { 3587 logical_sb_block = sb_block; 3588 } 3589 3590 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3591 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3592 goto out_fail; 3593 } 3594 /* 3595 * Note: s_es must be initialized as soon as possible because 3596 * some ext4 macro-instructions depend on its value 3597 */ 3598 es = (struct ext4_super_block *) (bh->b_data + offset); 3599 sbi->s_es = es; 3600 sb->s_magic = le16_to_cpu(es->s_magic); 3601 if (sb->s_magic != EXT4_SUPER_MAGIC) 3602 goto cantfind_ext4; 3603 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3604 3605 /* Warn if metadata_csum and gdt_csum are both set. */ 3606 if (ext4_has_feature_metadata_csum(sb) && 3607 ext4_has_feature_gdt_csum(sb)) 3608 ext4_warning(sb, "metadata_csum and uninit_bg are " 3609 "redundant flags; please run fsck."); 3610 3611 /* Check for a known checksum algorithm */ 3612 if (!ext4_verify_csum_type(sb, es)) { 3613 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3614 "unknown checksum algorithm."); 3615 silent = 1; 3616 goto cantfind_ext4; 3617 } 3618 3619 /* Load the checksum driver */ 3620 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3621 if (IS_ERR(sbi->s_chksum_driver)) { 3622 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3623 ret = PTR_ERR(sbi->s_chksum_driver); 3624 sbi->s_chksum_driver = NULL; 3625 goto failed_mount; 3626 } 3627 3628 /* Check superblock checksum */ 3629 if (!ext4_superblock_csum_verify(sb, es)) { 3630 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3631 "invalid superblock checksum. Run e2fsck?"); 3632 silent = 1; 3633 ret = -EFSBADCRC; 3634 goto cantfind_ext4; 3635 } 3636 3637 /* Precompute checksum seed for all metadata */ 3638 if (ext4_has_feature_csum_seed(sb)) 3639 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3640 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3641 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3642 sizeof(es->s_uuid)); 3643 3644 /* Set defaults before we parse the mount options */ 3645 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3646 set_opt(sb, INIT_INODE_TABLE); 3647 if (def_mount_opts & EXT4_DEFM_DEBUG) 3648 set_opt(sb, DEBUG); 3649 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3650 set_opt(sb, GRPID); 3651 if (def_mount_opts & EXT4_DEFM_UID16) 3652 set_opt(sb, NO_UID32); 3653 /* xattr user namespace & acls are now defaulted on */ 3654 set_opt(sb, XATTR_USER); 3655 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3656 set_opt(sb, POSIX_ACL); 3657 #endif 3658 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 3659 if (ext4_has_metadata_csum(sb)) 3660 set_opt(sb, JOURNAL_CHECKSUM); 3661 3662 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3663 set_opt(sb, JOURNAL_DATA); 3664 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3665 set_opt(sb, ORDERED_DATA); 3666 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3667 set_opt(sb, WRITEBACK_DATA); 3668 3669 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3670 set_opt(sb, ERRORS_PANIC); 3671 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3672 set_opt(sb, ERRORS_CONT); 3673 else 3674 set_opt(sb, ERRORS_RO); 3675 /* block_validity enabled by default; disable with noblock_validity */ 3676 set_opt(sb, BLOCK_VALIDITY); 3677 if (def_mount_opts & EXT4_DEFM_DISCARD) 3678 set_opt(sb, DISCARD); 3679 3680 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3681 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3682 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3683 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3684 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3685 3686 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3687 set_opt(sb, BARRIER); 3688 3689 /* 3690 * enable delayed allocation by default 3691 * Use -o nodelalloc to turn it off 3692 */ 3693 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3694 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3695 set_opt(sb, DELALLOC); 3696 3697 /* 3698 * set default s_li_wait_mult for lazyinit, for the case there is 3699 * no mount option specified. 3700 */ 3701 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3702 3703 if (sbi->s_es->s_mount_opts[0]) { 3704 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 3705 sizeof(sbi->s_es->s_mount_opts), 3706 GFP_KERNEL); 3707 if (!s_mount_opts) 3708 goto failed_mount; 3709 if (!parse_options(s_mount_opts, sb, &journal_devnum, 3710 &journal_ioprio, 0)) { 3711 ext4_msg(sb, KERN_WARNING, 3712 "failed to parse options in superblock: %s", 3713 s_mount_opts); 3714 } 3715 kfree(s_mount_opts); 3716 } 3717 sbi->s_def_mount_opt = sbi->s_mount_opt; 3718 if (!parse_options((char *) data, sb, &journal_devnum, 3719 &journal_ioprio, 0)) 3720 goto failed_mount; 3721 3722 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3723 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3724 "with data=journal disables delayed " 3725 "allocation and O_DIRECT support!\n"); 3726 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3727 ext4_msg(sb, KERN_ERR, "can't mount with " 3728 "both data=journal and delalloc"); 3729 goto failed_mount; 3730 } 3731 if (test_opt(sb, DIOREAD_NOLOCK)) { 3732 ext4_msg(sb, KERN_ERR, "can't mount with " 3733 "both data=journal and dioread_nolock"); 3734 goto failed_mount; 3735 } 3736 if (test_opt(sb, DAX)) { 3737 ext4_msg(sb, KERN_ERR, "can't mount with " 3738 "both data=journal and dax"); 3739 goto failed_mount; 3740 } 3741 if (ext4_has_feature_encrypt(sb)) { 3742 ext4_msg(sb, KERN_WARNING, 3743 "encrypted files will use data=ordered " 3744 "instead of data journaling mode"); 3745 } 3746 if (test_opt(sb, DELALLOC)) 3747 clear_opt(sb, DELALLOC); 3748 } else { 3749 sb->s_iflags |= SB_I_CGROUPWB; 3750 } 3751 3752 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3753 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 3754 3755 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3756 (ext4_has_compat_features(sb) || 3757 ext4_has_ro_compat_features(sb) || 3758 ext4_has_incompat_features(sb))) 3759 ext4_msg(sb, KERN_WARNING, 3760 "feature flags set on rev 0 fs, " 3761 "running e2fsck is recommended"); 3762 3763 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3764 set_opt2(sb, HURD_COMPAT); 3765 if (ext4_has_feature_64bit(sb)) { 3766 ext4_msg(sb, KERN_ERR, 3767 "The Hurd can't support 64-bit file systems"); 3768 goto failed_mount; 3769 } 3770 3771 /* 3772 * ea_inode feature uses l_i_version field which is not 3773 * available in HURD_COMPAT mode. 3774 */ 3775 if (ext4_has_feature_ea_inode(sb)) { 3776 ext4_msg(sb, KERN_ERR, 3777 "ea_inode feature is not supported for Hurd"); 3778 goto failed_mount; 3779 } 3780 } 3781 3782 if (IS_EXT2_SB(sb)) { 3783 if (ext2_feature_set_ok(sb)) 3784 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3785 "using the ext4 subsystem"); 3786 else { 3787 /* 3788 * If we're probing be silent, if this looks like 3789 * it's actually an ext[34] filesystem. 3790 */ 3791 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3792 goto failed_mount; 3793 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3794 "to feature incompatibilities"); 3795 goto failed_mount; 3796 } 3797 } 3798 3799 if (IS_EXT3_SB(sb)) { 3800 if (ext3_feature_set_ok(sb)) 3801 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3802 "using the ext4 subsystem"); 3803 else { 3804 /* 3805 * If we're probing be silent, if this looks like 3806 * it's actually an ext4 filesystem. 3807 */ 3808 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3809 goto failed_mount; 3810 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3811 "to feature incompatibilities"); 3812 goto failed_mount; 3813 } 3814 } 3815 3816 /* 3817 * Check feature flags regardless of the revision level, since we 3818 * previously didn't change the revision level when setting the flags, 3819 * so there is a chance incompat flags are set on a rev 0 filesystem. 3820 */ 3821 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 3822 goto failed_mount; 3823 3824 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3825 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3826 blocksize > EXT4_MAX_BLOCK_SIZE) { 3827 ext4_msg(sb, KERN_ERR, 3828 "Unsupported filesystem blocksize %d (%d log_block_size)", 3829 blocksize, le32_to_cpu(es->s_log_block_size)); 3830 goto failed_mount; 3831 } 3832 if (le32_to_cpu(es->s_log_block_size) > 3833 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3834 ext4_msg(sb, KERN_ERR, 3835 "Invalid log block size: %u", 3836 le32_to_cpu(es->s_log_block_size)); 3837 goto failed_mount; 3838 } 3839 if (le32_to_cpu(es->s_log_cluster_size) > 3840 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 3841 ext4_msg(sb, KERN_ERR, 3842 "Invalid log cluster size: %u", 3843 le32_to_cpu(es->s_log_cluster_size)); 3844 goto failed_mount; 3845 } 3846 3847 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 3848 ext4_msg(sb, KERN_ERR, 3849 "Number of reserved GDT blocks insanely large: %d", 3850 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 3851 goto failed_mount; 3852 } 3853 3854 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 3855 if (ext4_has_feature_inline_data(sb)) { 3856 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 3857 " that may contain inline data"); 3858 goto failed_mount; 3859 } 3860 if (!bdev_dax_supported(sb->s_bdev, blocksize)) { 3861 ext4_msg(sb, KERN_ERR, 3862 "DAX unsupported by block device."); 3863 goto failed_mount; 3864 } 3865 } 3866 3867 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 3868 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 3869 es->s_encryption_level); 3870 goto failed_mount; 3871 } 3872 3873 if (sb->s_blocksize != blocksize) { 3874 /* Validate the filesystem blocksize */ 3875 if (!sb_set_blocksize(sb, blocksize)) { 3876 ext4_msg(sb, KERN_ERR, "bad block size %d", 3877 blocksize); 3878 goto failed_mount; 3879 } 3880 3881 brelse(bh); 3882 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3883 offset = do_div(logical_sb_block, blocksize); 3884 bh = sb_bread_unmovable(sb, logical_sb_block); 3885 if (!bh) { 3886 ext4_msg(sb, KERN_ERR, 3887 "Can't read superblock on 2nd try"); 3888 goto failed_mount; 3889 } 3890 es = (struct ext4_super_block *)(bh->b_data + offset); 3891 sbi->s_es = es; 3892 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3893 ext4_msg(sb, KERN_ERR, 3894 "Magic mismatch, very weird!"); 3895 goto failed_mount; 3896 } 3897 } 3898 3899 has_huge_files = ext4_has_feature_huge_file(sb); 3900 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3901 has_huge_files); 3902 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3903 3904 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3905 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3906 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3907 } else { 3908 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3909 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3910 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 3911 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 3912 sbi->s_first_ino); 3913 goto failed_mount; 3914 } 3915 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3916 (!is_power_of_2(sbi->s_inode_size)) || 3917 (sbi->s_inode_size > blocksize)) { 3918 ext4_msg(sb, KERN_ERR, 3919 "unsupported inode size: %d", 3920 sbi->s_inode_size); 3921 goto failed_mount; 3922 } 3923 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3924 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3925 } 3926 3927 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3928 if (ext4_has_feature_64bit(sb)) { 3929 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 3930 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3931 !is_power_of_2(sbi->s_desc_size)) { 3932 ext4_msg(sb, KERN_ERR, 3933 "unsupported descriptor size %lu", 3934 sbi->s_desc_size); 3935 goto failed_mount; 3936 } 3937 } else 3938 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 3939 3940 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3941 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3942 3943 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3944 if (sbi->s_inodes_per_block == 0) 3945 goto cantfind_ext4; 3946 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 3947 sbi->s_inodes_per_group > blocksize * 8) { 3948 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 3949 sbi->s_blocks_per_group); 3950 goto failed_mount; 3951 } 3952 sbi->s_itb_per_group = sbi->s_inodes_per_group / 3953 sbi->s_inodes_per_block; 3954 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3955 sbi->s_sbh = bh; 3956 sbi->s_mount_state = le16_to_cpu(es->s_state); 3957 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3958 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 3959 3960 for (i = 0; i < 4; i++) 3961 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3962 sbi->s_def_hash_version = es->s_def_hash_version; 3963 if (ext4_has_feature_dir_index(sb)) { 3964 i = le32_to_cpu(es->s_flags); 3965 if (i & EXT2_FLAGS_UNSIGNED_HASH) 3966 sbi->s_hash_unsigned = 3; 3967 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3968 #ifdef __CHAR_UNSIGNED__ 3969 if (!sb_rdonly(sb)) 3970 es->s_flags |= 3971 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3972 sbi->s_hash_unsigned = 3; 3973 #else 3974 if (!sb_rdonly(sb)) 3975 es->s_flags |= 3976 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3977 #endif 3978 } 3979 } 3980 3981 /* Handle clustersize */ 3982 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3983 has_bigalloc = ext4_has_feature_bigalloc(sb); 3984 if (has_bigalloc) { 3985 if (clustersize < blocksize) { 3986 ext4_msg(sb, KERN_ERR, 3987 "cluster size (%d) smaller than " 3988 "block size (%d)", clustersize, blocksize); 3989 goto failed_mount; 3990 } 3991 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3992 le32_to_cpu(es->s_log_block_size); 3993 sbi->s_clusters_per_group = 3994 le32_to_cpu(es->s_clusters_per_group); 3995 if (sbi->s_clusters_per_group > blocksize * 8) { 3996 ext4_msg(sb, KERN_ERR, 3997 "#clusters per group too big: %lu", 3998 sbi->s_clusters_per_group); 3999 goto failed_mount; 4000 } 4001 if (sbi->s_blocks_per_group != 4002 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4003 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4004 "clusters per group (%lu) inconsistent", 4005 sbi->s_blocks_per_group, 4006 sbi->s_clusters_per_group); 4007 goto failed_mount; 4008 } 4009 } else { 4010 if (clustersize != blocksize) { 4011 ext4_msg(sb, KERN_ERR, 4012 "fragment/cluster size (%d) != " 4013 "block size (%d)", clustersize, blocksize); 4014 goto failed_mount; 4015 } 4016 if (sbi->s_blocks_per_group > blocksize * 8) { 4017 ext4_msg(sb, KERN_ERR, 4018 "#blocks per group too big: %lu", 4019 sbi->s_blocks_per_group); 4020 goto failed_mount; 4021 } 4022 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4023 sbi->s_cluster_bits = 0; 4024 } 4025 sbi->s_cluster_ratio = clustersize / blocksize; 4026 4027 /* Do we have standard group size of clustersize * 8 blocks ? */ 4028 if (sbi->s_blocks_per_group == clustersize << 3) 4029 set_opt2(sb, STD_GROUP_SIZE); 4030 4031 /* 4032 * Test whether we have more sectors than will fit in sector_t, 4033 * and whether the max offset is addressable by the page cache. 4034 */ 4035 err = generic_check_addressable(sb->s_blocksize_bits, 4036 ext4_blocks_count(es)); 4037 if (err) { 4038 ext4_msg(sb, KERN_ERR, "filesystem" 4039 " too large to mount safely on this system"); 4040 if (sizeof(sector_t) < 8) 4041 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 4042 goto failed_mount; 4043 } 4044 4045 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4046 goto cantfind_ext4; 4047 4048 /* check blocks count against device size */ 4049 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 4050 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4051 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4052 "exceeds size of device (%llu blocks)", 4053 ext4_blocks_count(es), blocks_count); 4054 goto failed_mount; 4055 } 4056 4057 /* 4058 * It makes no sense for the first data block to be beyond the end 4059 * of the filesystem. 4060 */ 4061 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4062 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4063 "block %u is beyond end of filesystem (%llu)", 4064 le32_to_cpu(es->s_first_data_block), 4065 ext4_blocks_count(es)); 4066 goto failed_mount; 4067 } 4068 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4069 (sbi->s_cluster_ratio == 1)) { 4070 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4071 "block is 0 with a 1k block and cluster size"); 4072 goto failed_mount; 4073 } 4074 4075 blocks_count = (ext4_blocks_count(es) - 4076 le32_to_cpu(es->s_first_data_block) + 4077 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4078 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4079 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4080 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 4081 "(block count %llu, first data block %u, " 4082 "blocks per group %lu)", sbi->s_groups_count, 4083 ext4_blocks_count(es), 4084 le32_to_cpu(es->s_first_data_block), 4085 EXT4_BLOCKS_PER_GROUP(sb)); 4086 goto failed_mount; 4087 } 4088 sbi->s_groups_count = blocks_count; 4089 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4090 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4091 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4092 le32_to_cpu(es->s_inodes_count)) { 4093 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4094 le32_to_cpu(es->s_inodes_count), 4095 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4096 ret = -EINVAL; 4097 goto failed_mount; 4098 } 4099 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4100 EXT4_DESC_PER_BLOCK(sb); 4101 if (ext4_has_feature_meta_bg(sb)) { 4102 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4103 ext4_msg(sb, KERN_WARNING, 4104 "first meta block group too large: %u " 4105 "(group descriptor block count %u)", 4106 le32_to_cpu(es->s_first_meta_bg), db_count); 4107 goto failed_mount; 4108 } 4109 } 4110 sbi->s_group_desc = kvmalloc_array(db_count, 4111 sizeof(struct buffer_head *), 4112 GFP_KERNEL); 4113 if (sbi->s_group_desc == NULL) { 4114 ext4_msg(sb, KERN_ERR, "not enough memory"); 4115 ret = -ENOMEM; 4116 goto failed_mount; 4117 } 4118 4119 bgl_lock_init(sbi->s_blockgroup_lock); 4120 4121 /* Pre-read the descriptors into the buffer cache */ 4122 for (i = 0; i < db_count; i++) { 4123 block = descriptor_loc(sb, logical_sb_block, i); 4124 sb_breadahead(sb, block); 4125 } 4126 4127 for (i = 0; i < db_count; i++) { 4128 block = descriptor_loc(sb, logical_sb_block, i); 4129 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 4130 if (!sbi->s_group_desc[i]) { 4131 ext4_msg(sb, KERN_ERR, 4132 "can't read group descriptor %d", i); 4133 db_count = i; 4134 goto failed_mount2; 4135 } 4136 } 4137 sbi->s_gdb_count = db_count; 4138 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4139 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4140 ret = -EFSCORRUPTED; 4141 goto failed_mount2; 4142 } 4143 4144 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4145 4146 /* Register extent status tree shrinker */ 4147 if (ext4_es_register_shrinker(sbi)) 4148 goto failed_mount3; 4149 4150 sbi->s_stripe = ext4_get_stripe_size(sbi); 4151 sbi->s_extent_max_zeroout_kb = 32; 4152 4153 /* 4154 * set up enough so that it can read an inode 4155 */ 4156 sb->s_op = &ext4_sops; 4157 sb->s_export_op = &ext4_export_ops; 4158 sb->s_xattr = ext4_xattr_handlers; 4159 #ifdef CONFIG_EXT4_FS_ENCRYPTION 4160 sb->s_cop = &ext4_cryptops; 4161 #endif 4162 #ifdef CONFIG_QUOTA 4163 sb->dq_op = &ext4_quota_operations; 4164 if (ext4_has_feature_quota(sb)) 4165 sb->s_qcop = &dquot_quotactl_sysfile_ops; 4166 else 4167 sb->s_qcop = &ext4_qctl_operations; 4168 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4169 #endif 4170 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4171 4172 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 4173 mutex_init(&sbi->s_orphan_lock); 4174 4175 sb->s_root = NULL; 4176 4177 needs_recovery = (es->s_last_orphan != 0 || 4178 ext4_has_feature_journal_needs_recovery(sb)); 4179 4180 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4181 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 4182 goto failed_mount3a; 4183 4184 /* 4185 * The first inode we look at is the journal inode. Don't try 4186 * root first: it may be modified in the journal! 4187 */ 4188 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 4189 err = ext4_load_journal(sb, es, journal_devnum); 4190 if (err) 4191 goto failed_mount3a; 4192 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4193 ext4_has_feature_journal_needs_recovery(sb)) { 4194 ext4_msg(sb, KERN_ERR, "required journal recovery " 4195 "suppressed and not mounted read-only"); 4196 goto failed_mount_wq; 4197 } else { 4198 /* Nojournal mode, all journal mount options are illegal */ 4199 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 4200 ext4_msg(sb, KERN_ERR, "can't mount with " 4201 "journal_checksum, fs mounted w/o journal"); 4202 goto failed_mount_wq; 4203 } 4204 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4205 ext4_msg(sb, KERN_ERR, "can't mount with " 4206 "journal_async_commit, fs mounted w/o journal"); 4207 goto failed_mount_wq; 4208 } 4209 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 4210 ext4_msg(sb, KERN_ERR, "can't mount with " 4211 "commit=%lu, fs mounted w/o journal", 4212 sbi->s_commit_interval / HZ); 4213 goto failed_mount_wq; 4214 } 4215 if (EXT4_MOUNT_DATA_FLAGS & 4216 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 4217 ext4_msg(sb, KERN_ERR, "can't mount with " 4218 "data=, fs mounted w/o journal"); 4219 goto failed_mount_wq; 4220 } 4221 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM; 4222 clear_opt(sb, JOURNAL_CHECKSUM); 4223 clear_opt(sb, DATA_FLAGS); 4224 sbi->s_journal = NULL; 4225 needs_recovery = 0; 4226 goto no_journal; 4227 } 4228 4229 if (ext4_has_feature_64bit(sb) && 4230 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4231 JBD2_FEATURE_INCOMPAT_64BIT)) { 4232 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4233 goto failed_mount_wq; 4234 } 4235 4236 if (!set_journal_csum_feature_set(sb)) { 4237 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4238 "feature set"); 4239 goto failed_mount_wq; 4240 } 4241 4242 /* We have now updated the journal if required, so we can 4243 * validate the data journaling mode. */ 4244 switch (test_opt(sb, DATA_FLAGS)) { 4245 case 0: 4246 /* No mode set, assume a default based on the journal 4247 * capabilities: ORDERED_DATA if the journal can 4248 * cope, else JOURNAL_DATA 4249 */ 4250 if (jbd2_journal_check_available_features 4251 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4252 set_opt(sb, ORDERED_DATA); 4253 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4254 } else { 4255 set_opt(sb, JOURNAL_DATA); 4256 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4257 } 4258 break; 4259 4260 case EXT4_MOUNT_ORDERED_DATA: 4261 case EXT4_MOUNT_WRITEBACK_DATA: 4262 if (!jbd2_journal_check_available_features 4263 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4264 ext4_msg(sb, KERN_ERR, "Journal does not support " 4265 "requested data journaling mode"); 4266 goto failed_mount_wq; 4267 } 4268 default: 4269 break; 4270 } 4271 4272 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4273 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4274 ext4_msg(sb, KERN_ERR, "can't mount with " 4275 "journal_async_commit in data=ordered mode"); 4276 goto failed_mount_wq; 4277 } 4278 4279 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4280 4281 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4282 4283 no_journal: 4284 if (!test_opt(sb, NO_MBCACHE)) { 4285 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 4286 if (!sbi->s_ea_block_cache) { 4287 ext4_msg(sb, KERN_ERR, 4288 "Failed to create ea_block_cache"); 4289 goto failed_mount_wq; 4290 } 4291 4292 if (ext4_has_feature_ea_inode(sb)) { 4293 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4294 if (!sbi->s_ea_inode_cache) { 4295 ext4_msg(sb, KERN_ERR, 4296 "Failed to create ea_inode_cache"); 4297 goto failed_mount_wq; 4298 } 4299 } 4300 } 4301 4302 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && 4303 (blocksize != PAGE_SIZE)) { 4304 ext4_msg(sb, KERN_ERR, 4305 "Unsupported blocksize for fs encryption"); 4306 goto failed_mount_wq; 4307 } 4308 4309 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4310 !ext4_has_feature_encrypt(sb)) { 4311 ext4_set_feature_encrypt(sb); 4312 ext4_commit_super(sb, 1); 4313 } 4314 4315 /* 4316 * Get the # of file system overhead blocks from the 4317 * superblock if present. 4318 */ 4319 if (es->s_overhead_clusters) 4320 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4321 else { 4322 err = ext4_calculate_overhead(sb); 4323 if (err) 4324 goto failed_mount_wq; 4325 } 4326 4327 /* 4328 * The maximum number of concurrent works can be high and 4329 * concurrency isn't really necessary. Limit it to 1. 4330 */ 4331 EXT4_SB(sb)->rsv_conversion_wq = 4332 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4333 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4334 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4335 ret = -ENOMEM; 4336 goto failed_mount4; 4337 } 4338 4339 /* 4340 * The jbd2_journal_load will have done any necessary log recovery, 4341 * so we can safely mount the rest of the filesystem now. 4342 */ 4343 4344 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 4345 if (IS_ERR(root)) { 4346 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4347 ret = PTR_ERR(root); 4348 root = NULL; 4349 goto failed_mount4; 4350 } 4351 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4352 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4353 iput(root); 4354 goto failed_mount4; 4355 } 4356 sb->s_root = d_make_root(root); 4357 if (!sb->s_root) { 4358 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4359 ret = -ENOMEM; 4360 goto failed_mount4; 4361 } 4362 4363 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4364 if (ret == -EROFS) { 4365 sb->s_flags |= SB_RDONLY; 4366 ret = 0; 4367 } else if (ret) 4368 goto failed_mount4a; 4369 4370 /* determine the minimum size of new large inodes, if present */ 4371 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE && 4372 sbi->s_want_extra_isize == 0) { 4373 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4374 EXT4_GOOD_OLD_INODE_SIZE; 4375 if (ext4_has_feature_extra_isize(sb)) { 4376 if (sbi->s_want_extra_isize < 4377 le16_to_cpu(es->s_want_extra_isize)) 4378 sbi->s_want_extra_isize = 4379 le16_to_cpu(es->s_want_extra_isize); 4380 if (sbi->s_want_extra_isize < 4381 le16_to_cpu(es->s_min_extra_isize)) 4382 sbi->s_want_extra_isize = 4383 le16_to_cpu(es->s_min_extra_isize); 4384 } 4385 } 4386 /* Check if enough inode space is available */ 4387 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 4388 sbi->s_inode_size) { 4389 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4390 EXT4_GOOD_OLD_INODE_SIZE; 4391 ext4_msg(sb, KERN_INFO, "required extra inode space not" 4392 "available"); 4393 } 4394 4395 ext4_set_resv_clusters(sb); 4396 4397 err = ext4_setup_system_zone(sb); 4398 if (err) { 4399 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4400 "zone (%d)", err); 4401 goto failed_mount4a; 4402 } 4403 4404 ext4_ext_init(sb); 4405 err = ext4_mb_init(sb); 4406 if (err) { 4407 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4408 err); 4409 goto failed_mount5; 4410 } 4411 4412 block = ext4_count_free_clusters(sb); 4413 ext4_free_blocks_count_set(sbi->s_es, 4414 EXT4_C2B(sbi, block)); 4415 ext4_superblock_csum_set(sb); 4416 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4417 GFP_KERNEL); 4418 if (!err) { 4419 unsigned long freei = ext4_count_free_inodes(sb); 4420 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4421 ext4_superblock_csum_set(sb); 4422 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4423 GFP_KERNEL); 4424 } 4425 if (!err) 4426 err = percpu_counter_init(&sbi->s_dirs_counter, 4427 ext4_count_dirs(sb), GFP_KERNEL); 4428 if (!err) 4429 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4430 GFP_KERNEL); 4431 if (!err) 4432 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 4433 4434 if (err) { 4435 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4436 goto failed_mount6; 4437 } 4438 4439 if (ext4_has_feature_flex_bg(sb)) 4440 if (!ext4_fill_flex_info(sb)) { 4441 ext4_msg(sb, KERN_ERR, 4442 "unable to initialize " 4443 "flex_bg meta info!"); 4444 goto failed_mount6; 4445 } 4446 4447 err = ext4_register_li_request(sb, first_not_zeroed); 4448 if (err) 4449 goto failed_mount6; 4450 4451 err = ext4_register_sysfs(sb); 4452 if (err) 4453 goto failed_mount7; 4454 4455 #ifdef CONFIG_QUOTA 4456 /* Enable quota usage during mount. */ 4457 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 4458 err = ext4_enable_quotas(sb); 4459 if (err) 4460 goto failed_mount8; 4461 } 4462 #endif /* CONFIG_QUOTA */ 4463 4464 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4465 ext4_orphan_cleanup(sb, es); 4466 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4467 if (needs_recovery) { 4468 ext4_msg(sb, KERN_INFO, "recovery complete"); 4469 ext4_mark_recovery_complete(sb, es); 4470 } 4471 if (EXT4_SB(sb)->s_journal) { 4472 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4473 descr = " journalled data mode"; 4474 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4475 descr = " ordered data mode"; 4476 else 4477 descr = " writeback data mode"; 4478 } else 4479 descr = "out journal"; 4480 4481 if (test_opt(sb, DISCARD)) { 4482 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4483 if (!blk_queue_discard(q)) 4484 ext4_msg(sb, KERN_WARNING, 4485 "mounting with \"discard\" option, but " 4486 "the device does not support discard"); 4487 } 4488 4489 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4490 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4491 "Opts: %.*s%s%s", descr, 4492 (int) sizeof(sbi->s_es->s_mount_opts), 4493 sbi->s_es->s_mount_opts, 4494 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4495 4496 if (es->s_error_count) 4497 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4498 4499 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4500 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4501 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4502 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4503 4504 kfree(orig_data); 4505 return 0; 4506 4507 cantfind_ext4: 4508 if (!silent) 4509 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4510 goto failed_mount; 4511 4512 #ifdef CONFIG_QUOTA 4513 failed_mount8: 4514 ext4_unregister_sysfs(sb); 4515 #endif 4516 failed_mount7: 4517 ext4_unregister_li_request(sb); 4518 failed_mount6: 4519 ext4_mb_release(sb); 4520 if (sbi->s_flex_groups) 4521 kvfree(sbi->s_flex_groups); 4522 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4523 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4524 percpu_counter_destroy(&sbi->s_dirs_counter); 4525 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4526 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 4527 failed_mount5: 4528 ext4_ext_release(sb); 4529 ext4_release_system_zone(sb); 4530 failed_mount4a: 4531 dput(sb->s_root); 4532 sb->s_root = NULL; 4533 failed_mount4: 4534 ext4_msg(sb, KERN_ERR, "mount failed"); 4535 if (EXT4_SB(sb)->rsv_conversion_wq) 4536 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4537 failed_mount_wq: 4538 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4539 sbi->s_ea_inode_cache = NULL; 4540 4541 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 4542 sbi->s_ea_block_cache = NULL; 4543 4544 if (sbi->s_journal) { 4545 jbd2_journal_destroy(sbi->s_journal); 4546 sbi->s_journal = NULL; 4547 } 4548 failed_mount3a: 4549 ext4_es_unregister_shrinker(sbi); 4550 failed_mount3: 4551 del_timer_sync(&sbi->s_err_report); 4552 if (sbi->s_mmp_tsk) 4553 kthread_stop(sbi->s_mmp_tsk); 4554 failed_mount2: 4555 for (i = 0; i < db_count; i++) 4556 brelse(sbi->s_group_desc[i]); 4557 kvfree(sbi->s_group_desc); 4558 failed_mount: 4559 if (sbi->s_chksum_driver) 4560 crypto_free_shash(sbi->s_chksum_driver); 4561 #ifdef CONFIG_QUOTA 4562 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4563 kfree(sbi->s_qf_names[i]); 4564 #endif 4565 ext4_blkdev_remove(sbi); 4566 brelse(bh); 4567 out_fail: 4568 sb->s_fs_info = NULL; 4569 kfree(sbi->s_blockgroup_lock); 4570 out_free_base: 4571 kfree(sbi); 4572 kfree(orig_data); 4573 fs_put_dax(dax_dev); 4574 return err ? err : ret; 4575 } 4576 4577 /* 4578 * Setup any per-fs journal parameters now. We'll do this both on 4579 * initial mount, once the journal has been initialised but before we've 4580 * done any recovery; and again on any subsequent remount. 4581 */ 4582 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4583 { 4584 struct ext4_sb_info *sbi = EXT4_SB(sb); 4585 4586 journal->j_commit_interval = sbi->s_commit_interval; 4587 journal->j_min_batch_time = sbi->s_min_batch_time; 4588 journal->j_max_batch_time = sbi->s_max_batch_time; 4589 4590 write_lock(&journal->j_state_lock); 4591 if (test_opt(sb, BARRIER)) 4592 journal->j_flags |= JBD2_BARRIER; 4593 else 4594 journal->j_flags &= ~JBD2_BARRIER; 4595 if (test_opt(sb, DATA_ERR_ABORT)) 4596 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4597 else 4598 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4599 write_unlock(&journal->j_state_lock); 4600 } 4601 4602 static struct inode *ext4_get_journal_inode(struct super_block *sb, 4603 unsigned int journal_inum) 4604 { 4605 struct inode *journal_inode; 4606 4607 /* 4608 * Test for the existence of a valid inode on disk. Bad things 4609 * happen if we iget() an unused inode, as the subsequent iput() 4610 * will try to delete it. 4611 */ 4612 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 4613 if (IS_ERR(journal_inode)) { 4614 ext4_msg(sb, KERN_ERR, "no journal found"); 4615 return NULL; 4616 } 4617 if (!journal_inode->i_nlink) { 4618 make_bad_inode(journal_inode); 4619 iput(journal_inode); 4620 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4621 return NULL; 4622 } 4623 4624 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4625 journal_inode, journal_inode->i_size); 4626 if (!S_ISREG(journal_inode->i_mode)) { 4627 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4628 iput(journal_inode); 4629 return NULL; 4630 } 4631 return journal_inode; 4632 } 4633 4634 static journal_t *ext4_get_journal(struct super_block *sb, 4635 unsigned int journal_inum) 4636 { 4637 struct inode *journal_inode; 4638 journal_t *journal; 4639 4640 BUG_ON(!ext4_has_feature_journal(sb)); 4641 4642 journal_inode = ext4_get_journal_inode(sb, journal_inum); 4643 if (!journal_inode) 4644 return NULL; 4645 4646 journal = jbd2_journal_init_inode(journal_inode); 4647 if (!journal) { 4648 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4649 iput(journal_inode); 4650 return NULL; 4651 } 4652 journal->j_private = sb; 4653 ext4_init_journal_params(sb, journal); 4654 return journal; 4655 } 4656 4657 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4658 dev_t j_dev) 4659 { 4660 struct buffer_head *bh; 4661 journal_t *journal; 4662 ext4_fsblk_t start; 4663 ext4_fsblk_t len; 4664 int hblock, blocksize; 4665 ext4_fsblk_t sb_block; 4666 unsigned long offset; 4667 struct ext4_super_block *es; 4668 struct block_device *bdev; 4669 4670 BUG_ON(!ext4_has_feature_journal(sb)); 4671 4672 bdev = ext4_blkdev_get(j_dev, sb); 4673 if (bdev == NULL) 4674 return NULL; 4675 4676 blocksize = sb->s_blocksize; 4677 hblock = bdev_logical_block_size(bdev); 4678 if (blocksize < hblock) { 4679 ext4_msg(sb, KERN_ERR, 4680 "blocksize too small for journal device"); 4681 goto out_bdev; 4682 } 4683 4684 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4685 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4686 set_blocksize(bdev, blocksize); 4687 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4688 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4689 "external journal"); 4690 goto out_bdev; 4691 } 4692 4693 es = (struct ext4_super_block *) (bh->b_data + offset); 4694 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4695 !(le32_to_cpu(es->s_feature_incompat) & 4696 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4697 ext4_msg(sb, KERN_ERR, "external journal has " 4698 "bad superblock"); 4699 brelse(bh); 4700 goto out_bdev; 4701 } 4702 4703 if ((le32_to_cpu(es->s_feature_ro_compat) & 4704 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4705 es->s_checksum != ext4_superblock_csum(sb, es)) { 4706 ext4_msg(sb, KERN_ERR, "external journal has " 4707 "corrupt superblock"); 4708 brelse(bh); 4709 goto out_bdev; 4710 } 4711 4712 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4713 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4714 brelse(bh); 4715 goto out_bdev; 4716 } 4717 4718 len = ext4_blocks_count(es); 4719 start = sb_block + 1; 4720 brelse(bh); /* we're done with the superblock */ 4721 4722 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4723 start, len, blocksize); 4724 if (!journal) { 4725 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4726 goto out_bdev; 4727 } 4728 journal->j_private = sb; 4729 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4730 wait_on_buffer(journal->j_sb_buffer); 4731 if (!buffer_uptodate(journal->j_sb_buffer)) { 4732 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4733 goto out_journal; 4734 } 4735 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4736 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4737 "user (unsupported) - %d", 4738 be32_to_cpu(journal->j_superblock->s_nr_users)); 4739 goto out_journal; 4740 } 4741 EXT4_SB(sb)->journal_bdev = bdev; 4742 ext4_init_journal_params(sb, journal); 4743 return journal; 4744 4745 out_journal: 4746 jbd2_journal_destroy(journal); 4747 out_bdev: 4748 ext4_blkdev_put(bdev); 4749 return NULL; 4750 } 4751 4752 static int ext4_load_journal(struct super_block *sb, 4753 struct ext4_super_block *es, 4754 unsigned long journal_devnum) 4755 { 4756 journal_t *journal; 4757 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4758 dev_t journal_dev; 4759 int err = 0; 4760 int really_read_only; 4761 4762 BUG_ON(!ext4_has_feature_journal(sb)); 4763 4764 if (journal_devnum && 4765 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4766 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4767 "numbers have changed"); 4768 journal_dev = new_decode_dev(journal_devnum); 4769 } else 4770 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4771 4772 really_read_only = bdev_read_only(sb->s_bdev); 4773 4774 /* 4775 * Are we loading a blank journal or performing recovery after a 4776 * crash? For recovery, we need to check in advance whether we 4777 * can get read-write access to the device. 4778 */ 4779 if (ext4_has_feature_journal_needs_recovery(sb)) { 4780 if (sb_rdonly(sb)) { 4781 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4782 "required on readonly filesystem"); 4783 if (really_read_only) { 4784 ext4_msg(sb, KERN_ERR, "write access " 4785 "unavailable, cannot proceed " 4786 "(try mounting with noload)"); 4787 return -EROFS; 4788 } 4789 ext4_msg(sb, KERN_INFO, "write access will " 4790 "be enabled during recovery"); 4791 } 4792 } 4793 4794 if (journal_inum && journal_dev) { 4795 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4796 "and inode journals!"); 4797 return -EINVAL; 4798 } 4799 4800 if (journal_inum) { 4801 if (!(journal = ext4_get_journal(sb, journal_inum))) 4802 return -EINVAL; 4803 } else { 4804 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4805 return -EINVAL; 4806 } 4807 4808 if (!(journal->j_flags & JBD2_BARRIER)) 4809 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4810 4811 if (!ext4_has_feature_journal_needs_recovery(sb)) 4812 err = jbd2_journal_wipe(journal, !really_read_only); 4813 if (!err) { 4814 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4815 if (save) 4816 memcpy(save, ((char *) es) + 4817 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4818 err = jbd2_journal_load(journal); 4819 if (save) 4820 memcpy(((char *) es) + EXT4_S_ERR_START, 4821 save, EXT4_S_ERR_LEN); 4822 kfree(save); 4823 } 4824 4825 if (err) { 4826 ext4_msg(sb, KERN_ERR, "error loading journal"); 4827 jbd2_journal_destroy(journal); 4828 return err; 4829 } 4830 4831 EXT4_SB(sb)->s_journal = journal; 4832 ext4_clear_journal_err(sb, es); 4833 4834 if (!really_read_only && journal_devnum && 4835 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4836 es->s_journal_dev = cpu_to_le32(journal_devnum); 4837 4838 /* Make sure we flush the recovery flag to disk. */ 4839 ext4_commit_super(sb, 1); 4840 } 4841 4842 return 0; 4843 } 4844 4845 static int ext4_commit_super(struct super_block *sb, int sync) 4846 { 4847 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4848 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4849 int error = 0; 4850 4851 if (!sbh || block_device_ejected(sb)) 4852 return error; 4853 4854 /* 4855 * The superblock bh should be mapped, but it might not be if the 4856 * device was hot-removed. Not much we can do but fail the I/O. 4857 */ 4858 if (!buffer_mapped(sbh)) 4859 return error; 4860 4861 /* 4862 * If the file system is mounted read-only, don't update the 4863 * superblock write time. This avoids updating the superblock 4864 * write time when we are mounting the root file system 4865 * read/only but we need to replay the journal; at that point, 4866 * for people who are east of GMT and who make their clock 4867 * tick in localtime for Windows bug-for-bug compatibility, 4868 * the clock is set in the future, and this will cause e2fsck 4869 * to complain and force a full file system check. 4870 */ 4871 if (!(sb->s_flags & SB_RDONLY)) 4872 ext4_update_tstamp(es, s_wtime); 4873 if (sb->s_bdev->bd_part) 4874 es->s_kbytes_written = 4875 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4876 ((part_stat_read(sb->s_bdev->bd_part, 4877 sectors[STAT_WRITE]) - 4878 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4879 else 4880 es->s_kbytes_written = 4881 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4882 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 4883 ext4_free_blocks_count_set(es, 4884 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 4885 &EXT4_SB(sb)->s_freeclusters_counter))); 4886 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 4887 es->s_free_inodes_count = 4888 cpu_to_le32(percpu_counter_sum_positive( 4889 &EXT4_SB(sb)->s_freeinodes_counter)); 4890 BUFFER_TRACE(sbh, "marking dirty"); 4891 ext4_superblock_csum_set(sb); 4892 if (sync) 4893 lock_buffer(sbh); 4894 if (buffer_write_io_error(sbh)) { 4895 /* 4896 * Oh, dear. A previous attempt to write the 4897 * superblock failed. This could happen because the 4898 * USB device was yanked out. Or it could happen to 4899 * be a transient write error and maybe the block will 4900 * be remapped. Nothing we can do but to retry the 4901 * write and hope for the best. 4902 */ 4903 ext4_msg(sb, KERN_ERR, "previous I/O error to " 4904 "superblock detected"); 4905 clear_buffer_write_io_error(sbh); 4906 set_buffer_uptodate(sbh); 4907 } 4908 mark_buffer_dirty(sbh); 4909 if (sync) { 4910 unlock_buffer(sbh); 4911 error = __sync_dirty_buffer(sbh, 4912 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 4913 if (buffer_write_io_error(sbh)) { 4914 ext4_msg(sb, KERN_ERR, "I/O error while writing " 4915 "superblock"); 4916 clear_buffer_write_io_error(sbh); 4917 set_buffer_uptodate(sbh); 4918 } 4919 } 4920 return error; 4921 } 4922 4923 /* 4924 * Have we just finished recovery? If so, and if we are mounting (or 4925 * remounting) the filesystem readonly, then we will end up with a 4926 * consistent fs on disk. Record that fact. 4927 */ 4928 static void ext4_mark_recovery_complete(struct super_block *sb, 4929 struct ext4_super_block *es) 4930 { 4931 journal_t *journal = EXT4_SB(sb)->s_journal; 4932 4933 if (!ext4_has_feature_journal(sb)) { 4934 BUG_ON(journal != NULL); 4935 return; 4936 } 4937 jbd2_journal_lock_updates(journal); 4938 if (jbd2_journal_flush(journal) < 0) 4939 goto out; 4940 4941 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 4942 ext4_clear_feature_journal_needs_recovery(sb); 4943 ext4_commit_super(sb, 1); 4944 } 4945 4946 out: 4947 jbd2_journal_unlock_updates(journal); 4948 } 4949 4950 /* 4951 * If we are mounting (or read-write remounting) a filesystem whose journal 4952 * has recorded an error from a previous lifetime, move that error to the 4953 * main filesystem now. 4954 */ 4955 static void ext4_clear_journal_err(struct super_block *sb, 4956 struct ext4_super_block *es) 4957 { 4958 journal_t *journal; 4959 int j_errno; 4960 const char *errstr; 4961 4962 BUG_ON(!ext4_has_feature_journal(sb)); 4963 4964 journal = EXT4_SB(sb)->s_journal; 4965 4966 /* 4967 * Now check for any error status which may have been recorded in the 4968 * journal by a prior ext4_error() or ext4_abort() 4969 */ 4970 4971 j_errno = jbd2_journal_errno(journal); 4972 if (j_errno) { 4973 char nbuf[16]; 4974 4975 errstr = ext4_decode_error(sb, j_errno, nbuf); 4976 ext4_warning(sb, "Filesystem error recorded " 4977 "from previous mount: %s", errstr); 4978 ext4_warning(sb, "Marking fs in need of filesystem check."); 4979 4980 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4981 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4982 ext4_commit_super(sb, 1); 4983 4984 jbd2_journal_clear_err(journal); 4985 jbd2_journal_update_sb_errno(journal); 4986 } 4987 } 4988 4989 /* 4990 * Force the running and committing transactions to commit, 4991 * and wait on the commit. 4992 */ 4993 int ext4_force_commit(struct super_block *sb) 4994 { 4995 journal_t *journal; 4996 4997 if (sb_rdonly(sb)) 4998 return 0; 4999 5000 journal = EXT4_SB(sb)->s_journal; 5001 return ext4_journal_force_commit(journal); 5002 } 5003 5004 static int ext4_sync_fs(struct super_block *sb, int wait) 5005 { 5006 int ret = 0; 5007 tid_t target; 5008 bool needs_barrier = false; 5009 struct ext4_sb_info *sbi = EXT4_SB(sb); 5010 5011 if (unlikely(ext4_forced_shutdown(sbi))) 5012 return 0; 5013 5014 trace_ext4_sync_fs(sb, wait); 5015 flush_workqueue(sbi->rsv_conversion_wq); 5016 /* 5017 * Writeback quota in non-journalled quota case - journalled quota has 5018 * no dirty dquots 5019 */ 5020 dquot_writeback_dquots(sb, -1); 5021 /* 5022 * Data writeback is possible w/o journal transaction, so barrier must 5023 * being sent at the end of the function. But we can skip it if 5024 * transaction_commit will do it for us. 5025 */ 5026 if (sbi->s_journal) { 5027 target = jbd2_get_latest_transaction(sbi->s_journal); 5028 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 5029 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 5030 needs_barrier = true; 5031 5032 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5033 if (wait) 5034 ret = jbd2_log_wait_commit(sbi->s_journal, 5035 target); 5036 } 5037 } else if (wait && test_opt(sb, BARRIER)) 5038 needs_barrier = true; 5039 if (needs_barrier) { 5040 int err; 5041 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 5042 if (!ret) 5043 ret = err; 5044 } 5045 5046 return ret; 5047 } 5048 5049 /* 5050 * LVM calls this function before a (read-only) snapshot is created. This 5051 * gives us a chance to flush the journal completely and mark the fs clean. 5052 * 5053 * Note that only this function cannot bring a filesystem to be in a clean 5054 * state independently. It relies on upper layer to stop all data & metadata 5055 * modifications. 5056 */ 5057 static int ext4_freeze(struct super_block *sb) 5058 { 5059 int error = 0; 5060 journal_t *journal; 5061 5062 if (sb_rdonly(sb)) 5063 return 0; 5064 5065 journal = EXT4_SB(sb)->s_journal; 5066 5067 if (journal) { 5068 /* Now we set up the journal barrier. */ 5069 jbd2_journal_lock_updates(journal); 5070 5071 /* 5072 * Don't clear the needs_recovery flag if we failed to 5073 * flush the journal. 5074 */ 5075 error = jbd2_journal_flush(journal); 5076 if (error < 0) 5077 goto out; 5078 5079 /* Journal blocked and flushed, clear needs_recovery flag. */ 5080 ext4_clear_feature_journal_needs_recovery(sb); 5081 } 5082 5083 error = ext4_commit_super(sb, 1); 5084 out: 5085 if (journal) 5086 /* we rely on upper layer to stop further updates */ 5087 jbd2_journal_unlock_updates(journal); 5088 return error; 5089 } 5090 5091 /* 5092 * Called by LVM after the snapshot is done. We need to reset the RECOVER 5093 * flag here, even though the filesystem is not technically dirty yet. 5094 */ 5095 static int ext4_unfreeze(struct super_block *sb) 5096 { 5097 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 5098 return 0; 5099 5100 if (EXT4_SB(sb)->s_journal) { 5101 /* Reset the needs_recovery flag before the fs is unlocked. */ 5102 ext4_set_feature_journal_needs_recovery(sb); 5103 } 5104 5105 ext4_commit_super(sb, 1); 5106 return 0; 5107 } 5108 5109 /* 5110 * Structure to save mount options for ext4_remount's benefit 5111 */ 5112 struct ext4_mount_options { 5113 unsigned long s_mount_opt; 5114 unsigned long s_mount_opt2; 5115 kuid_t s_resuid; 5116 kgid_t s_resgid; 5117 unsigned long s_commit_interval; 5118 u32 s_min_batch_time, s_max_batch_time; 5119 #ifdef CONFIG_QUOTA 5120 int s_jquota_fmt; 5121 char *s_qf_names[EXT4_MAXQUOTAS]; 5122 #endif 5123 }; 5124 5125 static int ext4_remount(struct super_block *sb, int *flags, char *data) 5126 { 5127 struct ext4_super_block *es; 5128 struct ext4_sb_info *sbi = EXT4_SB(sb); 5129 unsigned long old_sb_flags; 5130 struct ext4_mount_options old_opts; 5131 int enable_quota = 0; 5132 ext4_group_t g; 5133 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5134 int err = 0; 5135 #ifdef CONFIG_QUOTA 5136 int i, j; 5137 char *to_free[EXT4_MAXQUOTAS]; 5138 #endif 5139 char *orig_data = kstrdup(data, GFP_KERNEL); 5140 5141 if (data && !orig_data) 5142 return -ENOMEM; 5143 5144 /* Store the original options */ 5145 old_sb_flags = sb->s_flags; 5146 old_opts.s_mount_opt = sbi->s_mount_opt; 5147 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5148 old_opts.s_resuid = sbi->s_resuid; 5149 old_opts.s_resgid = sbi->s_resgid; 5150 old_opts.s_commit_interval = sbi->s_commit_interval; 5151 old_opts.s_min_batch_time = sbi->s_min_batch_time; 5152 old_opts.s_max_batch_time = sbi->s_max_batch_time; 5153 #ifdef CONFIG_QUOTA 5154 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5155 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5156 if (sbi->s_qf_names[i]) { 5157 char *qf_name = get_qf_name(sb, sbi, i); 5158 5159 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 5160 if (!old_opts.s_qf_names[i]) { 5161 for (j = 0; j < i; j++) 5162 kfree(old_opts.s_qf_names[j]); 5163 kfree(orig_data); 5164 return -ENOMEM; 5165 } 5166 } else 5167 old_opts.s_qf_names[i] = NULL; 5168 #endif 5169 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5170 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5171 5172 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5173 err = -EINVAL; 5174 goto restore_opts; 5175 } 5176 5177 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 5178 test_opt(sb, JOURNAL_CHECKSUM)) { 5179 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 5180 "during remount not supported; ignoring"); 5181 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5182 } 5183 5184 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5185 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5186 ext4_msg(sb, KERN_ERR, "can't mount with " 5187 "both data=journal and delalloc"); 5188 err = -EINVAL; 5189 goto restore_opts; 5190 } 5191 if (test_opt(sb, DIOREAD_NOLOCK)) { 5192 ext4_msg(sb, KERN_ERR, "can't mount with " 5193 "both data=journal and dioread_nolock"); 5194 err = -EINVAL; 5195 goto restore_opts; 5196 } 5197 if (test_opt(sb, DAX)) { 5198 ext4_msg(sb, KERN_ERR, "can't mount with " 5199 "both data=journal and dax"); 5200 err = -EINVAL; 5201 goto restore_opts; 5202 } 5203 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5204 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5205 ext4_msg(sb, KERN_ERR, "can't mount with " 5206 "journal_async_commit in data=ordered mode"); 5207 err = -EINVAL; 5208 goto restore_opts; 5209 } 5210 } 5211 5212 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5213 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5214 err = -EINVAL; 5215 goto restore_opts; 5216 } 5217 5218 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 5219 ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 5220 "dax flag with busy inodes while remounting"); 5221 sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 5222 } 5223 5224 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 5225 ext4_abort(sb, "Abort forced by user"); 5226 5227 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5228 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5229 5230 es = sbi->s_es; 5231 5232 if (sbi->s_journal) { 5233 ext4_init_journal_params(sb, sbi->s_journal); 5234 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5235 } 5236 5237 if (*flags & SB_LAZYTIME) 5238 sb->s_flags |= SB_LAZYTIME; 5239 5240 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 5241 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5242 err = -EROFS; 5243 goto restore_opts; 5244 } 5245 5246 if (*flags & SB_RDONLY) { 5247 err = sync_filesystem(sb); 5248 if (err < 0) 5249 goto restore_opts; 5250 err = dquot_suspend(sb, -1); 5251 if (err < 0) 5252 goto restore_opts; 5253 5254 /* 5255 * First of all, the unconditional stuff we have to do 5256 * to disable replay of the journal when we next remount 5257 */ 5258 sb->s_flags |= SB_RDONLY; 5259 5260 /* 5261 * OK, test if we are remounting a valid rw partition 5262 * readonly, and if so set the rdonly flag and then 5263 * mark the partition as valid again. 5264 */ 5265 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5266 (sbi->s_mount_state & EXT4_VALID_FS)) 5267 es->s_state = cpu_to_le16(sbi->s_mount_state); 5268 5269 if (sbi->s_journal) 5270 ext4_mark_recovery_complete(sb, es); 5271 if (sbi->s_mmp_tsk) 5272 kthread_stop(sbi->s_mmp_tsk); 5273 } else { 5274 /* Make sure we can mount this feature set readwrite */ 5275 if (ext4_has_feature_readonly(sb) || 5276 !ext4_feature_set_ok(sb, 0)) { 5277 err = -EROFS; 5278 goto restore_opts; 5279 } 5280 /* 5281 * Make sure the group descriptor checksums 5282 * are sane. If they aren't, refuse to remount r/w. 5283 */ 5284 for (g = 0; g < sbi->s_groups_count; g++) { 5285 struct ext4_group_desc *gdp = 5286 ext4_get_group_desc(sb, g, NULL); 5287 5288 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5289 ext4_msg(sb, KERN_ERR, 5290 "ext4_remount: Checksum for group %u failed (%u!=%u)", 5291 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 5292 le16_to_cpu(gdp->bg_checksum)); 5293 err = -EFSBADCRC; 5294 goto restore_opts; 5295 } 5296 } 5297 5298 /* 5299 * If we have an unprocessed orphan list hanging 5300 * around from a previously readonly bdev mount, 5301 * require a full umount/remount for now. 5302 */ 5303 if (es->s_last_orphan) { 5304 ext4_msg(sb, KERN_WARNING, "Couldn't " 5305 "remount RDWR because of unprocessed " 5306 "orphan inode list. Please " 5307 "umount/remount instead"); 5308 err = -EINVAL; 5309 goto restore_opts; 5310 } 5311 5312 /* 5313 * Mounting a RDONLY partition read-write, so reread 5314 * and store the current valid flag. (It may have 5315 * been changed by e2fsck since we originally mounted 5316 * the partition.) 5317 */ 5318 if (sbi->s_journal) 5319 ext4_clear_journal_err(sb, es); 5320 sbi->s_mount_state = le16_to_cpu(es->s_state); 5321 5322 err = ext4_setup_super(sb, es, 0); 5323 if (err) 5324 goto restore_opts; 5325 5326 sb->s_flags &= ~SB_RDONLY; 5327 if (ext4_has_feature_mmp(sb)) 5328 if (ext4_multi_mount_protect(sb, 5329 le64_to_cpu(es->s_mmp_block))) { 5330 err = -EROFS; 5331 goto restore_opts; 5332 } 5333 enable_quota = 1; 5334 } 5335 } 5336 5337 /* 5338 * Reinitialize lazy itable initialization thread based on 5339 * current settings 5340 */ 5341 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5342 ext4_unregister_li_request(sb); 5343 else { 5344 ext4_group_t first_not_zeroed; 5345 first_not_zeroed = ext4_has_uninit_itable(sb); 5346 ext4_register_li_request(sb, first_not_zeroed); 5347 } 5348 5349 ext4_setup_system_zone(sb); 5350 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5351 err = ext4_commit_super(sb, 1); 5352 if (err) 5353 goto restore_opts; 5354 } 5355 5356 #ifdef CONFIG_QUOTA 5357 /* Release old quota file names */ 5358 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5359 kfree(old_opts.s_qf_names[i]); 5360 if (enable_quota) { 5361 if (sb_any_quota_suspended(sb)) 5362 dquot_resume(sb, -1); 5363 else if (ext4_has_feature_quota(sb)) { 5364 err = ext4_enable_quotas(sb); 5365 if (err) 5366 goto restore_opts; 5367 } 5368 } 5369 #endif 5370 5371 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 5372 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5373 kfree(orig_data); 5374 return 0; 5375 5376 restore_opts: 5377 sb->s_flags = old_sb_flags; 5378 sbi->s_mount_opt = old_opts.s_mount_opt; 5379 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5380 sbi->s_resuid = old_opts.s_resuid; 5381 sbi->s_resgid = old_opts.s_resgid; 5382 sbi->s_commit_interval = old_opts.s_commit_interval; 5383 sbi->s_min_batch_time = old_opts.s_min_batch_time; 5384 sbi->s_max_batch_time = old_opts.s_max_batch_time; 5385 #ifdef CONFIG_QUOTA 5386 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5387 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 5388 to_free[i] = get_qf_name(sb, sbi, i); 5389 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5390 } 5391 synchronize_rcu(); 5392 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5393 kfree(to_free[i]); 5394 #endif 5395 kfree(orig_data); 5396 return err; 5397 } 5398 5399 #ifdef CONFIG_QUOTA 5400 static int ext4_statfs_project(struct super_block *sb, 5401 kprojid_t projid, struct kstatfs *buf) 5402 { 5403 struct kqid qid; 5404 struct dquot *dquot; 5405 u64 limit; 5406 u64 curblock; 5407 5408 qid = make_kqid_projid(projid); 5409 dquot = dqget(sb, qid); 5410 if (IS_ERR(dquot)) 5411 return PTR_ERR(dquot); 5412 spin_lock(&dquot->dq_dqb_lock); 5413 5414 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 5415 dquot->dq_dqb.dqb_bsoftlimit : 5416 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 5417 if (limit && buf->f_blocks > limit) { 5418 curblock = (dquot->dq_dqb.dqb_curspace + 5419 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5420 buf->f_blocks = limit; 5421 buf->f_bfree = buf->f_bavail = 5422 (buf->f_blocks > curblock) ? 5423 (buf->f_blocks - curblock) : 0; 5424 } 5425 5426 limit = dquot->dq_dqb.dqb_isoftlimit ? 5427 dquot->dq_dqb.dqb_isoftlimit : 5428 dquot->dq_dqb.dqb_ihardlimit; 5429 if (limit && buf->f_files > limit) { 5430 buf->f_files = limit; 5431 buf->f_ffree = 5432 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5433 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5434 } 5435 5436 spin_unlock(&dquot->dq_dqb_lock); 5437 dqput(dquot); 5438 return 0; 5439 } 5440 #endif 5441 5442 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5443 { 5444 struct super_block *sb = dentry->d_sb; 5445 struct ext4_sb_info *sbi = EXT4_SB(sb); 5446 struct ext4_super_block *es = sbi->s_es; 5447 ext4_fsblk_t overhead = 0, resv_blocks; 5448 u64 fsid; 5449 s64 bfree; 5450 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5451 5452 if (!test_opt(sb, MINIX_DF)) 5453 overhead = sbi->s_overhead; 5454 5455 buf->f_type = EXT4_SUPER_MAGIC; 5456 buf->f_bsize = sb->s_blocksize; 5457 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5458 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5459 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5460 /* prevent underflow in case that few free space is available */ 5461 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5462 buf->f_bavail = buf->f_bfree - 5463 (ext4_r_blocks_count(es) + resv_blocks); 5464 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5465 buf->f_bavail = 0; 5466 buf->f_files = le32_to_cpu(es->s_inodes_count); 5467 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5468 buf->f_namelen = EXT4_NAME_LEN; 5469 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5470 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5471 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5472 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5473 5474 #ifdef CONFIG_QUOTA 5475 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5476 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5477 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5478 #endif 5479 return 0; 5480 } 5481 5482 5483 #ifdef CONFIG_QUOTA 5484 5485 /* 5486 * Helper functions so that transaction is started before we acquire dqio_sem 5487 * to keep correct lock ordering of transaction > dqio_sem 5488 */ 5489 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5490 { 5491 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5492 } 5493 5494 static int ext4_write_dquot(struct dquot *dquot) 5495 { 5496 int ret, err; 5497 handle_t *handle; 5498 struct inode *inode; 5499 5500 inode = dquot_to_inode(dquot); 5501 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5502 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5503 if (IS_ERR(handle)) 5504 return PTR_ERR(handle); 5505 ret = dquot_commit(dquot); 5506 err = ext4_journal_stop(handle); 5507 if (!ret) 5508 ret = err; 5509 return ret; 5510 } 5511 5512 static int ext4_acquire_dquot(struct dquot *dquot) 5513 { 5514 int ret, err; 5515 handle_t *handle; 5516 5517 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5518 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5519 if (IS_ERR(handle)) 5520 return PTR_ERR(handle); 5521 ret = dquot_acquire(dquot); 5522 err = ext4_journal_stop(handle); 5523 if (!ret) 5524 ret = err; 5525 return ret; 5526 } 5527 5528 static int ext4_release_dquot(struct dquot *dquot) 5529 { 5530 int ret, err; 5531 handle_t *handle; 5532 5533 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5534 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5535 if (IS_ERR(handle)) { 5536 /* Release dquot anyway to avoid endless cycle in dqput() */ 5537 dquot_release(dquot); 5538 return PTR_ERR(handle); 5539 } 5540 ret = dquot_release(dquot); 5541 err = ext4_journal_stop(handle); 5542 if (!ret) 5543 ret = err; 5544 return ret; 5545 } 5546 5547 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5548 { 5549 struct super_block *sb = dquot->dq_sb; 5550 struct ext4_sb_info *sbi = EXT4_SB(sb); 5551 5552 /* Are we journaling quotas? */ 5553 if (ext4_has_feature_quota(sb) || 5554 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5555 dquot_mark_dquot_dirty(dquot); 5556 return ext4_write_dquot(dquot); 5557 } else { 5558 return dquot_mark_dquot_dirty(dquot); 5559 } 5560 } 5561 5562 static int ext4_write_info(struct super_block *sb, int type) 5563 { 5564 int ret, err; 5565 handle_t *handle; 5566 5567 /* Data block + inode block */ 5568 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5569 if (IS_ERR(handle)) 5570 return PTR_ERR(handle); 5571 ret = dquot_commit_info(sb, type); 5572 err = ext4_journal_stop(handle); 5573 if (!ret) 5574 ret = err; 5575 return ret; 5576 } 5577 5578 /* 5579 * Turn on quotas during mount time - we need to find 5580 * the quota file and such... 5581 */ 5582 static int ext4_quota_on_mount(struct super_block *sb, int type) 5583 { 5584 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 5585 EXT4_SB(sb)->s_jquota_fmt, type); 5586 } 5587 5588 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5589 { 5590 struct ext4_inode_info *ei = EXT4_I(inode); 5591 5592 /* The first argument of lockdep_set_subclass has to be 5593 * *exactly* the same as the argument to init_rwsem() --- in 5594 * this case, in init_once() --- or lockdep gets unhappy 5595 * because the name of the lock is set using the 5596 * stringification of the argument to init_rwsem(). 5597 */ 5598 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5599 lockdep_set_subclass(&ei->i_data_sem, subclass); 5600 } 5601 5602 /* 5603 * Standard function to be called on quota_on 5604 */ 5605 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5606 const struct path *path) 5607 { 5608 int err; 5609 5610 if (!test_opt(sb, QUOTA)) 5611 return -EINVAL; 5612 5613 /* Quotafile not on the same filesystem? */ 5614 if (path->dentry->d_sb != sb) 5615 return -EXDEV; 5616 /* Journaling quota? */ 5617 if (EXT4_SB(sb)->s_qf_names[type]) { 5618 /* Quotafile not in fs root? */ 5619 if (path->dentry->d_parent != sb->s_root) 5620 ext4_msg(sb, KERN_WARNING, 5621 "Quota file not on filesystem root. " 5622 "Journaled quota will not work"); 5623 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 5624 } else { 5625 /* 5626 * Clear the flag just in case mount options changed since 5627 * last time. 5628 */ 5629 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 5630 } 5631 5632 /* 5633 * When we journal data on quota file, we have to flush journal to see 5634 * all updates to the file when we bypass pagecache... 5635 */ 5636 if (EXT4_SB(sb)->s_journal && 5637 ext4_should_journal_data(d_inode(path->dentry))) { 5638 /* 5639 * We don't need to lock updates but journal_flush() could 5640 * otherwise be livelocked... 5641 */ 5642 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5643 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5644 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5645 if (err) 5646 return err; 5647 } 5648 5649 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5650 err = dquot_quota_on(sb, type, format_id, path); 5651 if (err) { 5652 lockdep_set_quota_inode(path->dentry->d_inode, 5653 I_DATA_SEM_NORMAL); 5654 } else { 5655 struct inode *inode = d_inode(path->dentry); 5656 handle_t *handle; 5657 5658 /* 5659 * Set inode flags to prevent userspace from messing with quota 5660 * files. If this fails, we return success anyway since quotas 5661 * are already enabled and this is not a hard failure. 5662 */ 5663 inode_lock(inode); 5664 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5665 if (IS_ERR(handle)) 5666 goto unlock_inode; 5667 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 5668 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 5669 S_NOATIME | S_IMMUTABLE); 5670 ext4_mark_inode_dirty(handle, inode); 5671 ext4_journal_stop(handle); 5672 unlock_inode: 5673 inode_unlock(inode); 5674 } 5675 return err; 5676 } 5677 5678 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5679 unsigned int flags) 5680 { 5681 int err; 5682 struct inode *qf_inode; 5683 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5684 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5685 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5686 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5687 }; 5688 5689 BUG_ON(!ext4_has_feature_quota(sb)); 5690 5691 if (!qf_inums[type]) 5692 return -EPERM; 5693 5694 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 5695 if (IS_ERR(qf_inode)) { 5696 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5697 return PTR_ERR(qf_inode); 5698 } 5699 5700 /* Don't account quota for quota files to avoid recursion */ 5701 qf_inode->i_flags |= S_NOQUOTA; 5702 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 5703 err = dquot_enable(qf_inode, type, format_id, flags); 5704 if (err) 5705 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 5706 iput(qf_inode); 5707 5708 return err; 5709 } 5710 5711 /* Enable usage tracking for all quota types. */ 5712 static int ext4_enable_quotas(struct super_block *sb) 5713 { 5714 int type, err = 0; 5715 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5716 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5717 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5718 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5719 }; 5720 bool quota_mopt[EXT4_MAXQUOTAS] = { 5721 test_opt(sb, USRQUOTA), 5722 test_opt(sb, GRPQUOTA), 5723 test_opt(sb, PRJQUOTA), 5724 }; 5725 5726 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 5727 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 5728 if (qf_inums[type]) { 5729 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5730 DQUOT_USAGE_ENABLED | 5731 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 5732 if (err) { 5733 ext4_warning(sb, 5734 "Failed to enable quota tracking " 5735 "(type=%d, err=%d). Please run " 5736 "e2fsck to fix.", type, err); 5737 for (type--; type >= 0; type--) 5738 dquot_quota_off(sb, type); 5739 5740 return err; 5741 } 5742 } 5743 } 5744 return 0; 5745 } 5746 5747 static int ext4_quota_off(struct super_block *sb, int type) 5748 { 5749 struct inode *inode = sb_dqopt(sb)->files[type]; 5750 handle_t *handle; 5751 int err; 5752 5753 /* Force all delayed allocation blocks to be allocated. 5754 * Caller already holds s_umount sem */ 5755 if (test_opt(sb, DELALLOC)) 5756 sync_filesystem(sb); 5757 5758 if (!inode || !igrab(inode)) 5759 goto out; 5760 5761 err = dquot_quota_off(sb, type); 5762 if (err || ext4_has_feature_quota(sb)) 5763 goto out_put; 5764 5765 inode_lock(inode); 5766 /* 5767 * Update modification times of quota files when userspace can 5768 * start looking at them. If we fail, we return success anyway since 5769 * this is not a hard failure and quotas are already disabled. 5770 */ 5771 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5772 if (IS_ERR(handle)) 5773 goto out_unlock; 5774 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 5775 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 5776 inode->i_mtime = inode->i_ctime = current_time(inode); 5777 ext4_mark_inode_dirty(handle, inode); 5778 ext4_journal_stop(handle); 5779 out_unlock: 5780 inode_unlock(inode); 5781 out_put: 5782 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 5783 iput(inode); 5784 return err; 5785 out: 5786 return dquot_quota_off(sb, type); 5787 } 5788 5789 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5790 * acquiring the locks... As quota files are never truncated and quota code 5791 * itself serializes the operations (and no one else should touch the files) 5792 * we don't have to be afraid of races */ 5793 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5794 size_t len, loff_t off) 5795 { 5796 struct inode *inode = sb_dqopt(sb)->files[type]; 5797 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5798 int offset = off & (sb->s_blocksize - 1); 5799 int tocopy; 5800 size_t toread; 5801 struct buffer_head *bh; 5802 loff_t i_size = i_size_read(inode); 5803 5804 if (off > i_size) 5805 return 0; 5806 if (off+len > i_size) 5807 len = i_size-off; 5808 toread = len; 5809 while (toread > 0) { 5810 tocopy = sb->s_blocksize - offset < toread ? 5811 sb->s_blocksize - offset : toread; 5812 bh = ext4_bread(NULL, inode, blk, 0); 5813 if (IS_ERR(bh)) 5814 return PTR_ERR(bh); 5815 if (!bh) /* A hole? */ 5816 memset(data, 0, tocopy); 5817 else 5818 memcpy(data, bh->b_data+offset, tocopy); 5819 brelse(bh); 5820 offset = 0; 5821 toread -= tocopy; 5822 data += tocopy; 5823 blk++; 5824 } 5825 return len; 5826 } 5827 5828 /* Write to quotafile (we know the transaction is already started and has 5829 * enough credits) */ 5830 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5831 const char *data, size_t len, loff_t off) 5832 { 5833 struct inode *inode = sb_dqopt(sb)->files[type]; 5834 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5835 int err, offset = off & (sb->s_blocksize - 1); 5836 int retries = 0; 5837 struct buffer_head *bh; 5838 handle_t *handle = journal_current_handle(); 5839 5840 if (EXT4_SB(sb)->s_journal && !handle) { 5841 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5842 " cancelled because transaction is not started", 5843 (unsigned long long)off, (unsigned long long)len); 5844 return -EIO; 5845 } 5846 /* 5847 * Since we account only one data block in transaction credits, 5848 * then it is impossible to cross a block boundary. 5849 */ 5850 if (sb->s_blocksize - offset < len) { 5851 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5852 " cancelled because not block aligned", 5853 (unsigned long long)off, (unsigned long long)len); 5854 return -EIO; 5855 } 5856 5857 do { 5858 bh = ext4_bread(handle, inode, blk, 5859 EXT4_GET_BLOCKS_CREATE | 5860 EXT4_GET_BLOCKS_METADATA_NOFAIL); 5861 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 5862 ext4_should_retry_alloc(inode->i_sb, &retries)); 5863 if (IS_ERR(bh)) 5864 return PTR_ERR(bh); 5865 if (!bh) 5866 goto out; 5867 BUFFER_TRACE(bh, "get write access"); 5868 err = ext4_journal_get_write_access(handle, bh); 5869 if (err) { 5870 brelse(bh); 5871 return err; 5872 } 5873 lock_buffer(bh); 5874 memcpy(bh->b_data+offset, data, len); 5875 flush_dcache_page(bh->b_page); 5876 unlock_buffer(bh); 5877 err = ext4_handle_dirty_metadata(handle, NULL, bh); 5878 brelse(bh); 5879 out: 5880 if (inode->i_size < off + len) { 5881 i_size_write(inode, off + len); 5882 EXT4_I(inode)->i_disksize = inode->i_size; 5883 ext4_mark_inode_dirty(handle, inode); 5884 } 5885 return len; 5886 } 5887 5888 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) 5889 { 5890 const struct quota_format_ops *ops; 5891 5892 if (!sb_has_quota_loaded(sb, qid->type)) 5893 return -ESRCH; 5894 ops = sb_dqopt(sb)->ops[qid->type]; 5895 if (!ops || !ops->get_next_id) 5896 return -ENOSYS; 5897 return dquot_get_next_id(sb, qid); 5898 } 5899 #endif 5900 5901 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5902 const char *dev_name, void *data) 5903 { 5904 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5905 } 5906 5907 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 5908 static inline void register_as_ext2(void) 5909 { 5910 int err = register_filesystem(&ext2_fs_type); 5911 if (err) 5912 printk(KERN_WARNING 5913 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 5914 } 5915 5916 static inline void unregister_as_ext2(void) 5917 { 5918 unregister_filesystem(&ext2_fs_type); 5919 } 5920 5921 static inline int ext2_feature_set_ok(struct super_block *sb) 5922 { 5923 if (ext4_has_unknown_ext2_incompat_features(sb)) 5924 return 0; 5925 if (sb_rdonly(sb)) 5926 return 1; 5927 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 5928 return 0; 5929 return 1; 5930 } 5931 #else 5932 static inline void register_as_ext2(void) { } 5933 static inline void unregister_as_ext2(void) { } 5934 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 5935 #endif 5936 5937 static inline void register_as_ext3(void) 5938 { 5939 int err = register_filesystem(&ext3_fs_type); 5940 if (err) 5941 printk(KERN_WARNING 5942 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 5943 } 5944 5945 static inline void unregister_as_ext3(void) 5946 { 5947 unregister_filesystem(&ext3_fs_type); 5948 } 5949 5950 static inline int ext3_feature_set_ok(struct super_block *sb) 5951 { 5952 if (ext4_has_unknown_ext3_incompat_features(sb)) 5953 return 0; 5954 if (!ext4_has_feature_journal(sb)) 5955 return 0; 5956 if (sb_rdonly(sb)) 5957 return 1; 5958 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 5959 return 0; 5960 return 1; 5961 } 5962 5963 static struct file_system_type ext4_fs_type = { 5964 .owner = THIS_MODULE, 5965 .name = "ext4", 5966 .mount = ext4_mount, 5967 .kill_sb = kill_block_super, 5968 .fs_flags = FS_REQUIRES_DEV, 5969 }; 5970 MODULE_ALIAS_FS("ext4"); 5971 5972 /* Shared across all ext4 file systems */ 5973 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5974 5975 static int __init ext4_init_fs(void) 5976 { 5977 int i, err; 5978 5979 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 5980 ext4_li_info = NULL; 5981 mutex_init(&ext4_li_mtx); 5982 5983 /* Build-time check for flags consistency */ 5984 ext4_check_flag_values(); 5985 5986 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 5987 init_waitqueue_head(&ext4__ioend_wq[i]); 5988 5989 err = ext4_init_es(); 5990 if (err) 5991 return err; 5992 5993 err = ext4_init_pending(); 5994 if (err) 5995 goto out6; 5996 5997 err = ext4_init_pageio(); 5998 if (err) 5999 goto out5; 6000 6001 err = ext4_init_system_zone(); 6002 if (err) 6003 goto out4; 6004 6005 err = ext4_init_sysfs(); 6006 if (err) 6007 goto out3; 6008 6009 err = ext4_init_mballoc(); 6010 if (err) 6011 goto out2; 6012 err = init_inodecache(); 6013 if (err) 6014 goto out1; 6015 register_as_ext3(); 6016 register_as_ext2(); 6017 err = register_filesystem(&ext4_fs_type); 6018 if (err) 6019 goto out; 6020 6021 return 0; 6022 out: 6023 unregister_as_ext2(); 6024 unregister_as_ext3(); 6025 destroy_inodecache(); 6026 out1: 6027 ext4_exit_mballoc(); 6028 out2: 6029 ext4_exit_sysfs(); 6030 out3: 6031 ext4_exit_system_zone(); 6032 out4: 6033 ext4_exit_pageio(); 6034 out5: 6035 ext4_exit_pending(); 6036 out6: 6037 ext4_exit_es(); 6038 6039 return err; 6040 } 6041 6042 static void __exit ext4_exit_fs(void) 6043 { 6044 ext4_destroy_lazyinit_thread(); 6045 unregister_as_ext2(); 6046 unregister_as_ext3(); 6047 unregister_filesystem(&ext4_fs_type); 6048 destroy_inodecache(); 6049 ext4_exit_mballoc(); 6050 ext4_exit_sysfs(); 6051 ext4_exit_system_zone(); 6052 ext4_exit_pageio(); 6053 ext4_exit_es(); 6054 ext4_exit_pending(); 6055 } 6056 6057 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 6058 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6059 MODULE_LICENSE("GPL"); 6060 MODULE_SOFTDEP("pre: crc32c"); 6061 module_init(ext4_init_fs) 6062 module_exit(ext4_exit_fs) 6063