fs/ext4/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  *  linux/fs/ext4/super.c
   4  *
   5  * Copyright (C) 1992, 1993, 1994, 1995
   6  * Remy Card (card@masi.ibp.fr)
   7  * Laboratoire MASI - Institut Blaise Pascal
   8  * Universite Pierre et Marie Curie (Paris VI)
   9  *
  10  *  from
  11  *
  12  *  linux/fs/minix/inode.c
  13  *
  14  *  Copyright (C) 1991, 1992  Linus Torvalds
  15  *
  16  *  Big-endian to little-endian byte-swapping/bitmaps by
  17  *        David S. Miller (davem@caip.rutgers.edu), 1995
  18  */
  19
  20 #include <linux/module.h>
  21 #include <linux/string.h>
  22 #include <linux/fs.h>
  23 #include <linux/time.h>
  24 #include <linux/vmalloc.h>
  25 #include <linux/slab.h>
  26 #include <linux/init.h>
  27 #include <linux/blkdev.h>
  28 #include <linux/backing-dev.h>
  29 #include <linux/parser.h>
  30 #include <linux/buffer_head.h>
  31 #include <linux/exportfs.h>
  32 #include <linux/vfs.h>
  33 #include <linux/random.h>
  34 #include <linux/mount.h>
  35 #include <linux/namei.h>
  36 #include <linux/quotaops.h>
  37 #include <linux/seq_file.h>
  38 #include <linux/ctype.h>
  39 #include <linux/log2.h>
  40 #include <linux/crc16.h>
  41 #include <linux/dax.h>
  42 #include <linux/cleancache.h>
  43 #include <linux/uaccess.h>
  44 #include <linux/iversion.h>
  45 #include <linux/unicode.h>
  46 #include <linux/part_stat.h>
  47 #include <linux/kthread.h>
  48 #include <linux/freezer.h>
  49
  50 #include "ext4.h"
  51 #include "ext4_extents.h"       /* Needed for trace points definition */
  52 #include "ext4_jbd2.h"
  53 #include "xattr.h"
  54 #include "acl.h"
  55 #include "mballoc.h"
  56 #include "fsmap.h"
  57
  58 #define CREATE_TRACE_POINTS
  59 #include <trace/events/ext4.h>
  60
  61 static struct ext4_lazy_init *ext4_li_info;
  62 static DEFINE_MUTEX(ext4_li_mtx);
  63 static struct ratelimit_state ext4_mount_msg_ratelimit;
  64
  65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  66                              unsigned long journal_devnum);
  67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
  68 static void ext4_update_super(struct super_block *sb);
  69 static int ext4_commit_super(struct super_block *sb);
  70 static int ext4_mark_recovery_complete(struct super_block *sb,
  71                                         struct ext4_super_block *es);
  72 static int ext4_clear_journal_err(struct super_block *sb,
  73                                   struct ext4_super_block *es);
  74 static int ext4_sync_fs(struct super_block *sb, int wait);
  75 static int ext4_remount(struct super_block *sb, int *flags, char *data);
  76 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
  77 static int ext4_unfreeze(struct super_block *sb);
  78 static int ext4_freeze(struct super_block *sb);
  79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
  80                        const char *dev_name, void *data);
  81 static inline int ext2_feature_set_ok(struct super_block *sb);
  82 static inline int ext3_feature_set_ok(struct super_block *sb);
  83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
  84 static void ext4_destroy_lazyinit_thread(void);
  85 static void ext4_unregister_li_request(struct super_block *sb);
  86 static void ext4_clear_request_list(void);
  87 static struct inode *ext4_get_journal_inode(struct super_block *sb,
  88                                             unsigned int journal_inum);
  89
  90 /*
  91  * Lock ordering
  92  *
  93  * page fault path:
  94  * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
  95  *   -> page lock -> i_data_sem (rw)
  96  *
  97  * buffered write path:
  98  * sb_start_write -> i_mutex -> mmap_lock
  99  * sb_start_write -> i_mutex -> transaction start -> page lock ->
 100  *   i_data_sem (rw)
 101  *
 102  * truncate:
 103  * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
 104  *   page lock
 105  * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
 106  *   i_data_sem (rw)
 107  *
 108  * direct IO:
 109  * sb_start_write -> i_mutex -> mmap_lock
 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 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
 145                                   bh_end_io_t *end_io)
 146 {
 147         /*
 148          * buffer's verified bit is no longer valid after reading from
 149          * disk again due to write out error, clear it to make sure we
 150          * recheck the buffer contents.
 151          */
 152         clear_buffer_verified(bh);
 153
 154         bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
 155         get_bh(bh);
 156         submit_bh(REQ_OP_READ, op_flags, bh);
 157 }
 158
 159 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
 160                          bh_end_io_t *end_io)
 161 {
 162         BUG_ON(!buffer_locked(bh));
 163
 164         if (ext4_buffer_uptodate(bh)) {
 165                 unlock_buffer(bh);
 166                 return;
 167         }
 168         __ext4_read_bh(bh, op_flags, end_io);
 169 }
 170
 171 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
 172 {
 173         BUG_ON(!buffer_locked(bh));
 174
 175         if (ext4_buffer_uptodate(bh)) {
 176                 unlock_buffer(bh);
 177                 return 0;
 178         }
 179
 180         __ext4_read_bh(bh, op_flags, end_io);
 181
 182         wait_on_buffer(bh);
 183         if (buffer_uptodate(bh))
 184                 return 0;
 185         return -EIO;
 186 }
 187
 188 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
 189 {
 190         if (trylock_buffer(bh)) {
 191                 if (wait)
 192                         return ext4_read_bh(bh, op_flags, NULL);
 193                 ext4_read_bh_nowait(bh, op_flags, NULL);
 194                 return 0;
 195         }
 196         if (wait) {
 197                 wait_on_buffer(bh);
 198                 if (buffer_uptodate(bh))
 199                         return 0;
 200                 return -EIO;
 201         }
 202         return 0;
 203 }
 204
 205 /*
 206  * This works like __bread_gfp() except it uses ERR_PTR for error
 207  * returns.  Currently with sb_bread it's impossible to distinguish
 208  * between ENOMEM and EIO situations (since both result in a NULL
 209  * return.
 210  */
 211 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
 212                                                sector_t block, int op_flags,
 213                                                gfp_t gfp)
 214 {
 215         struct buffer_head *bh;
 216         int ret;
 217
 218         bh = sb_getblk_gfp(sb, block, gfp);
 219         if (bh == NULL)
 220                 return ERR_PTR(-ENOMEM);
 221         if (ext4_buffer_uptodate(bh))
 222                 return bh;
 223
 224         ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
 225         if (ret) {
 226                 put_bh(bh);
 227                 return ERR_PTR(ret);
 228         }
 229         return bh;
 230 }
 231
 232 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
 233                                    int op_flags)
 234 {
 235         return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
 236 }
 237
 238 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
 239                                             sector_t block)
 240 {
 241         return __ext4_sb_bread_gfp(sb, block, 0, 0);
 242 }
 243
 244 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
 245 {
 246         struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
 247
 248         if (likely(bh)) {
 249                 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
 250                 brelse(bh);
 251         }
 252 }
 253
 254 static int ext4_verify_csum_type(struct super_block *sb,
 255                                  struct ext4_super_block *es)
 256 {
 257         if (!ext4_has_feature_metadata_csum(sb))
 258                 return 1;
 259
 260         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
 261 }
 262
 263 static __le32 ext4_superblock_csum(struct super_block *sb,
 264                                    struct ext4_super_block *es)
 265 {
 266         struct ext4_sb_info *sbi = EXT4_SB(sb);
 267         int offset = offsetof(struct ext4_super_block, s_checksum);
 268         __u32 csum;
 269
 270         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
 271
 272         return cpu_to_le32(csum);
 273 }
 274
 275 static int ext4_superblock_csum_verify(struct super_block *sb,
 276                                        struct ext4_super_block *es)
 277 {
 278         if (!ext4_has_metadata_csum(sb))
 279                 return 1;
 280
 281         return es->s_checksum == ext4_superblock_csum(sb, es);
 282 }
 283
 284 void ext4_superblock_csum_set(struct super_block *sb)
 285 {
 286         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 287
 288         if (!ext4_has_metadata_csum(sb))
 289                 return;
 290
 291         es->s_checksum = ext4_superblock_csum(sb, es);
 292 }
 293
 294 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
 295                                struct ext4_group_desc *bg)
 296 {
 297         return le32_to_cpu(bg->bg_block_bitmap_lo) |
 298                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 299                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
 300 }
 301
 302 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
 303                                struct ext4_group_desc *bg)
 304 {
 305         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
 306                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 307                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
 308 }
 309
 310 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
 311                               struct ext4_group_desc *bg)
 312 {
 313         return le32_to_cpu(bg->bg_inode_table_lo) |
 314                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 315                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
 316 }
 317
 318 __u32 ext4_free_group_clusters(struct super_block *sb,
 319                                struct ext4_group_desc *bg)
 320 {
 321         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
 322                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 323                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
 324 }
 325
 326 __u32 ext4_free_inodes_count(struct super_block *sb,
 327                               struct ext4_group_desc *bg)
 328 {
 329         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
 330                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 331                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
 332 }
 333
 334 __u32 ext4_used_dirs_count(struct super_block *sb,
 335                               struct ext4_group_desc *bg)
 336 {
 337         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
 338                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 339                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
 340 }
 341
 342 __u32 ext4_itable_unused_count(struct super_block *sb,
 343                               struct ext4_group_desc *bg)
 344 {
 345         return le16_to_cpu(bg->bg_itable_unused_lo) |
 346                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 347                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
 348 }
 349
 350 void ext4_block_bitmap_set(struct super_block *sb,
 351                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 352 {
 353         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
 354         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 355                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
 356 }
 357
 358 void ext4_inode_bitmap_set(struct super_block *sb,
 359                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 360 {
 361         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
 362         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 363                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
 364 }
 365
 366 void ext4_inode_table_set(struct super_block *sb,
 367                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
 368 {
 369         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
 370         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 371                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
 372 }
 373
 374 void ext4_free_group_clusters_set(struct super_block *sb,
 375                                   struct ext4_group_desc *bg, __u32 count)
 376 {
 377         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
 378         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 379                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
 380 }
 381
 382 void ext4_free_inodes_set(struct super_block *sb,
 383                           struct ext4_group_desc *bg, __u32 count)
 384 {
 385         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
 386         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 387                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
 388 }
 389
 390 void ext4_used_dirs_set(struct super_block *sb,
 391                           struct ext4_group_desc *bg, __u32 count)
 392 {
 393         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
 394         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 395                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
 396 }
 397
 398 void ext4_itable_unused_set(struct super_block *sb,
 399                           struct ext4_group_desc *bg, __u32 count)
 400 {
 401         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
 402         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 403                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
 404 }
 405
 406 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
 407 {
 408         now = clamp_val(now, 0, (1ull << 40) - 1);
 409
 410         *lo = cpu_to_le32(lower_32_bits(now));
 411         *hi = upper_32_bits(now);
 412 }
 413
 414 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
 415 {
 416         return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
 417 }
 418 #define ext4_update_tstamp(es, tstamp) \
 419         __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
 420                              ktime_get_real_seconds())
 421 #define ext4_get_tstamp(es, tstamp) \
 422         __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
 423
 424 /*
 425  * The del_gendisk() function uninitializes the disk-specific data
 426  * structures, including the bdi structure, without telling anyone
 427  * else.  Once this happens, any attempt to call mark_buffer_dirty()
 428  * (for example, by ext4_commit_super), will cause a kernel OOPS.
 429  * This is a kludge to prevent these oops until we can put in a proper
 430  * hook in del_gendisk() to inform the VFS and file system layers.
 431  */
 432 static int block_device_ejected(struct super_block *sb)
 433 {
 434         struct inode *bd_inode = sb->s_bdev->bd_inode;
 435         struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
 436
 437         return bdi->dev == NULL;
 438 }
 439
 440 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
 441 {
 442         struct super_block              *sb = journal->j_private;
 443         struct ext4_sb_info             *sbi = EXT4_SB(sb);
 444         int                             error = is_journal_aborted(journal);
 445         struct ext4_journal_cb_entry    *jce;
 446
 447         BUG_ON(txn->t_state == T_FINISHED);
 448
 449         ext4_process_freed_data(sb, txn->t_tid);
 450
 451         spin_lock(&sbi->s_md_lock);
 452         while (!list_empty(&txn->t_private_list)) {
 453                 jce = list_entry(txn->t_private_list.next,
 454                                  struct ext4_journal_cb_entry, jce_list);
 455                 list_del_init(&jce->jce_list);
 456                 spin_unlock(&sbi->s_md_lock);
 457                 jce->jce_func(sb, jce, error);
 458                 spin_lock(&sbi->s_md_lock);
 459         }
 460         spin_unlock(&sbi->s_md_lock);
 461 }
 462
 463 /*
 464  * This writepage callback for write_cache_pages()
 465  * takes care of a few cases after page cleaning.
 466  *
 467  * write_cache_pages() already checks for dirty pages
 468  * and calls clear_page_dirty_for_io(), which we want,
 469  * to write protect the pages.
 470  *
 471  * However, we may have to redirty a page (see below.)
 472  */
 473 static int ext4_journalled_writepage_callback(struct page *page,
 474                                               struct writeback_control *wbc,
 475                                               void *data)
 476 {
 477         transaction_t *transaction = (transaction_t *) data;
 478         struct buffer_head *bh, *head;
 479         struct journal_head *jh;
 480
 481         bh = head = page_buffers(page);
 482         do {
 483                 /*
 484                  * We have to redirty a page in these cases:
 485                  * 1) If buffer is dirty, it means the page was dirty because it
 486                  * contains a buffer that needs checkpointing. So the dirty bit
 487                  * needs to be preserved so that checkpointing writes the buffer
 488                  * properly.
 489                  * 2) If buffer is not part of the committing transaction
 490                  * (we may have just accidentally come across this buffer because
 491                  * inode range tracking is not exact) or if the currently running
 492                  * transaction already contains this buffer as well, dirty bit
 493                  * needs to be preserved so that the buffer gets writeprotected
 494                  * properly on running transaction's commit.
 495                  */
 496                 jh = bh2jh(bh);
 497                 if (buffer_dirty(bh) ||
 498                     (jh && (jh->b_transaction != transaction ||
 499                             jh->b_next_transaction))) {
 500                         redirty_page_for_writepage(wbc, page);
 501                         goto out;
 502                 }
 503         } while ((bh = bh->b_this_page) != head);
 504
 505 out:
 506         return AOP_WRITEPAGE_ACTIVATE;
 507 }
 508
 509 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
 510 {
 511         struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
 512         struct writeback_control wbc = {
 513                 .sync_mode =  WB_SYNC_ALL,
 514                 .nr_to_write = LONG_MAX,
 515                 .range_start = jinode->i_dirty_start,
 516                 .range_end = jinode->i_dirty_end,
 517         };
 518
 519         return write_cache_pages(mapping, &wbc,
 520                                  ext4_journalled_writepage_callback,
 521                                  jinode->i_transaction);
 522 }
 523
 524 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
 525 {
 526         int ret;
 527
 528         if (ext4_should_journal_data(jinode->i_vfs_inode))
 529                 ret = ext4_journalled_submit_inode_data_buffers(jinode);
 530         else
 531                 ret = jbd2_journal_submit_inode_data_buffers(jinode);
 532
 533         return ret;
 534 }
 535
 536 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
 537 {
 538         int ret = 0;
 539
 540         if (!ext4_should_journal_data(jinode->i_vfs_inode))
 541                 ret = jbd2_journal_finish_inode_data_buffers(jinode);
 542
 543         return ret;
 544 }
 545
 546 static bool system_going_down(void)
 547 {
 548         return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
 549                 || system_state == SYSTEM_RESTART;
 550 }
 551
 552 struct ext4_err_translation {
 553         int code;
 554         int errno;
 555 };
 556
 557 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
 558
 559 static struct ext4_err_translation err_translation[] = {
 560         EXT4_ERR_TRANSLATE(EIO),
 561         EXT4_ERR_TRANSLATE(ENOMEM),
 562         EXT4_ERR_TRANSLATE(EFSBADCRC),
 563         EXT4_ERR_TRANSLATE(EFSCORRUPTED),
 564         EXT4_ERR_TRANSLATE(ENOSPC),
 565         EXT4_ERR_TRANSLATE(ENOKEY),
 566         EXT4_ERR_TRANSLATE(EROFS),
 567         EXT4_ERR_TRANSLATE(EFBIG),
 568         EXT4_ERR_TRANSLATE(EEXIST),
 569         EXT4_ERR_TRANSLATE(ERANGE),
 570         EXT4_ERR_TRANSLATE(EOVERFLOW),
 571         EXT4_ERR_TRANSLATE(EBUSY),
 572         EXT4_ERR_TRANSLATE(ENOTDIR),
 573         EXT4_ERR_TRANSLATE(ENOTEMPTY),
 574         EXT4_ERR_TRANSLATE(ESHUTDOWN),
 575         EXT4_ERR_TRANSLATE(EFAULT),
 576 };
 577
 578 static int ext4_errno_to_code(int errno)
 579 {
 580         int i;
 581
 582         for (i = 0; i < ARRAY_SIZE(err_translation); i++)
 583                 if (err_translation[i].errno == errno)
 584                         return err_translation[i].code;
 585         return EXT4_ERR_UNKNOWN;
 586 }
 587
 588 static void save_error_info(struct super_block *sb, int error,
 589                             __u32 ino, __u64 block,
 590                             const char *func, unsigned int line)
 591 {
 592         struct ext4_sb_info *sbi = EXT4_SB(sb);
 593
 594         /* We default to EFSCORRUPTED error... */
 595         if (error == 0)
 596                 error = EFSCORRUPTED;
 597
 598         spin_lock(&sbi->s_error_lock);
 599         sbi->s_add_error_count++;
 600         sbi->s_last_error_code = error;
 601         sbi->s_last_error_line = line;
 602         sbi->s_last_error_ino = ino;
 603         sbi->s_last_error_block = block;
 604         sbi->s_last_error_func = func;
 605         sbi->s_last_error_time = ktime_get_real_seconds();
 606         if (!sbi->s_first_error_time) {
 607                 sbi->s_first_error_code = error;
 608                 sbi->s_first_error_line = line;
 609                 sbi->s_first_error_ino = ino;
 610                 sbi->s_first_error_block = block;
 611                 sbi->s_first_error_func = func;
 612                 sbi->s_first_error_time = sbi->s_last_error_time;
 613         }
 614         spin_unlock(&sbi->s_error_lock);
 615 }
 616
 617 /* Deal with the reporting of failure conditions on a filesystem such as
 618  * inconsistencies detected or read IO failures.
 619  *
 620  * On ext2, we can store the error state of the filesystem in the
 621  * superblock.  That is not possible on ext4, because we may have other
 622  * write ordering constraints on the superblock which prevent us from
 623  * writing it out straight away; and given that the journal is about to
 624  * be aborted, we can't rely on the current, or future, transactions to
 625  * write out the superblock safely.
 626  *
 627  * We'll just use the jbd2_journal_abort() error code to record an error in
 628  * the journal instead.  On recovery, the journal will complain about
 629  * that error until we've noted it down and cleared it.
 630  *
 631  * If force_ro is set, we unconditionally force the filesystem into an
 632  * ABORT|READONLY state, unless the error response on the fs has been set to
 633  * panic in which case we take the easy way out and panic immediately. This is
 634  * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
 635  * at a critical moment in log management.
 636  */
 637 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
 638                               __u32 ino, __u64 block,
 639                               const char *func, unsigned int line)
 640 {
 641         journal_t *journal = EXT4_SB(sb)->s_journal;
 642         bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
 643
 644         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 645         if (test_opt(sb, WARN_ON_ERROR))
 646                 WARN_ON_ONCE(1);
 647
 648         if (!continue_fs && !sb_rdonly(sb)) {
 649                 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
 650                 if (journal)
 651                         jbd2_journal_abort(journal, -EIO);
 652         }
 653
 654         if (!bdev_read_only(sb->s_bdev)) {
 655                 save_error_info(sb, error, ino, block, func, line);
 656                 /*
 657                  * In case the fs should keep running, we need to writeout
 658                  * superblock through the journal. Due to lock ordering
 659                  * constraints, it may not be safe to do it right here so we
 660                  * defer superblock flushing to a workqueue.
 661                  */
 662                 if (continue_fs)
 663                         schedule_work(&EXT4_SB(sb)->s_error_work);
 664                 else
 665                         ext4_commit_super(sb);
 666         }
 667
 668         /*
 669          * We force ERRORS_RO behavior when system is rebooting. Otherwise we
 670          * could panic during 'reboot -f' as the underlying device got already
 671          * disabled.
 672          */
 673         if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
 674                 panic("EXT4-fs (device %s): panic forced after error\n",
 675                         sb->s_id);
 676         }
 677
 678         if (sb_rdonly(sb) || continue_fs)
 679                 return;
 680
 681         ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 682         /*
 683          * Make sure updated value of ->s_mount_flags will be visible before
 684          * ->s_flags update
 685          */
 686         smp_wmb();
 687         sb->s_flags |= SB_RDONLY;
 688 }
 689
 690 static void flush_stashed_error_work(struct work_struct *work)
 691 {
 692         struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
 693                                                 s_error_work);
 694         journal_t *journal = sbi->s_journal;
 695         handle_t *handle;
 696
 697         /*
 698          * If the journal is still running, we have to write out superblock
 699          * through the journal to avoid collisions of other journalled sb
 700          * updates.
 701          *
 702          * We use directly jbd2 functions here to avoid recursing back into
 703          * ext4 error handling code during handling of previous errors.
 704          */
 705         if (!sb_rdonly(sbi->s_sb) && journal) {
 706                 struct buffer_head *sbh = sbi->s_sbh;
 707                 handle = jbd2_journal_start(journal, 1);
 708                 if (IS_ERR(handle))
 709                         goto write_directly;
 710                 if (jbd2_journal_get_write_access(handle, sbh)) {
 711                         jbd2_journal_stop(handle);
 712                         goto write_directly;
 713                 }
 714                 ext4_update_super(sbi->s_sb);
 715                 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
 716                         ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
 717                                  "superblock detected");
 718                         clear_buffer_write_io_error(sbh);
 719                         set_buffer_uptodate(sbh);
 720                 }
 721
 722                 if (jbd2_journal_dirty_metadata(handle, sbh)) {
 723                         jbd2_journal_stop(handle);
 724                         goto write_directly;
 725                 }
 726                 jbd2_journal_stop(handle);
 727                 ext4_notify_error_sysfs(sbi);
 728                 return;
 729         }
 730 write_directly:
 731         /*
 732          * Write through journal failed. Write sb directly to get error info
 733          * out and hope for the best.
 734          */
 735         ext4_commit_super(sbi->s_sb);
 736         ext4_notify_error_sysfs(sbi);
 737 }
 738
 739 #define ext4_error_ratelimit(sb)                                        \
 740                 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),     \
 741                              "EXT4-fs error")
 742
 743 void __ext4_error(struct super_block *sb, const char *function,
 744                   unsigned int line, bool force_ro, int error, __u64 block,
 745                   const char *fmt, ...)
 746 {
 747         struct va_format vaf;
 748         va_list args;
 749
 750         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
 751                 return;
 752
 753         trace_ext4_error(sb, function, line);
 754         if (ext4_error_ratelimit(sb)) {
 755                 va_start(args, fmt);
 756                 vaf.fmt = fmt;
 757                 vaf.va = &args;
 758                 printk(KERN_CRIT
 759                        "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
 760                        sb->s_id, function, line, current->comm, &vaf);
 761                 va_end(args);
 762         }
 763         ext4_handle_error(sb, force_ro, error, 0, block, function, line);
 764 }
 765
 766 void __ext4_error_inode(struct inode *inode, const char *function,
 767                         unsigned int line, ext4_fsblk_t block, int error,
 768                         const char *fmt, ...)
 769 {
 770         va_list args;
 771         struct va_format vaf;
 772
 773         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
 774                 return;
 775
 776         trace_ext4_error(inode->i_sb, function, line);
 777         if (ext4_error_ratelimit(inode->i_sb)) {
 778                 va_start(args, fmt);
 779                 vaf.fmt = fmt;
 780                 vaf.va = &args;
 781                 if (block)
 782                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 783                                "inode #%lu: block %llu: comm %s: %pV\n",
 784                                inode->i_sb->s_id, function, line, inode->i_ino,
 785                                block, current->comm, &vaf);
 786                 else
 787                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 788                                "inode #%lu: comm %s: %pV\n",
 789                                inode->i_sb->s_id, function, line, inode->i_ino,
 790                                current->comm, &vaf);
 791                 va_end(args);
 792         }
 793         ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
 794                           function, line);
 795 }
 796
 797 void __ext4_error_file(struct file *file, const char *function,
 798                        unsigned int line, ext4_fsblk_t block,
 799                        const char *fmt, ...)
 800 {
 801         va_list args;
 802         struct va_format vaf;
 803         struct inode *inode = file_inode(file);
 804         char pathname[80], *path;
 805
 806         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
 807                 return;
 808
 809         trace_ext4_error(inode->i_sb, function, line);
 810         if (ext4_error_ratelimit(inode->i_sb)) {
 811                 path = file_path(file, pathname, sizeof(pathname));
 812                 if (IS_ERR(path))
 813                         path = "(unknown)";
 814                 va_start(args, fmt);
 815                 vaf.fmt = fmt;
 816                 vaf.va = &args;
 817                 if (block)
 818                         printk(KERN_CRIT
 819                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 820                                "block %llu: comm %s: path %s: %pV\n",
 821                                inode->i_sb->s_id, function, line, inode->i_ino,
 822                                block, current->comm, path, &vaf);
 823                 else
 824                         printk(KERN_CRIT
 825                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 826                                "comm %s: path %s: %pV\n",
 827                                inode->i_sb->s_id, function, line, inode->i_ino,
 828                                current->comm, path, &vaf);
 829                 va_end(args);
 830         }
 831         ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
 832                           function, line);
 833 }
 834
 835 const char *ext4_decode_error(struct super_block *sb, int errno,
 836                               char nbuf[16])
 837 {
 838         char *errstr = NULL;
 839
 840         switch (errno) {
 841         case -EFSCORRUPTED:
 842                 errstr = "Corrupt filesystem";
 843                 break;
 844         case -EFSBADCRC:
 845                 errstr = "Filesystem failed CRC";
 846                 break;
 847         case -EIO:
 848                 errstr = "IO failure";
 849                 break;
 850         case -ENOMEM:
 851                 errstr = "Out of memory";
 852                 break;
 853         case -EROFS:
 854                 if (!sb || (EXT4_SB(sb)->s_journal &&
 855                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
 856                         errstr = "Journal has aborted";
 857                 else
 858                         errstr = "Readonly filesystem";
 859                 break;
 860         default:
 861                 /* If the caller passed in an extra buffer for unknown
 862                  * errors, textualise them now.  Else we just return
 863                  * NULL. */
 864                 if (nbuf) {
 865                         /* Check for truncated error codes... */
 866                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
 867                                 errstr = nbuf;
 868                 }
 869                 break;
 870         }
 871
 872         return errstr;
 873 }
 874
 875 /* __ext4_std_error decodes expected errors from journaling functions
 876  * automatically and invokes the appropriate error response.  */
 877
 878 void __ext4_std_error(struct super_block *sb, const char *function,
 879                       unsigned int line, int errno)
 880 {
 881         char nbuf[16];
 882         const char *errstr;
 883
 884         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
 885                 return;
 886
 887         /* Special case: if the error is EROFS, and we're not already
 888          * inside a transaction, then there's really no point in logging
 889          * an error. */
 890         if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
 891                 return;
 892
 893         if (ext4_error_ratelimit(sb)) {
 894                 errstr = ext4_decode_error(sb, errno, nbuf);
 895                 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
 896                        sb->s_id, function, line, errstr);
 897         }
 898
 899         ext4_handle_error(sb, false, -errno, 0, 0, function, line);
 900 }
 901
 902 void __ext4_msg(struct super_block *sb,
 903                 const char *prefix, const char *fmt, ...)
 904 {
 905         struct va_format vaf;
 906         va_list args;
 907
 908         atomic_inc(&EXT4_SB(sb)->s_msg_count);
 909         if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
 910                 return;
 911
 912         va_start(args, fmt);
 913         vaf.fmt = fmt;
 914         vaf.va = &args;
 915         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
 916         va_end(args);
 917 }
 918
 919 static int ext4_warning_ratelimit(struct super_block *sb)
 920 {
 921         atomic_inc(&EXT4_SB(sb)->s_warning_count);
 922         return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
 923                             "EXT4-fs warning");
 924 }
 925
 926 void __ext4_warning(struct super_block *sb, const char *function,
 927                     unsigned int line, const char *fmt, ...)
 928 {
 929         struct va_format vaf;
 930         va_list args;
 931
 932         if (!ext4_warning_ratelimit(sb))
 933                 return;
 934
 935         va_start(args, fmt);
 936         vaf.fmt = fmt;
 937         vaf.va = &args;
 938         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
 939                sb->s_id, function, line, &vaf);
 940         va_end(args);
 941 }
 942
 943 void __ext4_warning_inode(const struct inode *inode, const char *function,
 944                           unsigned int line, const char *fmt, ...)
 945 {
 946         struct va_format vaf;
 947         va_list args;
 948
 949         if (!ext4_warning_ratelimit(inode->i_sb))
 950                 return;
 951
 952         va_start(args, fmt);
 953         vaf.fmt = fmt;
 954         vaf.va = &args;
 955         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
 956                "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
 957                function, line, inode->i_ino, current->comm, &vaf);
 958         va_end(args);
 959 }
 960
 961 void __ext4_grp_locked_error(const char *function, unsigned int line,
 962                              struct super_block *sb, ext4_group_t grp,
 963                              unsigned long ino, ext4_fsblk_t block,
 964                              const char *fmt, ...)
 965 __releases(bitlock)
 966 __acquires(bitlock)
 967 {
 968         struct va_format vaf;
 969         va_list args;
 970
 971         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
 972                 return;
 973
 974         trace_ext4_error(sb, function, line);
 975         if (ext4_error_ratelimit(sb)) {
 976                 va_start(args, fmt);
 977                 vaf.fmt = fmt;
 978                 vaf.va = &args;
 979                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
 980                        sb->s_id, function, line, grp);
 981                 if (ino)
 982                         printk(KERN_CONT "inode %lu: ", ino);
 983                 if (block)
 984                         printk(KERN_CONT "block %llu:",
 985                                (unsigned long long) block);
 986                 printk(KERN_CONT "%pV\n", &vaf);
 987                 va_end(args);
 988         }
 989
 990         if (test_opt(sb, ERRORS_CONT)) {
 991                 if (test_opt(sb, WARN_ON_ERROR))
 992                         WARN_ON_ONCE(1);
 993                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 994                 if (!bdev_read_only(sb->s_bdev)) {
 995                         save_error_info(sb, EFSCORRUPTED, ino, block, function,
 996                                         line);
 997                         schedule_work(&EXT4_SB(sb)->s_error_work);
 998                 }
 999                 return;
1000         }
1001         ext4_unlock_group(sb, grp);
1002         ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1003         /*
1004          * We only get here in the ERRORS_RO case; relocking the group
1005          * may be dangerous, but nothing bad will happen since the
1006          * filesystem will have already been marked read/only and the
1007          * journal has been aborted.  We return 1 as a hint to callers
1008          * who might what to use the return value from
1009          * ext4_grp_locked_error() to distinguish between the
1010          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1011          * aggressively from the ext4 function in question, with a
1012          * more appropriate error code.
1013          */
1014         ext4_lock_group(sb, grp);
1015         return;
1016 }
1017
1018 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1019                                      ext4_group_t group,
1020                                      unsigned int flags)
1021 {
1022         struct ext4_sb_info *sbi = EXT4_SB(sb);
1023         struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1024         struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1025         int ret;
1026
1027         if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1028                 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1029                                             &grp->bb_state);
1030                 if (!ret)
1031                         percpu_counter_sub(&sbi->s_freeclusters_counter,
1032                                            grp->bb_free);
1033         }
1034
1035         if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1036                 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1037                                             &grp->bb_state);
1038                 if (!ret && gdp) {
1039                         int count;
1040
1041                         count = ext4_free_inodes_count(sb, gdp);
1042                         percpu_counter_sub(&sbi->s_freeinodes_counter,
1043                                            count);
1044                 }
1045         }
1046 }
1047
1048 void ext4_update_dynamic_rev(struct super_block *sb)
1049 {
1050         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1051
1052         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1053                 return;
1054
1055         ext4_warning(sb,
1056                      "updating to rev %d because of new feature flag, "
1057                      "running e2fsck is recommended",
1058                      EXT4_DYNAMIC_REV);
1059
1060         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1061         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1062         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1063         /* leave es->s_feature_*compat flags alone */
1064         /* es->s_uuid will be set by e2fsck if empty */
1065
1066         /*
1067          * The rest of the superblock fields should be zero, and if not it
1068          * means they are likely already in use, so leave them alone.  We
1069          * can leave it up to e2fsck to clean up any inconsistencies there.
1070          */
1071 }
1072
1073 /*
1074  * Open the external journal device
1075  */
1076 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1077 {
1078         struct block_device *bdev;
1079
1080         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1081         if (IS_ERR(bdev))
1082                 goto fail;
1083         return bdev;
1084
1085 fail:
1086         ext4_msg(sb, KERN_ERR,
1087                  "failed to open journal device unknown-block(%u,%u) %ld",
1088                  MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1089         return NULL;
1090 }
1091
1092 /*
1093  * Release the journal device
1094  */
1095 static void ext4_blkdev_put(struct block_device *bdev)
1096 {
1097         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1098 }
1099
1100 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1101 {
1102         struct block_device *bdev;
1103         bdev = sbi->s_journal_bdev;
1104         if (bdev) {
1105                 ext4_blkdev_put(bdev);
1106                 sbi->s_journal_bdev = NULL;
1107         }
1108 }
1109
1110 static inline struct inode *orphan_list_entry(struct list_head *l)
1111 {
1112         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1113 }
1114
1115 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1116 {
1117         struct list_head *l;
1118
1119         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1120                  le32_to_cpu(sbi->s_es->s_last_orphan));
1121
1122         printk(KERN_ERR "sb_info orphan list:\n");
1123         list_for_each(l, &sbi->s_orphan) {
1124                 struct inode *inode = orphan_list_entry(l);
1125                 printk(KERN_ERR "  "
1126                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1127                        inode->i_sb->s_id, inode->i_ino, inode,
1128                        inode->i_mode, inode->i_nlink,
1129                        NEXT_ORPHAN(inode));
1130         }
1131 }
1132
1133 #ifdef CONFIG_QUOTA
1134 static int ext4_quota_off(struct super_block *sb, int type);
1135
1136 static inline void ext4_quota_off_umount(struct super_block *sb)
1137 {
1138         int type;
1139
1140         /* Use our quota_off function to clear inode flags etc. */
1141         for (type = 0; type < EXT4_MAXQUOTAS; type++)
1142                 ext4_quota_off(sb, type);
1143 }
1144
1145 /*
1146  * This is a helper function which is used in the mount/remount
1147  * codepaths (which holds s_umount) to fetch the quota file name.
1148  */
1149 static inline char *get_qf_name(struct super_block *sb,
1150                                 struct ext4_sb_info *sbi,
1151                                 int type)
1152 {
1153         return rcu_dereference_protected(sbi->s_qf_names[type],
1154                                          lockdep_is_held(&sb->s_umount));
1155 }
1156 #else
1157 static inline void ext4_quota_off_umount(struct super_block *sb)
1158 {
1159 }
1160 #endif
1161
1162 static void ext4_put_super(struct super_block *sb)
1163 {
1164         struct ext4_sb_info *sbi = EXT4_SB(sb);
1165         struct ext4_super_block *es = sbi->s_es;
1166         struct buffer_head **group_desc;
1167         struct flex_groups **flex_groups;
1168         int aborted = 0;
1169         int i, err;
1170
1171         ext4_unregister_li_request(sb);
1172         ext4_quota_off_umount(sb);
1173
1174         flush_work(&sbi->s_error_work);
1175         destroy_workqueue(sbi->rsv_conversion_wq);
1176
1177         /*
1178          * Unregister sysfs before destroying jbd2 journal.
1179          * Since we could still access attr_journal_task attribute via sysfs
1180          * path which could have sbi->s_journal->j_task as NULL
1181          */
1182         ext4_unregister_sysfs(sb);
1183
1184         if (sbi->s_journal) {
1185                 aborted = is_journal_aborted(sbi->s_journal);
1186                 err = jbd2_journal_destroy(sbi->s_journal);
1187                 sbi->s_journal = NULL;
1188                 if ((err < 0) && !aborted) {
1189                         ext4_abort(sb, -err, "Couldn't clean up the journal");
1190                 }
1191         }
1192
1193         ext4_es_unregister_shrinker(sbi);
1194         del_timer_sync(&sbi->s_err_report);
1195         ext4_release_system_zone(sb);
1196         ext4_mb_release(sb);
1197         ext4_ext_release(sb);
1198
1199         if (!sb_rdonly(sb) && !aborted) {
1200                 ext4_clear_feature_journal_needs_recovery(sb);
1201                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1202         }
1203         if (!sb_rdonly(sb))
1204                 ext4_commit_super(sb);
1205
1206         rcu_read_lock();
1207         group_desc = rcu_dereference(sbi->s_group_desc);
1208         for (i = 0; i < sbi->s_gdb_count; i++)
1209                 brelse(group_desc[i]);
1210         kvfree(group_desc);
1211         flex_groups = rcu_dereference(sbi->s_flex_groups);
1212         if (flex_groups) {
1213                 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1214                         kvfree(flex_groups[i]);
1215                 kvfree(flex_groups);
1216         }
1217         rcu_read_unlock();
1218         percpu_counter_destroy(&sbi->s_freeclusters_counter);
1219         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1220         percpu_counter_destroy(&sbi->s_dirs_counter);
1221         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1222         percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1223         percpu_free_rwsem(&sbi->s_writepages_rwsem);
1224 #ifdef CONFIG_QUOTA
1225         for (i = 0; i < EXT4_MAXQUOTAS; i++)
1226                 kfree(get_qf_name(sb, sbi, i));
1227 #endif
1228
1229         /* Debugging code just in case the in-memory inode orphan list
1230          * isn't empty.  The on-disk one can be non-empty if we've
1231          * detected an error and taken the fs readonly, but the
1232          * in-memory list had better be clean by this point. */
1233         if (!list_empty(&sbi->s_orphan))
1234                 dump_orphan_list(sb, sbi);
1235         ASSERT(list_empty(&sbi->s_orphan));
1236
1237         sync_blockdev(sb->s_bdev);
1238         invalidate_bdev(sb->s_bdev);
1239         if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1240                 /*
1241                  * Invalidate the journal device's buffers.  We don't want them
1242                  * floating about in memory - the physical journal device may
1243                  * hotswapped, and it breaks the `ro-after' testing code.
1244                  */
1245                 sync_blockdev(sbi->s_journal_bdev);
1246                 invalidate_bdev(sbi->s_journal_bdev);
1247                 ext4_blkdev_remove(sbi);
1248         }
1249
1250         ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1251         sbi->s_ea_inode_cache = NULL;
1252
1253         ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1254         sbi->s_ea_block_cache = NULL;
1255
1256         ext4_stop_mmpd(sbi);
1257
1258         brelse(sbi->s_sbh);
1259         sb->s_fs_info = NULL;
1260         /*
1261          * Now that we are completely done shutting down the
1262          * superblock, we need to actually destroy the kobject.
1263          */
1264         kobject_put(&sbi->s_kobj);
1265         wait_for_completion(&sbi->s_kobj_unregister);
1266         if (sbi->s_chksum_driver)
1267                 crypto_free_shash(sbi->s_chksum_driver);
1268         kfree(sbi->s_blockgroup_lock);
1269         fs_put_dax(sbi->s_daxdev);
1270         fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1271 #ifdef CONFIG_UNICODE
1272         utf8_unload(sb->s_encoding);
1273 #endif
1274         kfree(sbi);
1275 }
1276
1277 static struct kmem_cache *ext4_inode_cachep;
1278
1279 /*
1280  * Called inside transaction, so use GFP_NOFS
1281  */
1282 static struct inode *ext4_alloc_inode(struct super_block *sb)
1283 {
1284         struct ext4_inode_info *ei;
1285
1286         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1287         if (!ei)
1288                 return NULL;
1289
1290         inode_set_iversion(&ei->vfs_inode, 1);
1291         spin_lock_init(&ei->i_raw_lock);
1292         INIT_LIST_HEAD(&ei->i_prealloc_list);
1293         atomic_set(&ei->i_prealloc_active, 0);
1294         spin_lock_init(&ei->i_prealloc_lock);
1295         ext4_es_init_tree(&ei->i_es_tree);
1296         rwlock_init(&ei->i_es_lock);
1297         INIT_LIST_HEAD(&ei->i_es_list);
1298         ei->i_es_all_nr = 0;
1299         ei->i_es_shk_nr = 0;
1300         ei->i_es_shrink_lblk = 0;
1301         ei->i_reserved_data_blocks = 0;
1302         spin_lock_init(&(ei->i_block_reservation_lock));
1303         ext4_init_pending_tree(&ei->i_pending_tree);
1304 #ifdef CONFIG_QUOTA
1305         ei->i_reserved_quota = 0;
1306         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1307 #endif
1308         ei->jinode = NULL;
1309         INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1310         spin_lock_init(&ei->i_completed_io_lock);
1311         ei->i_sync_tid = 0;
1312         ei->i_datasync_tid = 0;
1313         atomic_set(&ei->i_unwritten, 0);
1314         INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1315         ext4_fc_init_inode(&ei->vfs_inode);
1316         mutex_init(&ei->i_fc_lock);
1317         return &ei->vfs_inode;
1318 }
1319
1320 static int ext4_drop_inode(struct inode *inode)
1321 {
1322         int drop = generic_drop_inode(inode);
1323
1324         if (!drop)
1325                 drop = fscrypt_drop_inode(inode);
1326
1327         trace_ext4_drop_inode(inode, drop);
1328         return drop;
1329 }
1330
1331 static void ext4_free_in_core_inode(struct inode *inode)
1332 {
1333         fscrypt_free_inode(inode);
1334         if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1335                 pr_warn("%s: inode %ld still in fc list",
1336                         __func__, inode->i_ino);
1337         }
1338         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1339 }
1340
1341 static void ext4_destroy_inode(struct inode *inode)
1342 {
1343         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1344                 ext4_msg(inode->i_sb, KERN_ERR,
1345                          "Inode %lu (%p): orphan list check failed!",
1346                          inode->i_ino, EXT4_I(inode));
1347                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1348                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
1349                                 true);
1350                 dump_stack();
1351         }
1352 }
1353
1354 static void init_once(void *foo)
1355 {
1356         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1357
1358         INIT_LIST_HEAD(&ei->i_orphan);
1359         init_rwsem(&ei->xattr_sem);
1360         init_rwsem(&ei->i_data_sem);
1361         inode_init_once(&ei->vfs_inode);
1362         ext4_fc_init_inode(&ei->vfs_inode);
1363 }
1364
1365 static int __init init_inodecache(void)
1366 {
1367         ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1368                                 sizeof(struct ext4_inode_info), 0,
1369                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1370                                         SLAB_ACCOUNT),
1371                                 offsetof(struct ext4_inode_info, i_data),
1372                                 sizeof_field(struct ext4_inode_info, i_data),
1373                                 init_once);
1374         if (ext4_inode_cachep == NULL)
1375                 return -ENOMEM;
1376         return 0;
1377 }
1378
1379 static void destroy_inodecache(void)
1380 {
1381         /*
1382          * Make sure all delayed rcu free inodes are flushed before we
1383          * destroy cache.
1384          */
1385         rcu_barrier();
1386         kmem_cache_destroy(ext4_inode_cachep);
1387 }
1388
1389 void ext4_clear_inode(struct inode *inode)
1390 {
1391         ext4_fc_del(inode);
1392         invalidate_inode_buffers(inode);
1393         clear_inode(inode);
1394         ext4_discard_preallocations(inode, 0);
1395         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1396         dquot_drop(inode);
1397         if (EXT4_I(inode)->jinode) {
1398                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1399                                                EXT4_I(inode)->jinode);
1400                 jbd2_free_inode(EXT4_I(inode)->jinode);
1401                 EXT4_I(inode)->jinode = NULL;
1402         }
1403         fscrypt_put_encryption_info(inode);
1404         fsverity_cleanup_inode(inode);
1405 }
1406
1407 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1408                                         u64 ino, u32 generation)
1409 {
1410         struct inode *inode;
1411
1412         /*
1413          * Currently we don't know the generation for parent directory, so
1414          * a generation of 0 means "accept any"
1415          */
1416         inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1417         if (IS_ERR(inode))
1418                 return ERR_CAST(inode);
1419         if (generation && inode->i_generation != generation) {
1420                 iput(inode);
1421                 return ERR_PTR(-ESTALE);
1422         }
1423
1424         return inode;
1425 }
1426
1427 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1428                                         int fh_len, int fh_type)
1429 {
1430         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1431                                     ext4_nfs_get_inode);
1432 }
1433
1434 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1435                                         int fh_len, int fh_type)
1436 {
1437         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1438                                     ext4_nfs_get_inode);
1439 }
1440
1441 static int ext4_nfs_commit_metadata(struct inode *inode)
1442 {
1443         struct writeback_control wbc = {
1444                 .sync_mode = WB_SYNC_ALL
1445         };
1446
1447         trace_ext4_nfs_commit_metadata(inode);
1448         return ext4_write_inode(inode, &wbc);
1449 }
1450
1451 #ifdef CONFIG_FS_ENCRYPTION
1452 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1453 {
1454         return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1455                                  EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1456 }
1457
1458 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1459                                                         void *fs_data)
1460 {
1461         handle_t *handle = fs_data;
1462         int res, res2, credits, retries = 0;
1463
1464         /*
1465          * Encrypting the root directory is not allowed because e2fsck expects
1466          * lost+found to exist and be unencrypted, and encrypting the root
1467          * directory would imply encrypting the lost+found directory as well as
1468          * the filename "lost+found" itself.
1469          */
1470         if (inode->i_ino == EXT4_ROOT_INO)
1471                 return -EPERM;
1472
1473         if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1474                 return -EINVAL;
1475
1476         if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1477                 return -EOPNOTSUPP;
1478
1479         res = ext4_convert_inline_data(inode);
1480         if (res)
1481                 return res;
1482
1483         /*
1484          * If a journal handle was specified, then the encryption context is
1485          * being set on a new inode via inheritance and is part of a larger
1486          * transaction to create the inode.  Otherwise the encryption context is
1487          * being set on an existing inode in its own transaction.  Only in the
1488          * latter case should the "retry on ENOSPC" logic be used.
1489          */
1490
1491         if (handle) {
1492                 res = ext4_xattr_set_handle(handle, inode,
1493                                             EXT4_XATTR_INDEX_ENCRYPTION,
1494                                             EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1495                                             ctx, len, 0);
1496                 if (!res) {
1497                         ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1498                         ext4_clear_inode_state(inode,
1499                                         EXT4_STATE_MAY_INLINE_DATA);
1500                         /*
1501                          * Update inode->i_flags - S_ENCRYPTED will be enabled,
1502                          * S_DAX may be disabled
1503                          */
1504                         ext4_set_inode_flags(inode, false);
1505                 }
1506                 return res;
1507         }
1508
1509         res = dquot_initialize(inode);
1510         if (res)
1511                 return res;
1512 retry:
1513         res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1514                                      &credits);
1515         if (res)
1516                 return res;
1517
1518         handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1519         if (IS_ERR(handle))
1520                 return PTR_ERR(handle);
1521
1522         res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1523                                     EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1524                                     ctx, len, 0);
1525         if (!res) {
1526                 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1527                 /*
1528                  * Update inode->i_flags - S_ENCRYPTED will be enabled,
1529                  * S_DAX may be disabled
1530                  */
1531                 ext4_set_inode_flags(inode, false);
1532                 res = ext4_mark_inode_dirty(handle, inode);
1533                 if (res)
1534                         EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1535         }
1536         res2 = ext4_journal_stop(handle);
1537
1538         if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1539                 goto retry;
1540         if (!res)
1541                 res = res2;
1542         return res;
1543 }
1544
1545 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1546 {
1547         return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1548 }
1549
1550 static bool ext4_has_stable_inodes(struct super_block *sb)
1551 {
1552         return ext4_has_feature_stable_inodes(sb);
1553 }
1554
1555 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1556                                        int *ino_bits_ret, int *lblk_bits_ret)
1557 {
1558         *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1559         *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1560 }
1561
1562 static const struct fscrypt_operations ext4_cryptops = {
1563         .key_prefix             = "ext4:",
1564         .get_context            = ext4_get_context,
1565         .set_context            = ext4_set_context,
1566         .get_dummy_policy       = ext4_get_dummy_policy,
1567         .empty_dir              = ext4_empty_dir,
1568         .max_namelen            = EXT4_NAME_LEN,
1569         .has_stable_inodes      = ext4_has_stable_inodes,
1570         .get_ino_and_lblk_bits  = ext4_get_ino_and_lblk_bits,
1571 };
1572 #endif
1573
1574 #ifdef CONFIG_QUOTA
1575 static const char * const quotatypes[] = INITQFNAMES;
1576 #define QTYPE2NAME(t) (quotatypes[t])
1577
1578 static int ext4_write_dquot(struct dquot *dquot);
1579 static int ext4_acquire_dquot(struct dquot *dquot);
1580 static int ext4_release_dquot(struct dquot *dquot);
1581 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1582 static int ext4_write_info(struct super_block *sb, int type);
1583 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1584                          const struct path *path);
1585 static int ext4_quota_on_mount(struct super_block *sb, int type);
1586 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1587                                size_t len, loff_t off);
1588 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1589                                 const char *data, size_t len, loff_t off);
1590 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1591                              unsigned int flags);
1592 static int ext4_enable_quotas(struct super_block *sb);
1593
1594 static struct dquot **ext4_get_dquots(struct inode *inode)
1595 {
1596         return EXT4_I(inode)->i_dquot;
1597 }
1598
1599 static const struct dquot_operations ext4_quota_operations = {
1600         .get_reserved_space     = ext4_get_reserved_space,
1601         .write_dquot            = ext4_write_dquot,
1602         .acquire_dquot          = ext4_acquire_dquot,
1603         .release_dquot          = ext4_release_dquot,
1604         .mark_dirty             = ext4_mark_dquot_dirty,
1605         .write_info             = ext4_write_info,
1606         .alloc_dquot            = dquot_alloc,
1607         .destroy_dquot          = dquot_destroy,
1608         .get_projid             = ext4_get_projid,
1609         .get_inode_usage        = ext4_get_inode_usage,
1610         .get_next_id            = dquot_get_next_id,
1611 };
1612
1613 static const struct quotactl_ops ext4_qctl_operations = {
1614         .quota_on       = ext4_quota_on,
1615         .quota_off      = ext4_quota_off,
1616         .quota_sync     = dquot_quota_sync,
1617         .get_state      = dquot_get_state,
1618         .set_info       = dquot_set_dqinfo,
1619         .get_dqblk      = dquot_get_dqblk,
1620         .set_dqblk      = dquot_set_dqblk,
1621         .get_nextdqblk  = dquot_get_next_dqblk,
1622 };
1623 #endif
1624
1625 static const struct super_operations ext4_sops = {
1626         .alloc_inode    = ext4_alloc_inode,
1627         .free_inode     = ext4_free_in_core_inode,
1628         .destroy_inode  = ext4_destroy_inode,
1629         .write_inode    = ext4_write_inode,
1630         .dirty_inode    = ext4_dirty_inode,
1631         .drop_inode     = ext4_drop_inode,
1632         .evict_inode    = ext4_evict_inode,
1633         .put_super      = ext4_put_super,
1634         .sync_fs        = ext4_sync_fs,
1635         .freeze_fs      = ext4_freeze,
1636         .unfreeze_fs    = ext4_unfreeze,
1637         .statfs         = ext4_statfs,
1638         .remount_fs     = ext4_remount,
1639         .show_options   = ext4_show_options,
1640 #ifdef CONFIG_QUOTA
1641         .quota_read     = ext4_quota_read,
1642         .quota_write    = ext4_quota_write,
1643         .get_dquots     = ext4_get_dquots,
1644 #endif
1645 };
1646
1647 static const struct export_operations ext4_export_ops = {
1648         .fh_to_dentry = ext4_fh_to_dentry,
1649         .fh_to_parent = ext4_fh_to_parent,
1650         .get_parent = ext4_get_parent,
1651         .commit_metadata = ext4_nfs_commit_metadata,
1652 };
1653
1654 enum {
1655         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1656         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1657         Opt_nouid32, Opt_debug, Opt_removed,
1658         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1659         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1660         Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1661         Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1662         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1663         Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1664         Opt_inlinecrypt,
1665         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1666         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1667         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1668         Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1669         Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1670         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1671         Opt_nowarn_on_error, Opt_mblk_io_submit,
1672         Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1673         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1674         Opt_inode_readahead_blks, Opt_journal_ioprio,
1675         Opt_dioread_nolock, Opt_dioread_lock,
1676         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1677         Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1678         Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1679 #ifdef CONFIG_EXT4_DEBUG
1680         Opt_fc_debug_max_replay, Opt_fc_debug_force
1681 #endif
1682 };
1683
1684 static const match_table_t tokens = {
1685         {Opt_bsd_df, "bsddf"},
1686         {Opt_minix_df, "minixdf"},
1687         {Opt_grpid, "grpid"},
1688         {Opt_grpid, "bsdgroups"},
1689         {Opt_nogrpid, "nogrpid"},
1690         {Opt_nogrpid, "sysvgroups"},
1691         {Opt_resgid, "resgid=%u"},
1692         {Opt_resuid, "resuid=%u"},
1693         {Opt_sb, "sb=%u"},
1694         {Opt_err_cont, "errors=continue"},
1695         {Opt_err_panic, "errors=panic"},
1696         {Opt_err_ro, "errors=remount-ro"},
1697         {Opt_nouid32, "nouid32"},
1698         {Opt_debug, "debug"},
1699         {Opt_removed, "oldalloc"},
1700         {Opt_removed, "orlov"},
1701         {Opt_user_xattr, "user_xattr"},
1702         {Opt_nouser_xattr, "nouser_xattr"},
1703         {Opt_acl, "acl"},
1704         {Opt_noacl, "noacl"},
1705         {Opt_noload, "norecovery"},
1706         {Opt_noload, "noload"},
1707         {Opt_removed, "nobh"},
1708         {Opt_removed, "bh"},
1709         {Opt_commit, "commit=%u"},
1710         {Opt_min_batch_time, "min_batch_time=%u"},
1711         {Opt_max_batch_time, "max_batch_time=%u"},
1712         {Opt_journal_dev, "journal_dev=%u"},
1713         {Opt_journal_path, "journal_path=%s"},
1714         {Opt_journal_checksum, "journal_checksum"},
1715         {Opt_nojournal_checksum, "nojournal_checksum"},
1716         {Opt_journal_async_commit, "journal_async_commit"},
1717         {Opt_abort, "abort"},
1718         {Opt_data_journal, "data=journal"},
1719         {Opt_data_ordered, "data=ordered"},
1720         {Opt_data_writeback, "data=writeback"},
1721         {Opt_data_err_abort, "data_err=abort"},
1722         {Opt_data_err_ignore, "data_err=ignore"},
1723         {Opt_offusrjquota, "usrjquota="},
1724         {Opt_usrjquota, "usrjquota=%s"},
1725         {Opt_offgrpjquota, "grpjquota="},
1726         {Opt_grpjquota, "grpjquota=%s"},
1727         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1728         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1729         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1730         {Opt_grpquota, "grpquota"},
1731         {Opt_noquota, "noquota"},
1732         {Opt_quota, "quota"},
1733         {Opt_usrquota, "usrquota"},
1734         {Opt_prjquota, "prjquota"},
1735         {Opt_barrier, "barrier=%u"},
1736         {Opt_barrier, "barrier"},
1737         {Opt_nobarrier, "nobarrier"},
1738         {Opt_i_version, "i_version"},
1739         {Opt_dax, "dax"},
1740         {Opt_dax_always, "dax=always"},
1741         {Opt_dax_inode, "dax=inode"},
1742         {Opt_dax_never, "dax=never"},
1743         {Opt_stripe, "stripe=%u"},
1744         {Opt_delalloc, "delalloc"},
1745         {Opt_warn_on_error, "warn_on_error"},
1746         {Opt_nowarn_on_error, "nowarn_on_error"},
1747         {Opt_lazytime, "lazytime"},
1748         {Opt_nolazytime, "nolazytime"},
1749         {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1750         {Opt_nodelalloc, "nodelalloc"},
1751         {Opt_removed, "mblk_io_submit"},
1752         {Opt_removed, "nomblk_io_submit"},
1753         {Opt_block_validity, "block_validity"},
1754         {Opt_noblock_validity, "noblock_validity"},
1755         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1756         {Opt_journal_ioprio, "journal_ioprio=%u"},
1757         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1758         {Opt_auto_da_alloc, "auto_da_alloc"},
1759         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1760         {Opt_dioread_nolock, "dioread_nolock"},
1761         {Opt_dioread_lock, "nodioread_nolock"},
1762         {Opt_dioread_lock, "dioread_lock"},
1763         {Opt_discard, "discard"},
1764         {Opt_nodiscard, "nodiscard"},
1765         {Opt_init_itable, "init_itable=%u"},
1766         {Opt_init_itable, "init_itable"},
1767         {Opt_noinit_itable, "noinit_itable"},
1768 #ifdef CONFIG_EXT4_DEBUG
1769         {Opt_fc_debug_force, "fc_debug_force"},
1770         {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1771 #endif
1772         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1773         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1774         {Opt_test_dummy_encryption, "test_dummy_encryption"},
1775         {Opt_inlinecrypt, "inlinecrypt"},
1776         {Opt_nombcache, "nombcache"},
1777         {Opt_nombcache, "no_mbcache"},  /* for backward compatibility */
1778         {Opt_removed, "prefetch_block_bitmaps"},
1779         {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1780         {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1781         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1782         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1783         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1784         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1785         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1786         {Opt_err, NULL},
1787 };
1788
1789 static ext4_fsblk_t get_sb_block(void **data)
1790 {
1791         ext4_fsblk_t    sb_block;
1792         char            *options = (char *) *data;
1793
1794         if (!options || strncmp(options, "sb=", 3) != 0)
1795                 return 1;       /* Default location */
1796
1797         options += 3;
1798         /* TODO: use simple_strtoll with >32bit ext4 */
1799         sb_block = simple_strtoul(options, &options, 0);
1800         if (*options && *options != ',') {
1801                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1802                        (char *) *data);
1803                 return 1;
1804         }
1805         if (*options == ',')
1806                 options++;
1807         *data = (void *) options;
1808
1809         return sb_block;
1810 }
1811
1812 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1813 #define DEFAULT_MB_OPTIMIZE_SCAN        (-1)
1814
1815 static const char deprecated_msg[] =
1816         "Mount option \"%s\" will be removed by %s\n"
1817         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1818
1819 #ifdef CONFIG_QUOTA
1820 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1821 {
1822         struct ext4_sb_info *sbi = EXT4_SB(sb);
1823         char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1824         int ret = -1;
1825
1826         if (sb_any_quota_loaded(sb) && !old_qname) {
1827                 ext4_msg(sb, KERN_ERR,
1828                         "Cannot change journaled "
1829                         "quota options when quota turned on");
1830                 return -1;
1831         }
1832         if (ext4_has_feature_quota(sb)) {
1833                 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1834                          "ignored when QUOTA feature is enabled");
1835                 return 1;
1836         }
1837         qname = match_strdup(args);
1838         if (!qname) {
1839                 ext4_msg(sb, KERN_ERR,
1840                         "Not enough memory for storing quotafile name");
1841                 return -1;
1842         }
1843         if (old_qname) {
1844                 if (strcmp(old_qname, qname) == 0)
1845                         ret = 1;
1846                 else
1847                         ext4_msg(sb, KERN_ERR,
1848                                  "%s quota file already specified",
1849                                  QTYPE2NAME(qtype));
1850                 goto errout;
1851         }
1852         if (strchr(qname, '/')) {
1853                 ext4_msg(sb, KERN_ERR,
1854                         "quotafile must be on filesystem root");
1855                 goto errout;
1856         }
1857         rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1858         set_opt(sb, QUOTA);
1859         return 1;
1860 errout:
1861         kfree(qname);
1862         return ret;
1863 }
1864
1865 static int clear_qf_name(struct super_block *sb, int qtype)
1866 {
1867
1868         struct ext4_sb_info *sbi = EXT4_SB(sb);
1869         char *old_qname = get_qf_name(sb, sbi, qtype);
1870
1871         if (sb_any_quota_loaded(sb) && old_qname) {
1872                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1873                         " when quota turned on");
1874                 return -1;
1875         }
1876         rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1877         synchronize_rcu();
1878         kfree(old_qname);
1879         return 1;
1880 }
1881 #endif
1882
1883 #define MOPT_SET        0x0001
1884 #define MOPT_CLEAR      0x0002
1885 #define MOPT_NOSUPPORT  0x0004
1886 #define MOPT_EXPLICIT   0x0008
1887 #define MOPT_CLEAR_ERR  0x0010
1888 #define MOPT_GTE0       0x0020
1889 #ifdef CONFIG_QUOTA
1890 #define MOPT_Q          0
1891 #define MOPT_QFMT       0x0040
1892 #else
1893 #define MOPT_Q          MOPT_NOSUPPORT
1894 #define MOPT_QFMT       MOPT_NOSUPPORT
1895 #endif
1896 #define MOPT_DATAJ      0x0080
1897 #define MOPT_NO_EXT2    0x0100
1898 #define MOPT_NO_EXT3    0x0200
1899 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1900 #define MOPT_STRING     0x0400
1901 #define MOPT_SKIP       0x0800
1902 #define MOPT_2          0x1000
1903
1904 static const struct mount_opts {
1905         int     token;
1906         int     mount_opt;
1907         int     flags;
1908 } ext4_mount_opts[] = {
1909         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1910         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1911         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1912         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1913         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1914         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1915         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1916          MOPT_EXT4_ONLY | MOPT_SET},
1917         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1918          MOPT_EXT4_ONLY | MOPT_CLEAR},
1919         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1920         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1921         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1922          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1923         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1924          MOPT_EXT4_ONLY | MOPT_CLEAR},
1925         {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1926         {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1927         {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1928          MOPT_EXT4_ONLY | MOPT_CLEAR},
1929         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1930          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1931         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1932                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1933          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1934         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1935         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1936         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1937         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1938         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1939          MOPT_NO_EXT2},
1940         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1941          MOPT_NO_EXT2},
1942         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1943         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1944         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1945         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1946         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1947         {Opt_commit, 0, MOPT_GTE0},
1948         {Opt_max_batch_time, 0, MOPT_GTE0},
1949         {Opt_min_batch_time, 0, MOPT_GTE0},
1950         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1951         {Opt_init_itable, 0, MOPT_GTE0},
1952         {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1953         {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1954                 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1955         {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1956                 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1957         {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1958                 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1959         {Opt_stripe, 0, MOPT_GTE0},
1960         {Opt_resuid, 0, MOPT_GTE0},
1961         {Opt_resgid, 0, MOPT_GTE0},
1962         {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1963         {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1964         {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1965         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1966         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1967         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1968          MOPT_NO_EXT2 | MOPT_DATAJ},
1969         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1970         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1971 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1972         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1973         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1974 #else
1975         {Opt_acl, 0, MOPT_NOSUPPORT},
1976         {Opt_noacl, 0, MOPT_NOSUPPORT},
1977 #endif
1978         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1979         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1980         {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1981         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1982         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1983                                                         MOPT_SET | MOPT_Q},
1984         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1985                                                         MOPT_SET | MOPT_Q},
1986         {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1987                                                         MOPT_SET | MOPT_Q},
1988         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1989                        EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1990                                                         MOPT_CLEAR | MOPT_Q},
1991         {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1992         {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1993         {Opt_offusrjquota, 0, MOPT_Q},
1994         {Opt_offgrpjquota, 0, MOPT_Q},
1995         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1996         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1997         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1998         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1999         {Opt_test_dummy_encryption, 0, MOPT_STRING},
2000         {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2001         {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2002          MOPT_SET},
2003         {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2004 #ifdef CONFIG_EXT4_DEBUG
2005         {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2006          MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2007         {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2008 #endif
2009         {Opt_err, 0, 0}
2010 };
2011
2012 #ifdef CONFIG_UNICODE
2013 static const struct ext4_sb_encodings {
2014         __u16 magic;
2015         char *name;
2016         char *version;
2017 } ext4_sb_encoding_map[] = {
2018         {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2019 };
2020
2021 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2022                                  const struct ext4_sb_encodings **encoding,
2023                                  __u16 *flags)
2024 {
2025         __u16 magic = le16_to_cpu(es->s_encoding);
2026         int i;
2027
2028         for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2029                 if (magic == ext4_sb_encoding_map[i].magic)
2030                         break;
2031
2032         if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2033                 return -EINVAL;
2034
2035         *encoding = &ext4_sb_encoding_map[i];
2036         *flags = le16_to_cpu(es->s_encoding_flags);
2037
2038         return 0;
2039 }
2040 #endif
2041
2042 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2043                                           const char *opt,
2044                                           const substring_t *arg,
2045                                           bool is_remount)
2046 {
2047 #ifdef CONFIG_FS_ENCRYPTION
2048         struct ext4_sb_info *sbi = EXT4_SB(sb);
2049         int err;
2050
2051         /*
2052          * This mount option is just for testing, and it's not worthwhile to
2053          * implement the extra complexity (e.g. RCU protection) that would be
2054          * needed to allow it to be set or changed during remount.  We do allow
2055          * it to be specified during remount, but only if there is no change.
2056          */
2057         if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2058                 ext4_msg(sb, KERN_WARNING,
2059                          "Can't set test_dummy_encryption on remount");
2060                 return -1;
2061         }
2062         err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2063                                                 &sbi->s_dummy_enc_policy);
2064         if (err) {
2065                 if (err == -EEXIST)
2066                         ext4_msg(sb, KERN_WARNING,
2067                                  "Can't change test_dummy_encryption on remount");
2068                 else if (err == -EINVAL)
2069                         ext4_msg(sb, KERN_WARNING,
2070                                  "Value of option \"%s\" is unrecognized", opt);
2071                 else
2072                         ext4_msg(sb, KERN_WARNING,
2073                                  "Error processing option \"%s\" [%d]",
2074                                  opt, err);
2075                 return -1;
2076         }
2077         ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2078 #else
2079         ext4_msg(sb, KERN_WARNING,
2080                  "Test dummy encryption mount option ignored");
2081 #endif
2082         return 1;
2083 }
2084
2085 struct ext4_parsed_options {
2086         unsigned long journal_devnum;
2087         unsigned int journal_ioprio;
2088         int mb_optimize_scan;
2089 };
2090
2091 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2092                             substring_t *args, struct ext4_parsed_options *parsed_opts,
2093                             int is_remount)
2094 {
2095         struct ext4_sb_info *sbi = EXT4_SB(sb);
2096         const struct mount_opts *m;
2097         kuid_t uid;
2098         kgid_t gid;
2099         int arg = 0;
2100
2101 #ifdef CONFIG_QUOTA
2102         if (token == Opt_usrjquota)
2103                 return set_qf_name(sb, USRQUOTA, &args[0]);
2104         else if (token == Opt_grpjquota)
2105                 return set_qf_name(sb, GRPQUOTA, &args[0]);
2106         else if (token == Opt_offusrjquota)
2107                 return clear_qf_name(sb, USRQUOTA);
2108         else if (token == Opt_offgrpjquota)
2109                 return clear_qf_name(sb, GRPQUOTA);
2110 #endif
2111         switch (token) {
2112         case Opt_noacl:
2113         case Opt_nouser_xattr:
2114                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2115                 break;
2116         case Opt_sb:
2117                 return 1;       /* handled by get_sb_block() */
2118         case Opt_removed:
2119                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2120                 return 1;
2121         case Opt_abort:
2122                 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2123                 return 1;
2124         case Opt_i_version:
2125                 sb->s_flags |= SB_I_VERSION;
2126                 return 1;
2127         case Opt_lazytime:
2128                 sb->s_flags |= SB_LAZYTIME;
2129                 return 1;
2130         case Opt_nolazytime:
2131                 sb->s_flags &= ~SB_LAZYTIME;
2132                 return 1;
2133         case Opt_inlinecrypt:
2134 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2135                 sb->s_flags |= SB_INLINECRYPT;
2136 #else
2137                 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2138 #endif
2139                 return 1;
2140         }
2141
2142         for (m = ext4_mount_opts; m->token != Opt_err; m++)
2143                 if (token == m->token)
2144                         break;
2145
2146         if (m->token == Opt_err) {
2147                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2148                          "or missing value", opt);
2149                 return -1;
2150         }
2151
2152         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2153                 ext4_msg(sb, KERN_ERR,
2154                          "Mount option \"%s\" incompatible with ext2", opt);
2155                 return -1;
2156         }
2157         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2158                 ext4_msg(sb, KERN_ERR,
2159                          "Mount option \"%s\" incompatible with ext3", opt);
2160                 return -1;
2161         }
2162
2163         if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2164                 return -1;
2165         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2166                 return -1;
2167         if (m->flags & MOPT_EXPLICIT) {
2168                 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2169                         set_opt2(sb, EXPLICIT_DELALLOC);
2170                 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2171                         set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2172                 } else
2173                         return -1;
2174         }
2175         if (m->flags & MOPT_CLEAR_ERR)
2176                 clear_opt(sb, ERRORS_MASK);
2177         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2178                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2179                          "options when quota turned on");
2180                 return -1;
2181         }
2182
2183         if (m->flags & MOPT_NOSUPPORT) {
2184                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2185         } else if (token == Opt_commit) {
2186                 if (arg == 0)
2187                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2188                 else if (arg > INT_MAX / HZ) {
2189                         ext4_msg(sb, KERN_ERR,
2190                                  "Invalid commit interval %d, "
2191                                  "must be smaller than %d",
2192                                  arg, INT_MAX / HZ);
2193                         return -1;
2194                 }
2195                 sbi->s_commit_interval = HZ * arg;
2196         } else if (token == Opt_debug_want_extra_isize) {
2197                 if ((arg & 1) ||
2198                     (arg < 4) ||
2199                     (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2200                         ext4_msg(sb, KERN_ERR,
2201                                  "Invalid want_extra_isize %d", arg);
2202                         return -1;
2203                 }
2204                 sbi->s_want_extra_isize = arg;
2205         } else if (token == Opt_max_batch_time) {
2206                 sbi->s_max_batch_time = arg;
2207         } else if (token == Opt_min_batch_time) {
2208                 sbi->s_min_batch_time = arg;
2209         } else if (token == Opt_inode_readahead_blks) {
2210                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2211                         ext4_msg(sb, KERN_ERR,
2212                                  "EXT4-fs: inode_readahead_blks must be "
2213                                  "0 or a power of 2 smaller than 2^31");
2214                         return -1;
2215                 }
2216                 sbi->s_inode_readahead_blks = arg;
2217         } else if (token == Opt_init_itable) {
2218                 set_opt(sb, INIT_INODE_TABLE);
2219                 if (!args->from)
2220                         arg = EXT4_DEF_LI_WAIT_MULT;
2221                 sbi->s_li_wait_mult = arg;
2222         } else if (token == Opt_max_dir_size_kb) {
2223                 sbi->s_max_dir_size_kb = arg;
2224 #ifdef CONFIG_EXT4_DEBUG
2225         } else if (token == Opt_fc_debug_max_replay) {
2226                 sbi->s_fc_debug_max_replay = arg;
2227 #endif
2228         } else if (token == Opt_stripe) {
2229                 sbi->s_stripe = arg;
2230         } else if (token == Opt_resuid) {
2231                 uid = make_kuid(current_user_ns(), arg);
2232                 if (!uid_valid(uid)) {
2233                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2234                         return -1;
2235                 }
2236                 sbi->s_resuid = uid;
2237         } else if (token == Opt_resgid) {
2238                 gid = make_kgid(current_user_ns(), arg);
2239                 if (!gid_valid(gid)) {
2240                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2241                         return -1;
2242                 }
2243                 sbi->s_resgid = gid;
2244         } else if (token == Opt_journal_dev) {
2245                 if (is_remount) {
2246                         ext4_msg(sb, KERN_ERR,
2247                                  "Cannot specify journal on remount");
2248                         return -1;
2249                 }
2250                 parsed_opts->journal_devnum = arg;
2251         } else if (token == Opt_journal_path) {
2252                 char *journal_path;
2253                 struct inode *journal_inode;
2254                 struct path path;
2255                 int error;
2256
2257                 if (is_remount) {
2258                         ext4_msg(sb, KERN_ERR,
2259                                  "Cannot specify journal on remount");
2260                         return -1;
2261                 }
2262                 journal_path = match_strdup(&args[0]);
2263                 if (!journal_path) {
2264                         ext4_msg(sb, KERN_ERR, "error: could not dup "
2265                                 "journal device string");
2266                         return -1;
2267                 }
2268
2269                 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2270                 if (error) {
2271                         ext4_msg(sb, KERN_ERR, "error: could not find "
2272                                 "journal device path: error %d", error);
2273                         kfree(journal_path);
2274                         return -1;
2275                 }
2276
2277                 journal_inode = d_inode(path.dentry);
2278                 if (!S_ISBLK(journal_inode->i_mode)) {
2279                         ext4_msg(sb, KERN_ERR, "error: journal path %s "
2280                                 "is not a block device", journal_path);
2281                         path_put(&path);
2282                         kfree(journal_path);
2283                         return -1;
2284                 }
2285
2286                 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2287                 path_put(&path);
2288                 kfree(journal_path);
2289         } else if (token == Opt_journal_ioprio) {
2290                 if (arg > 7) {
2291                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2292                                  " (must be 0-7)");
2293                         return -1;
2294                 }
2295                 parsed_opts->journal_ioprio =
2296                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2297         } else if (token == Opt_test_dummy_encryption) {
2298                 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2299                                                       is_remount);
2300         } else if (m->flags & MOPT_DATAJ) {
2301                 if (is_remount) {
2302                         if (!sbi->s_journal)
2303                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2304                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2305                                 ext4_msg(sb, KERN_ERR,
2306                                          "Cannot change data mode on remount");
2307                                 return -1;
2308                         }
2309                 } else {
2310                         clear_opt(sb, DATA_FLAGS);
2311                         sbi->s_mount_opt |= m->mount_opt;
2312                 }
2313 #ifdef CONFIG_QUOTA
2314         } else if (m->flags & MOPT_QFMT) {
2315                 if (sb_any_quota_loaded(sb) &&
2316                     sbi->s_jquota_fmt != m->mount_opt) {
2317                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2318                                  "quota options when quota turned on");
2319                         return -1;
2320                 }
2321                 if (ext4_has_feature_quota(sb)) {
2322                         ext4_msg(sb, KERN_INFO,
2323                                  "Quota format mount options ignored "
2324                                  "when QUOTA feature is enabled");
2325                         return 1;
2326                 }
2327                 sbi->s_jquota_fmt = m->mount_opt;
2328 #endif
2329         } else if (token == Opt_dax || token == Opt_dax_always ||
2330                    token == Opt_dax_inode || token == Opt_dax_never) {
2331 #ifdef CONFIG_FS_DAX
2332                 switch (token) {
2333                 case Opt_dax:
2334                 case Opt_dax_always:
2335                         if (is_remount &&
2336                             (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2337                              (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2338                         fail_dax_change_remount:
2339                                 ext4_msg(sb, KERN_ERR, "can't change "
2340                                          "dax mount option while remounting");
2341                                 return -1;
2342                         }
2343                         if (is_remount &&
2344                             (test_opt(sb, DATA_FLAGS) ==
2345                              EXT4_MOUNT_JOURNAL_DATA)) {
2346                                     ext4_msg(sb, KERN_ERR, "can't mount with "
2347                                              "both data=journal and dax");
2348                                     return -1;
2349                         }
2350                         ext4_msg(sb, KERN_WARNING,
2351                                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2352                         sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2353                         sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2354                         break;
2355                 case Opt_dax_never:
2356                         if (is_remount &&
2357                             (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2358                              (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2359                                 goto fail_dax_change_remount;
2360                         sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2361                         sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2362                         break;
2363                 case Opt_dax_inode:
2364                         if (is_remount &&
2365                             ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2366                              (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2367                              !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2368                                 goto fail_dax_change_remount;
2369                         sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2370                         sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2371                         /* Strictly for printing options */
2372                         sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2373                         break;
2374                 }
2375 #else
2376                 ext4_msg(sb, KERN_INFO, "dax option not supported");
2377                 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2378                 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2379                 return -1;
2380 #endif
2381         } else if (token == Opt_data_err_abort) {
2382                 sbi->s_mount_opt |= m->mount_opt;
2383         } else if (token == Opt_data_err_ignore) {
2384                 sbi->s_mount_opt &= ~m->mount_opt;
2385         } else if (token == Opt_mb_optimize_scan) {
2386                 if (arg != 0 && arg != 1) {
2387                         ext4_msg(sb, KERN_WARNING,
2388                                  "mb_optimize_scan should be set to 0 or 1.");
2389                         return -1;
2390                 }
2391                 parsed_opts->mb_optimize_scan = arg;
2392         } else {
2393                 if (!args->from)
2394                         arg = 1;
2395                 if (m->flags & MOPT_CLEAR)
2396                         arg = !arg;
2397                 else if (unlikely(!(m->flags & MOPT_SET))) {
2398                         ext4_msg(sb, KERN_WARNING,
2399                                  "buggy handling of option %s", opt);
2400                         WARN_ON(1);
2401                         return -1;
2402                 }
2403                 if (m->flags & MOPT_2) {
2404                         if (arg != 0)
2405                                 sbi->s_mount_opt2 |= m->mount_opt;
2406                         else
2407                                 sbi->s_mount_opt2 &= ~m->mount_opt;
2408                 } else {
2409                         if (arg != 0)
2410                                 sbi->s_mount_opt |= m->mount_opt;
2411                         else
2412                                 sbi->s_mount_opt &= ~m->mount_opt;
2413                 }
2414         }
2415         return 1;
2416 }
2417
2418 static int parse_options(char *options, struct super_block *sb,
2419                          struct ext4_parsed_options *ret_opts,
2420                          int is_remount)
2421 {
2422         struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2423         char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2424         substring_t args[MAX_OPT_ARGS];
2425         int token;
2426
2427         if (!options)
2428                 return 1;
2429
2430         while ((p = strsep(&options, ",")) != NULL) {
2431                 if (!*p)
2432                         continue;
2433                 /*
2434                  * Initialize args struct so we know whether arg was
2435                  * found; some options take optional arguments.
2436                  */
2437                 args[0].to = args[0].from = NULL;
2438                 token = match_token(p, tokens, args);
2439                 if (handle_mount_opt(sb, p, token, args, ret_opts,
2440                                      is_remount) < 0)
2441                         return 0;
2442         }
2443 #ifdef CONFIG_QUOTA
2444         /*
2445          * We do the test below only for project quotas. 'usrquota' and
2446          * 'grpquota' mount options are allowed even without quota feature
2447          * to support legacy quotas in quota files.
2448          */
2449         if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2450                 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2451                          "Cannot enable project quota enforcement.");
2452                 return 0;
2453         }
2454         usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2455         grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2456         if (usr_qf_name || grp_qf_name) {
2457                 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2458                         clear_opt(sb, USRQUOTA);
2459
2460                 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2461                         clear_opt(sb, GRPQUOTA);
2462
2463                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2464                         ext4_msg(sb, KERN_ERR, "old and new quota "
2465                                         "format mixing");
2466                         return 0;
2467                 }
2468
2469                 if (!sbi->s_jquota_fmt) {
2470                         ext4_msg(sb, KERN_ERR, "journaled quota format "
2471                                         "not specified");
2472                         return 0;
2473                 }
2474         }
2475 #endif
2476         if (test_opt(sb, DIOREAD_NOLOCK)) {
2477                 int blocksize =
2478                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2479                 if (blocksize < PAGE_SIZE)
2480                         ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2481                                  "experimental mount option 'dioread_nolock' "
2482                                  "for blocksize < PAGE_SIZE");
2483         }
2484         return 1;
2485 }
2486
2487 static inline void ext4_show_quota_options(struct seq_file *seq,
2488                                            struct super_block *sb)
2489 {
2490 #if defined(CONFIG_QUOTA)
2491         struct ext4_sb_info *sbi = EXT4_SB(sb);
2492         char *usr_qf_name, *grp_qf_name;
2493
2494         if (sbi->s_jquota_fmt) {
2495                 char *fmtname = "";
2496
2497                 switch (sbi->s_jquota_fmt) {
2498                 case QFMT_VFS_OLD:
2499                         fmtname = "vfsold";
2500                         break;
2501                 case QFMT_VFS_V0:
2502                         fmtname = "vfsv0";
2503                         break;
2504                 case QFMT_VFS_V1:
2505                         fmtname = "vfsv1";
2506                         break;
2507                 }
2508                 seq_printf(seq, ",jqfmt=%s", fmtname);
2509         }
2510
2511         rcu_read_lock();
2512         usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2513         grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2514         if (usr_qf_name)
2515                 seq_show_option(seq, "usrjquota", usr_qf_name);
2516         if (grp_qf_name)
2517                 seq_show_option(seq, "grpjquota", grp_qf_name);
2518         rcu_read_unlock();
2519 #endif
2520 }
2521
2522 static const char *token2str(int token)
2523 {
2524         const struct match_token *t;
2525
2526         for (t = tokens; t->token != Opt_err; t++)
2527                 if (t->token == token && !strchr(t->pattern, '='))
2528                         break;
2529         return t->pattern;
2530 }
2531
2532 /*
2533  * Show an option if
2534  *  - it's set to a non-default value OR
2535  *  - if the per-sb default is different from the global default
2536  */
2537 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2538                               int nodefs)
2539 {
2540         struct ext4_sb_info *sbi = EXT4_SB(sb);
2541         struct ext4_super_block *es = sbi->s_es;
2542         int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2543         const struct mount_opts *m;
2544         char sep = nodefs ? '\n' : ',';
2545
2546 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2547 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2548
2549         if (sbi->s_sb_block != 1)
2550                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2551
2552         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2553                 int want_set = m->flags & MOPT_SET;
2554                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2555                     (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2556                         continue;
2557                 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2558                         continue; /* skip if same as the default */
2559                 if ((want_set &&
2560                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2561                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2562                         continue; /* select Opt_noFoo vs Opt_Foo */
2563                 SEQ_OPTS_PRINT("%s", token2str(m->token));
2564         }
2565
2566         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2567             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2568                 SEQ_OPTS_PRINT("resuid=%u",
2569                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2570         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2571             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2572                 SEQ_OPTS_PRINT("resgid=%u",
2573                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2574         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2575         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2576                 SEQ_OPTS_PUTS("errors=remount-ro");
2577         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2578                 SEQ_OPTS_PUTS("errors=continue");
2579         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2580                 SEQ_OPTS_PUTS("errors=panic");
2581         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2582                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2583         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2584                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2585         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2586                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2587         if (sb->s_flags & SB_I_VERSION)
2588                 SEQ_OPTS_PUTS("i_version");
2589         if (nodefs || sbi->s_stripe)
2590                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2591         if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2592                         (sbi->s_mount_opt ^ def_mount_opt)) {
2593                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2594                         SEQ_OPTS_PUTS("data=journal");
2595                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2596                         SEQ_OPTS_PUTS("data=ordered");
2597                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2598                         SEQ_OPTS_PUTS("data=writeback");
2599         }
2600         if (nodefs ||
2601             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2602                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2603                                sbi->s_inode_readahead_blks);
2604
2605         if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2606                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2607                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2608         if (nodefs || sbi->s_max_dir_size_kb)
2609                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2610         if (test_opt(sb, DATA_ERR_ABORT))
2611                 SEQ_OPTS_PUTS("data_err=abort");
2612
2613         fscrypt_show_test_dummy_encryption(seq, sep, sb);
2614
2615         if (sb->s_flags & SB_INLINECRYPT)
2616                 SEQ_OPTS_PUTS("inlinecrypt");
2617
2618         if (test_opt(sb, DAX_ALWAYS)) {
2619                 if (IS_EXT2_SB(sb))
2620                         SEQ_OPTS_PUTS("dax");
2621                 else
2622                         SEQ_OPTS_PUTS("dax=always");
2623         } else if (test_opt2(sb, DAX_NEVER)) {
2624                 SEQ_OPTS_PUTS("dax=never");
2625         } else if (test_opt2(sb, DAX_INODE)) {
2626                 SEQ_OPTS_PUTS("dax=inode");
2627         }
2628         ext4_show_quota_options(seq, sb);
2629         return 0;
2630 }
2631
2632 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2633 {
2634         return _ext4_show_options(seq, root->d_sb, 0);
2635 }
2636
2637 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2638 {
2639         struct super_block *sb = seq->private;
2640         int rc;
2641
2642         seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2643         rc = _ext4_show_options(seq, sb, 1);
2644         seq_puts(seq, "\n");
2645         return rc;
2646 }
2647
2648 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2649                             int read_only)
2650 {
2651         struct ext4_sb_info *sbi = EXT4_SB(sb);
2652         int err = 0;
2653
2654         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2655                 ext4_msg(sb, KERN_ERR, "revision level too high, "
2656                          "forcing read-only mode");
2657                 err = -EROFS;
2658                 goto done;
2659         }
2660         if (read_only)
2661                 goto done;
2662         if (!(sbi->s_mount_state & EXT4_VALID_FS))
2663                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2664                          "running e2fsck is recommended");
2665         else if (sbi->s_mount_state & EXT4_ERROR_FS)
2666                 ext4_msg(sb, KERN_WARNING,
2667                          "warning: mounting fs with errors, "
2668                          "running e2fsck is recommended");
2669         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2670                  le16_to_cpu(es->s_mnt_count) >=
2671                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2672                 ext4_msg(sb, KERN_WARNING,
2673                          "warning: maximal mount count reached, "
2674                          "running e2fsck is recommended");
2675         else if (le32_to_cpu(es->s_checkinterval) &&
2676                  (ext4_get_tstamp(es, s_lastcheck) +
2677                   le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2678                 ext4_msg(sb, KERN_WARNING,
2679                          "warning: checktime reached, "
2680                          "running e2fsck is recommended");
2681         if (!sbi->s_journal)
2682                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2683         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2684                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2685         le16_add_cpu(&es->s_mnt_count, 1);
2686         ext4_update_tstamp(es, s_mtime);
2687         if (sbi->s_journal)
2688                 ext4_set_feature_journal_needs_recovery(sb);
2689
2690         err = ext4_commit_super(sb);
2691 done:
2692         if (test_opt(sb, DEBUG))
2693                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2694                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2695                         sb->s_blocksize,
2696                         sbi->s_groups_count,
2697                         EXT4_BLOCKS_PER_GROUP(sb),
2698                         EXT4_INODES_PER_GROUP(sb),
2699                         sbi->s_mount_opt, sbi->s_mount_opt2);
2700
2701         cleancache_init_fs(sb);
2702         return err;
2703 }
2704
2705 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2706 {
2707         struct ext4_sb_info *sbi = EXT4_SB(sb);
2708         struct flex_groups **old_groups, **new_groups;
2709         int size, i, j;
2710
2711         if (!sbi->s_log_groups_per_flex)
2712                 return 0;
2713
2714         size = ext4_flex_group(sbi, ngroup - 1) + 1;
2715         if (size <= sbi->s_flex_groups_allocated)
2716                 return 0;
2717
2718         new_groups = kvzalloc(roundup_pow_of_two(size *
2719                               sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2720         if (!new_groups) {
2721                 ext4_msg(sb, KERN_ERR,
2722                          "not enough memory for %d flex group pointers", size);
2723                 return -ENOMEM;
2724         }
2725         for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2726                 new_groups[i] = kvzalloc(roundup_pow_of_two(
2727                                          sizeof(struct flex_groups)),
2728                                          GFP_KERNEL);
2729                 if (!new_groups[i]) {
2730                         for (j = sbi->s_flex_groups_allocated; j < i; j++)
2731                                 kvfree(new_groups[j]);
2732                         kvfree(new_groups);
2733                         ext4_msg(sb, KERN_ERR,
2734                                  "not enough memory for %d flex groups", size);
2735                         return -ENOMEM;
2736                 }
2737         }
2738         rcu_read_lock();
2739         old_groups = rcu_dereference(sbi->s_flex_groups);
2740         if (old_groups)
2741                 memcpy(new_groups, old_groups,
2742                        (sbi->s_flex_groups_allocated *
2743                         sizeof(struct flex_groups *)));
2744         rcu_read_unlock();
2745         rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2746         sbi->s_flex_groups_allocated = size;
2747         if (old_groups)
2748                 ext4_kvfree_array_rcu(old_groups);
2749         return 0;
2750 }
2751
2752 static int ext4_fill_flex_info(struct super_block *sb)
2753 {
2754         struct ext4_sb_info *sbi = EXT4_SB(sb);
2755         struct ext4_group_desc *gdp = NULL;
2756         struct flex_groups *fg;
2757         ext4_group_t flex_group;
2758         int i, err;
2759
2760         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2761         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2762                 sbi->s_log_groups_per_flex = 0;
2763                 return 1;
2764         }
2765
2766         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2767         if (err)
2768                 goto failed;
2769
2770         for (i = 0; i < sbi->s_groups_count; i++) {
2771                 gdp = ext4_get_group_desc(sb, i, NULL);
2772
2773                 flex_group = ext4_flex_group(sbi, i);
2774                 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2775                 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2776                 atomic64_add(ext4_free_group_clusters(sb, gdp),
2777                              &fg->free_clusters);
2778                 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2779         }
2780
2781         return 1;
2782 failed:
2783         return 0;
2784 }
2785
2786 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2787                                    struct ext4_group_desc *gdp)
2788 {
2789         int offset = offsetof(struct ext4_group_desc, bg_checksum);
2790         __u16 crc = 0;
2791         __le32 le_group = cpu_to_le32(block_group);
2792         struct ext4_sb_info *sbi = EXT4_SB(sb);
2793
2794         if (ext4_has_metadata_csum(sbi->s_sb)) {
2795                 /* Use new metadata_csum algorithm */
2796                 __u32 csum32;
2797                 __u16 dummy_csum = 0;
2798
2799                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2800                                      sizeof(le_group));
2801                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2802                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2803                                      sizeof(dummy_csum));
2804                 offset += sizeof(dummy_csum);
2805                 if (offset < sbi->s_desc_size)
2806                         csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2807                                              sbi->s_desc_size - offset);
2808
2809                 crc = csum32 & 0xFFFF;
2810                 goto out;
2811         }
2812
2813         /* old crc16 code */
2814         if (!ext4_has_feature_gdt_csum(sb))
2815                 return 0;
2816
2817         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2818         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2819         crc = crc16(crc, (__u8 *)gdp, offset);
2820         offset += sizeof(gdp->bg_checksum); /* skip checksum */
2821         /* for checksum of struct ext4_group_desc do the rest...*/
2822         if (ext4_has_feature_64bit(sb) &&
2823             offset < le16_to_cpu(sbi->s_es->s_desc_size))
2824                 crc = crc16(crc, (__u8 *)gdp + offset,
2825                             le16_to_cpu(sbi->s_es->s_desc_size) -
2826                                 offset);
2827
2828 out:
2829         return cpu_to_le16(crc);
2830 }
2831
2832 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2833                                 struct ext4_group_desc *gdp)
2834 {
2835         if (ext4_has_group_desc_csum(sb) &&
2836             (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2837                 return 0;
2838
2839         return 1;
2840 }
2841
2842 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2843                               struct ext4_group_desc *gdp)
2844 {
2845         if (!ext4_has_group_desc_csum(sb))
2846                 return;
2847         gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2848 }
2849
2850 /* Called at mount-time, super-block is locked */
2851 static int ext4_check_descriptors(struct super_block *sb,
2852                                   ext4_fsblk_t sb_block,
2853                                   ext4_group_t *first_not_zeroed)
2854 {
2855         struct ext4_sb_info *sbi = EXT4_SB(sb);
2856         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2857         ext4_fsblk_t last_block;
2858         ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2859         ext4_fsblk_t block_bitmap;
2860         ext4_fsblk_t inode_bitmap;
2861         ext4_fsblk_t inode_table;
2862         int flexbg_flag = 0;
2863         ext4_group_t i, grp = sbi->s_groups_count;
2864
2865         if (ext4_has_feature_flex_bg(sb))
2866                 flexbg_flag = 1;
2867
2868         ext4_debug("Checking group descriptors");
2869
2870         for (i = 0; i < sbi->s_groups_count; i++) {
2871                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2872
2873                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2874                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2875                 else
2876                         last_block = first_block +
2877                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2878
2879                 if ((grp == sbi->s_groups_count) &&
2880                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2881                         grp = i;
2882
2883                 block_bitmap = ext4_block_bitmap(sb, gdp);
2884                 if (block_bitmap == sb_block) {
2885                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2886                                  "Block bitmap for group %u overlaps "
2887                                  "superblock", i);
2888                         if (!sb_rdonly(sb))
2889                                 return 0;
2890                 }
2891                 if (block_bitmap >= sb_block + 1 &&
2892                     block_bitmap <= last_bg_block) {
2893                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2894                                  "Block bitmap for group %u overlaps "
2895                                  "block group descriptors", i);
2896                         if (!sb_rdonly(sb))
2897                                 return 0;
2898                 }
2899                 if (block_bitmap < first_block || block_bitmap > last_block) {
2900                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2901                                "Block bitmap for group %u not in group "
2902                                "(block %llu)!", i, block_bitmap);
2903                         return 0;
2904                 }
2905                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2906                 if (inode_bitmap == sb_block) {
2907                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2908                                  "Inode bitmap for group %u overlaps "
2909                                  "superblock", i);
2910                         if (!sb_rdonly(sb))
2911                                 return 0;
2912                 }
2913                 if (inode_bitmap >= sb_block + 1 &&
2914                     inode_bitmap <= last_bg_block) {
2915                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2916                                  "Inode bitmap for group %u overlaps "
2917                                  "block group descriptors", i);
2918                         if (!sb_rdonly(sb))
2919                                 return 0;
2920                 }
2921                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2922                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2923                                "Inode bitmap for group %u not in group "
2924                                "(block %llu)!", i, inode_bitmap);
2925                         return 0;
2926                 }
2927                 inode_table = ext4_inode_table(sb, gdp);
2928                 if (inode_table == sb_block) {
2929                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2930                                  "Inode table for group %u overlaps "
2931                                  "superblock", i);
2932                         if (!sb_rdonly(sb))
2933                                 return 0;
2934                 }
2935                 if (inode_table >= sb_block + 1 &&
2936                     inode_table <= last_bg_block) {
2937                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2938                                  "Inode table for group %u overlaps "
2939                                  "block group descriptors", i);
2940                         if (!sb_rdonly(sb))
2941                                 return 0;
2942                 }
2943                 if (inode_table < first_block ||
2944                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2945                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2946                                "Inode table for group %u not in group "
2947                                "(block %llu)!", i, inode_table);
2948                         return 0;
2949                 }
2950                 ext4_lock_group(sb, i);
2951                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2952                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2953                                  "Checksum for group %u failed (%u!=%u)",
2954                                  i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2955                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2956                         if (!sb_rdonly(sb)) {
2957                                 ext4_unlock_group(sb, i);
2958                                 return 0;
2959                         }
2960                 }
2961                 ext4_unlock_group(sb, i);
2962                 if (!flexbg_flag)
2963                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2964         }
2965         if (NULL != first_not_zeroed)
2966                 *first_not_zeroed = grp;
2967         return 1;
2968 }
2969
2970 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2971  * the superblock) which were deleted from all directories, but held open by
2972  * a process at the time of a crash.  We walk the list and try to delete these
2973  * inodes at recovery time (only with a read-write filesystem).
2974  *
2975  * In order to keep the orphan inode chain consistent during traversal (in
2976  * case of crash during recovery), we link each inode into the superblock
2977  * orphan list_head and handle it the same way as an inode deletion during
2978  * normal operation (which journals the operations for us).
2979  *
2980  * We only do an iget() and an iput() on each inode, which is very safe if we
2981  * accidentally point at an in-use or already deleted inode.  The worst that
2982  * can happen in this case is that we get a "bit already cleared" message from
2983  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2984  * e2fsck was run on this filesystem, and it must have already done the orphan
2985  * inode cleanup for us, so we can safely abort without any further action.
2986  */
2987 static void ext4_orphan_cleanup(struct super_block *sb,
2988                                 struct ext4_super_block *es)
2989 {
2990         unsigned int s_flags = sb->s_flags;
2991         int ret, nr_orphans = 0, nr_truncates = 0;
2992 #ifdef CONFIG_QUOTA
2993         int quota_update = 0;
2994         int i;
2995 #endif
2996         if (!es->s_last_orphan) {
2997                 jbd_debug(4, "no orphan inodes to clean up\n");
2998                 return;
2999         }
3000
3001         if (bdev_read_only(sb->s_bdev)) {
3002                 ext4_msg(sb, KERN_ERR, "write access "
3003                         "unavailable, skipping orphan cleanup");
3004                 return;
3005         }
3006
3007         /* Check if feature set would not allow a r/w mount */
3008         if (!ext4_feature_set_ok(sb, 0)) {
3009                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3010                          "unknown ROCOMPAT features");
3011                 return;
3012         }
3013
3014         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3015                 /* don't clear list on RO mount w/ errors */
3016                 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
3017                         ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
3018                                   "clearing orphan list.\n");
3019                         es->s_last_orphan = 0;
3020                 }
3021                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3022                 return;
3023         }
3024
3025         if (s_flags & SB_RDONLY) {
3026                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
3027                 sb->s_flags &= ~SB_RDONLY;
3028         }
3029 #ifdef CONFIG_QUOTA
3030         /*
3031          * Turn on quotas which were not enabled for read-only mounts if
3032          * filesystem has quota feature, so that they are updated correctly.
3033          */
3034         if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3035                 int ret = ext4_enable_quotas(sb);
3036
3037                 if (!ret)
3038                         quota_update = 1;
3039                 else
3040                         ext4_msg(sb, KERN_ERR,
3041                                 "Cannot turn on quotas: error %d", ret);
3042         }
3043
3044         /* Turn on journaled quotas used for old sytle */
3045         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3046                 if (EXT4_SB(sb)->s_qf_names[i]) {
3047                         int ret = ext4_quota_on_mount(sb, i);
3048
3049                         if (!ret)
3050                                 quota_update = 1;
3051                         else
3052                                 ext4_msg(sb, KERN_ERR,
3053                                         "Cannot turn on journaled "
3054                                         "quota: type %d: error %d", i, ret);
3055                 }
3056         }
3057 #endif
3058
3059         while (es->s_last_orphan) {
3060                 struct inode *inode;
3061
3062                 /*
3063                  * We may have encountered an error during cleanup; if
3064                  * so, skip the rest.
3065                  */
3066                 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3067                         jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3068                         es->s_last_orphan = 0;
3069                         break;
3070                 }
3071
3072                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3073                 if (IS_ERR(inode)) {
3074                         es->s_last_orphan = 0;
3075                         break;
3076                 }
3077
3078                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3079                 dquot_initialize(inode);
3080                 if (inode->i_nlink) {
3081                         if (test_opt(sb, DEBUG))
3082                                 ext4_msg(sb, KERN_DEBUG,
3083                                         "%s: truncating inode %lu to %lld bytes",
3084                                         __func__, inode->i_ino, inode->i_size);
3085                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3086                                   inode->i_ino, inode->i_size);
3087                         inode_lock(inode);
3088                         truncate_inode_pages(inode->i_mapping, inode->i_size);
3089                         ret = ext4_truncate(inode);
3090                         if (ret) {
3091                                 /*
3092                                  * We need to clean up the in-core orphan list
3093                                  * manually if ext4_truncate() failed to get a
3094                                  * transaction handle.
3095                                  */
3096                                 ext4_orphan_del(NULL, inode);
3097                                 ext4_std_error(inode->i_sb, ret);
3098                         }
3099                         inode_unlock(inode);
3100                         nr_truncates++;
3101                 } else {
3102                         if (test_opt(sb, DEBUG))
3103                                 ext4_msg(sb, KERN_DEBUG,
3104                                         "%s: deleting unreferenced inode %lu",
3105                                         __func__, inode->i_ino);
3106                         jbd_debug(2, "deleting unreferenced inode %lu\n",
3107                                   inode->i_ino);
3108                         nr_orphans++;
3109                 }
3110                 iput(inode);  /* The delete magic happens here! */
3111         }
3112
3113 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3114
3115         if (nr_orphans)
3116                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3117                        PLURAL(nr_orphans));
3118         if (nr_truncates)
3119                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3120                        PLURAL(nr_truncates));
3121 #ifdef CONFIG_QUOTA
3122         /* Turn off quotas if they were enabled for orphan cleanup */
3123         if (quota_update) {
3124                 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3125                         if (sb_dqopt(sb)->files[i])
3126                                 dquot_quota_off(sb, i);
3127                 }
3128         }
3129 #endif
3130         sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3131 }
3132
3133 /*
3134  * Maximal extent format file size.
3135  * Resulting logical blkno at s_maxbytes must fit in our on-disk
3136  * extent format containers, within a sector_t, and within i_blocks
3137  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
3138  * so that won't be a limiting factor.
3139  *
3140  * However there is other limiting factor. We do store extents in the form
3141  * of starting block and length, hence the resulting length of the extent
3142  * covering maximum file size must fit into on-disk format containers as
3143  * well. Given that length is always by 1 unit bigger than max unit (because
3144  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3145  *
3146  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3147  */
3148 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3149 {
3150         loff_t res;
3151         loff_t upper_limit = MAX_LFS_FILESIZE;
3152
3153         BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3154
3155         if (!has_huge_files) {
3156                 upper_limit = (1LL << 32) - 1;
3157
3158                 /* total blocks in file system block size */
3159                 upper_limit >>= (blkbits - 9);
3160                 upper_limit <<= blkbits;
3161         }
3162
3163         /*
3164          * 32-bit extent-start container, ee_block. We lower the maxbytes
3165          * by one fs block, so ee_len can cover the extent of maximum file
3166          * size
3167          */
3168         res = (1LL << 32) - 1;
3169         res <<= blkbits;
3170
3171         /* Sanity check against vm- & vfs- imposed limits */
3172         if (res > upper_limit)
3173                 res = upper_limit;
3174
3175         return res;
3176 }
3177
3178 /*
3179  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
3180  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3181  * We need to be 1 filesystem block less than the 2^48 sector limit.
3182  */
3183 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3184 {
3185         loff_t res = EXT4_NDIR_BLOCKS;
3186         int meta_blocks;
3187         loff_t upper_limit;
3188         /* This is calculated to be the largest file size for a dense, block
3189          * mapped file such that the file's total number of 512-byte sectors,
3190          * including data and all indirect blocks, does not exceed (2^48 - 1).
3191          *
3192          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3193          * number of 512-byte sectors of the file.
3194          */
3195
3196         if (!has_huge_files) {
3197                 /*
3198                  * !has_huge_files or implies that the inode i_block field
3199                  * represents total file blocks in 2^32 512-byte sectors ==
3200                  * size of vfs inode i_blocks * 8
3201                  */
3202                 upper_limit = (1LL << 32) - 1;
3203
3204                 /* total blocks in file system block size */
3205                 upper_limit >>= (bits - 9);
3206
3207         } else {
3208                 /*
3209                  * We use 48 bit ext4_inode i_blocks
3210                  * With EXT4_HUGE_FILE_FL set the i_blocks
3211                  * represent total number of blocks in
3212                  * file system block size
3213                  */
3214                 upper_limit = (1LL << 48) - 1;
3215
3216         }
3217
3218         /* indirect blocks */
3219         meta_blocks = 1;
3220         /* double indirect blocks */
3221         meta_blocks += 1 + (1LL << (bits-2));
3222         /* tripple indirect blocks */
3223         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3224
3225         upper_limit -= meta_blocks;
3226         upper_limit <<= bits;
3227
3228         res += 1LL << (bits-2);
3229         res += 1LL << (2*(bits-2));
3230         res += 1LL << (3*(bits-2));
3231         res <<= bits;
3232         if (res > upper_limit)
3233                 res = upper_limit;
3234
3235         if (res > MAX_LFS_FILESIZE)
3236                 res = MAX_LFS_FILESIZE;
3237
3238         return res;
3239 }
3240
3241 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3242                                    ext4_fsblk_t logical_sb_block, int nr)
3243 {
3244         struct ext4_sb_info *sbi = EXT4_SB(sb);
3245         ext4_group_t bg, first_meta_bg;
3246         int has_super = 0;
3247
3248         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3249
3250         if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3251                 return logical_sb_block + nr + 1;
3252         bg = sbi->s_desc_per_block * nr;
3253         if (ext4_bg_has_super(sb, bg))
3254                 has_super = 1;
3255
3256         /*
3257          * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3258          * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
3259          * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3260          * compensate.
3261          */
3262         if (sb->s_blocksize == 1024 && nr == 0 &&
3263             le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3264                 has_super++;
3265
3266         return (has_super + ext4_group_first_block_no(sb, bg));
3267 }
3268
3269 /**
3270  * ext4_get_stripe_size: Get the stripe size.
3271  * @sbi: In memory super block info
3272  *
3273  * If w