fs/ext2/super.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/fs/ext2/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/slab.h>
  24 #include <linux/init.h>
  25 #include <linux/blkdev.h>
  26 #include <linux/parser.h>
  27 #include <linux/random.h>
  28 #include <linux/buffer_head.h>
  29 #include <linux/exportfs.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <linux/quotaops.h>
  35 #include <linux/uaccess.h>
  36 #include <linux/dax.h>
  37 #include <linux/iversion.h>
  38 #include "ext2.h"
  39 #include "xattr.h"
  40 #include "acl.h"
  41
  42 static void ext2_write_super(struct super_block *sb);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45 static int ext2_sync_fs(struct super_block *sb, int wait);
  46 static int ext2_freeze(struct super_block *sb);
  47 static int ext2_unfreeze(struct super_block *sb);
  48
  49 void ext2_error(struct super_block *sb, const char *function,
  50                 const char *fmt, ...)
  51 {
  52         struct va_format vaf;
  53         va_list args;
  54         struct ext2_sb_info *sbi = EXT2_SB(sb);
  55         struct ext2_super_block *es = sbi->s_es;
  56
  57         if (!sb_rdonly(sb)) {
  58                 spin_lock(&sbi->s_lock);
  59                 sbi->s_mount_state |= EXT2_ERROR_FS;
  60                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  61                 spin_unlock(&sbi->s_lock);
  62                 ext2_sync_super(sb, es, 1);
  63         }
  64
  65         va_start(args, fmt);
  66
  67         vaf.fmt = fmt;
  68         vaf.va = &args;
  69
  70         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  71                sb->s_id, function, &vaf);
  72
  73         va_end(args);
  74
  75         if (test_opt(sb, ERRORS_PANIC))
  76                 panic("EXT2-fs: panic from previous error\n");
  77         if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
  78                 ext2_msg(sb, KERN_CRIT,
  79                              "error: remounting filesystem read-only");
  80                 sb->s_flags |= SB_RDONLY;
  81         }
  82 }
  83
  84 void ext2_msg(struct super_block *sb, const char *prefix,
  85                 const char *fmt, ...)
  86 {
  87         struct va_format vaf;
  88         va_list args;
  89
  90         va_start(args, fmt);
  91
  92         vaf.fmt = fmt;
  93         vaf.va = &args;
  94
  95         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  96
  97         va_end(args);
  98 }
  99
 100 /*
 101  * This must be called with sbi->s_lock held.
 102  */
 103 void ext2_update_dynamic_rev(struct super_block *sb)
 104 {
 105         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 106
 107         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 108                 return;
 109
 110         ext2_msg(sb, KERN_WARNING,
 111                      "warning: updating to rev %d because of "
 112                      "new feature flag, running e2fsck is recommended",
 113                      EXT2_DYNAMIC_REV);
 114
 115         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 116         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 117         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 118         /* leave es->s_feature_*compat flags alone */
 119         /* es->s_uuid will be set by e2fsck if empty */
 120
 121         /*
 122          * The rest of the superblock fields should be zero, and if not it
 123          * means they are likely already in use, so leave them alone.  We
 124          * can leave it up to e2fsck to clean up any inconsistencies there.
 125          */
 126 }
 127
 128 #ifdef CONFIG_QUOTA
 129 static int ext2_quota_off(struct super_block *sb, int type);
 130
 131 static void ext2_quota_off_umount(struct super_block *sb)
 132 {
 133         int type;
 134
 135         for (type = 0; type < MAXQUOTAS; type++)
 136                 ext2_quota_off(sb, type);
 137 }
 138 #else
 139 static inline void ext2_quota_off_umount(struct super_block *sb)
 140 {
 141 }
 142 #endif
 143
 144 static void ext2_put_super (struct super_block * sb)
 145 {
 146         int db_count;
 147         int i;
 148         struct ext2_sb_info *sbi = EXT2_SB(sb);
 149
 150         ext2_quota_off_umount(sb);
 151
 152         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
 153         sbi->s_ea_block_cache = NULL;
 154
 155         if (!sb_rdonly(sb)) {
 156                 struct ext2_super_block *es = sbi->s_es;
 157
 158                 spin_lock(&sbi->s_lock);
 159                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 160                 spin_unlock(&sbi->s_lock);
 161                 ext2_sync_super(sb, es, 1);
 162         }
 163         db_count = sbi->s_gdb_count;
 164         for (i = 0; i < db_count; i++)
 165                 brelse(sbi->s_group_desc[i]);
 166         kfree(sbi->s_group_desc);
 167         kfree(sbi->s_debts);
 168         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 169         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 170         percpu_counter_destroy(&sbi->s_dirs_counter);
 171         brelse (sbi->s_sbh);
 172         sb->s_fs_info = NULL;
 173         kfree(sbi->s_blockgroup_lock);
 174         fs_put_dax(sbi->s_daxdev);
 175         kfree(sbi);
 176 }
 177
 178 static struct kmem_cache * ext2_inode_cachep;
 179
 180 static struct inode *ext2_alloc_inode(struct super_block *sb)
 181 {
 182         struct ext2_inode_info *ei;
 183         ei = alloc_inode_sb(sb, ext2_inode_cachep, GFP_KERNEL);
 184         if (!ei)
 185                 return NULL;
 186         ei->i_block_alloc_info = NULL;
 187         inode_set_iversion(&ei->vfs_inode, 1);
 188 #ifdef CONFIG_QUOTA
 189         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 190 #endif
 191
 192         return &ei->vfs_inode;
 193 }
 194
 195 static void ext2_free_in_core_inode(struct inode *inode)
 196 {
 197         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 198 }
 199
 200 static void init_once(void *foo)
 201 {
 202         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 203
 204         rwlock_init(&ei->i_meta_lock);
 205 #ifdef CONFIG_EXT2_FS_XATTR
 206         init_rwsem(&ei->xattr_sem);
 207 #endif
 208         mutex_init(&ei->truncate_mutex);
 209         inode_init_once(&ei->vfs_inode);
 210 }
 211
 212 static int __init init_inodecache(void)
 213 {
 214         ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
 215                                 sizeof(struct ext2_inode_info), 0,
 216                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
 217                                         SLAB_ACCOUNT),
 218                                 offsetof(struct ext2_inode_info, i_data),
 219                                 sizeof_field(struct ext2_inode_info, i_data),
 220                                 init_once);
 221         if (ext2_inode_cachep == NULL)
 222                 return -ENOMEM;
 223         return 0;
 224 }
 225
 226 static void destroy_inodecache(void)
 227 {
 228         /*
 229          * Make sure all delayed rcu free inodes are flushed before we
 230          * destroy cache.
 231          */
 232         rcu_barrier();
 233         kmem_cache_destroy(ext2_inode_cachep);
 234 }
 235
 236 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 237 {
 238         struct super_block *sb = root->d_sb;
 239         struct ext2_sb_info *sbi = EXT2_SB(sb);
 240         struct ext2_super_block *es = sbi->s_es;
 241         unsigned long def_mount_opts;
 242
 243         spin_lock(&sbi->s_lock);
 244         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 245
 246         if (sbi->s_sb_block != 1)
 247                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 248         if (test_opt(sb, MINIX_DF))
 249                 seq_puts(seq, ",minixdf");
 250         if (test_opt(sb, GRPID))
 251                 seq_puts(seq, ",grpid");
 252         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 253                 seq_puts(seq, ",nogrpid");
 254         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 255             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 256                 seq_printf(seq, ",resuid=%u",
 257                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 258         }
 259         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 260             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 261                 seq_printf(seq, ",resgid=%u",
 262                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 263         }
 264         if (test_opt(sb, ERRORS_RO)) {
 265                 int def_errors = le16_to_cpu(es->s_errors);
 266
 267                 if (def_errors == EXT2_ERRORS_PANIC ||
 268                     def_errors == EXT2_ERRORS_CONTINUE) {
 269                         seq_puts(seq, ",errors=remount-ro");
 270                 }
 271         }
 272         if (test_opt(sb, ERRORS_CONT))
 273                 seq_puts(seq, ",errors=continue");
 274         if (test_opt(sb, ERRORS_PANIC))
 275                 seq_puts(seq, ",errors=panic");
 276         if (test_opt(sb, NO_UID32))
 277                 seq_puts(seq, ",nouid32");
 278         if (test_opt(sb, DEBUG))
 279                 seq_puts(seq, ",debug");
 280         if (test_opt(sb, OLDALLOC))
 281                 seq_puts(seq, ",oldalloc");
 282
 283 #ifdef CONFIG_EXT2_FS_XATTR
 284         if (test_opt(sb, XATTR_USER))
 285                 seq_puts(seq, ",user_xattr");
 286         if (!test_opt(sb, XATTR_USER) &&
 287             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 288                 seq_puts(seq, ",nouser_xattr");
 289         }
 290 #endif
 291
 292 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 293         if (test_opt(sb, POSIX_ACL))
 294                 seq_puts(seq, ",acl");
 295         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 296                 seq_puts(seq, ",noacl");
 297 #endif
 298
 299         if (test_opt(sb, NOBH))
 300                 seq_puts(seq, ",nobh");
 301
 302         if (test_opt(sb, USRQUOTA))
 303                 seq_puts(seq, ",usrquota");
 304
 305         if (test_opt(sb, GRPQUOTA))
 306                 seq_puts(seq, ",grpquota");
 307
 308         if (test_opt(sb, XIP))
 309                 seq_puts(seq, ",xip");
 310
 311         if (test_opt(sb, DAX))
 312                 seq_puts(seq, ",dax");
 313
 314         if (!test_opt(sb, RESERVATION))
 315                 seq_puts(seq, ",noreservation");
 316
 317         spin_unlock(&sbi->s_lock);
 318         return 0;
 319 }
 320
 321 #ifdef CONFIG_QUOTA
 322 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 323 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 324 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
 325                          const struct path *path);
 326 static struct dquot **ext2_get_dquots(struct inode *inode)
 327 {
 328         return EXT2_I(inode)->i_dquot;
 329 }
 330
 331 static const struct quotactl_ops ext2_quotactl_ops = {
 332         .quota_on       = ext2_quota_on,
 333         .quota_off      = ext2_quota_off,
 334         .quota_sync     = dquot_quota_sync,
 335         .get_state      = dquot_get_state,
 336         .set_info       = dquot_set_dqinfo,
 337         .get_dqblk      = dquot_get_dqblk,
 338         .set_dqblk      = dquot_set_dqblk,
 339         .get_nextdqblk  = dquot_get_next_dqblk,
 340 };
 341 #endif
 342
 343 static const struct super_operations ext2_sops = {
 344         .alloc_inode    = ext2_alloc_inode,
 345         .free_inode     = ext2_free_in_core_inode,
 346         .write_inode    = ext2_write_inode,
 347         .evict_inode    = ext2_evict_inode,
 348         .put_super      = ext2_put_super,
 349         .sync_fs        = ext2_sync_fs,
 350         .freeze_fs      = ext2_freeze,
 351         .unfreeze_fs    = ext2_unfreeze,
 352         .statfs         = ext2_statfs,
 353         .remount_fs     = ext2_remount,
 354         .show_options   = ext2_show_options,
 355 #ifdef CONFIG_QUOTA
 356         .quota_read     = ext2_quota_read,
 357         .quota_write    = ext2_quota_write,
 358         .get_dquots     = ext2_get_dquots,
 359 #endif
 360 };
 361
 362 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 363                 u64 ino, u32 generation)
 364 {
 365         struct inode *inode;
 366
 367         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 368                 return ERR_PTR(-ESTALE);
 369         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 370                 return ERR_PTR(-ESTALE);
 371
 372         /*
 373          * ext2_iget isn't quite right if the inode is currently unallocated!
 374          * However ext2_iget currently does appropriate checks to handle stale
 375          * inodes so everything is OK.
 376          */
 377         inode = ext2_iget(sb, ino);
 378         if (IS_ERR(inode))
 379                 return ERR_CAST(inode);
 380         if (generation && inode->i_generation != generation) {
 381                 /* we didn't find the right inode.. */
 382                 iput(inode);
 383                 return ERR_PTR(-ESTALE);
 384         }
 385         return inode;
 386 }
 387
 388 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 389                 int fh_len, int fh_type)
 390 {
 391         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 392                                     ext2_nfs_get_inode);
 393 }
 394
 395 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 396                 int fh_len, int fh_type)
 397 {
 398         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 399                                     ext2_nfs_get_inode);
 400 }
 401
 402 static const struct export_operations ext2_export_ops = {
 403         .fh_to_dentry = ext2_fh_to_dentry,
 404         .fh_to_parent = ext2_fh_to_parent,
 405         .get_parent = ext2_get_parent,
 406 };
 407
 408 static unsigned long get_sb_block(void **data)
 409 {
 410         unsigned long   sb_block;
 411         char            *options = (char *) *data;
 412
 413         if (!options || strncmp(options, "sb=", 3) != 0)
 414                 return 1;       /* Default location */
 415         options += 3;
 416         sb_block = simple_strtoul(options, &options, 0);
 417         if (*options && *options != ',') {
 418                 printk("EXT2-fs: Invalid sb specification: %s\n",
 419                        (char *) *data);
 420                 return 1;
 421         }
 422         if (*options == ',')
 423                 options++;
 424         *data = (void *) options;
 425         return sb_block;
 426 }
 427
 428 enum {
 429         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 430         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 431         Opt_err_ro, Opt_nouid32, Opt_debug,
 432         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 433         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 434         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 435 };
 436
 437 static const match_table_t tokens = {
 438         {Opt_bsd_df, "bsddf"},
 439         {Opt_minix_df, "minixdf"},
 440         {Opt_grpid, "grpid"},
 441         {Opt_grpid, "bsdgroups"},
 442         {Opt_nogrpid, "nogrpid"},
 443         {Opt_nogrpid, "sysvgroups"},
 444         {Opt_resgid, "resgid=%u"},
 445         {Opt_resuid, "resuid=%u"},
 446         {Opt_sb, "sb=%u"},
 447         {Opt_err_cont, "errors=continue"},
 448         {Opt_err_panic, "errors=panic"},
 449         {Opt_err_ro, "errors=remount-ro"},
 450         {Opt_nouid32, "nouid32"},
 451         {Opt_debug, "debug"},
 452         {Opt_oldalloc, "oldalloc"},
 453         {Opt_orlov, "orlov"},
 454         {Opt_nobh, "nobh"},
 455         {Opt_user_xattr, "user_xattr"},
 456         {Opt_nouser_xattr, "nouser_xattr"},
 457         {Opt_acl, "acl"},
 458         {Opt_noacl, "noacl"},
 459         {Opt_xip, "xip"},
 460         {Opt_dax, "dax"},
 461         {Opt_grpquota, "grpquota"},
 462         {Opt_ignore, "noquota"},
 463         {Opt_quota, "quota"},
 464         {Opt_usrquota, "usrquota"},
 465         {Opt_reservation, "reservation"},
 466         {Opt_noreservation, "noreservation"},
 467         {Opt_err, NULL}
 468 };
 469
 470 static int parse_options(char *options, struct super_block *sb,
 471                          struct ext2_mount_options *opts)
 472 {
 473         char *p;
 474         substring_t args[MAX_OPT_ARGS];
 475         int option;
 476         kuid_t uid;
 477         kgid_t gid;
 478
 479         if (!options)
 480                 return 1;
 481
 482         while ((p = strsep (&options, ",")) != NULL) {
 483                 int token;
 484                 if (!*p)
 485                         continue;
 486
 487                 token = match_token(p, tokens, args);
 488                 switch (token) {
 489                 case Opt_bsd_df:
 490                         clear_opt (opts->s_mount_opt, MINIX_DF);
 491                         break;
 492                 case Opt_minix_df:
 493                         set_opt (opts->s_mount_opt, MINIX_DF);
 494                         break;
 495                 case Opt_grpid:
 496                         set_opt (opts->s_mount_opt, GRPID);
 497                         break;
 498                 case Opt_nogrpid:
 499                         clear_opt (opts->s_mount_opt, GRPID);
 500                         break;
 501                 case Opt_resuid:
 502                         if (match_int(&args[0], &option))
 503                                 return 0;
 504                         uid = make_kuid(current_user_ns(), option);
 505                         if (!uid_valid(uid)) {
 506                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 507                                 return 0;
 508
 509                         }
 510                         opts->s_resuid = uid;
 511                         break;
 512                 case Opt_resgid:
 513                         if (match_int(&args[0], &option))
 514                                 return 0;
 515                         gid = make_kgid(current_user_ns(), option);
 516                         if (!gid_valid(gid)) {
 517                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 518                                 return 0;
 519                         }
 520                         opts->s_resgid = gid;
 521                         break;
 522                 case Opt_sb:
 523                         /* handled by get_sb_block() instead of here */
 524                         /* *sb_block = match_int(&args[0]); */
 525                         break;
 526                 case Opt_err_panic:
 527                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 528                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 529                         set_opt (opts->s_mount_opt, ERRORS_PANIC);
 530                         break;
 531                 case Opt_err_ro:
 532                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 533                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 534                         set_opt (opts->s_mount_opt, ERRORS_RO);
 535                         break;
 536                 case Opt_err_cont:
 537                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 538                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 539                         set_opt (opts->s_mount_opt, ERRORS_CONT);
 540                         break;
 541                 case Opt_nouid32:
 542                         set_opt (opts->s_mount_opt, NO_UID32);
 543                         break;
 544                 case Opt_debug:
 545                         set_opt (opts->s_mount_opt, DEBUG);
 546                         break;
 547                 case Opt_oldalloc:
 548                         set_opt (opts->s_mount_opt, OLDALLOC);
 549                         break;
 550                 case Opt_orlov:
 551                         clear_opt (opts->s_mount_opt, OLDALLOC);
 552                         break;
 553                 case Opt_nobh:
 554                         set_opt (opts->s_mount_opt, NOBH);
 555                         break;
 556 #ifdef CONFIG_EXT2_FS_XATTR
 557                 case Opt_user_xattr:
 558                         set_opt (opts->s_mount_opt, XATTR_USER);
 559                         break;
 560                 case Opt_nouser_xattr:
 561                         clear_opt (opts->s_mount_opt, XATTR_USER);
 562                         break;
 563 #else
 564                 case Opt_user_xattr:
 565                 case Opt_nouser_xattr:
 566                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 567                                 "not supported");
 568                         break;
 569 #endif
 570 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 571                 case Opt_acl:
 572                         set_opt(opts->s_mount_opt, POSIX_ACL);
 573                         break;
 574                 case Opt_noacl:
 575                         clear_opt(opts->s_mount_opt, POSIX_ACL);
 576                         break;
 577 #else
 578                 case Opt_acl:
 579                 case Opt_noacl:
 580                         ext2_msg(sb, KERN_INFO,
 581                                 "(no)acl options not supported");
 582                         break;
 583 #endif
 584                 case Opt_xip:
 585                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 586                         set_opt(opts->s_mount_opt, XIP);
 587                         fallthrough;
 588                 case Opt_dax:
 589 #ifdef CONFIG_FS_DAX
 590                         ext2_msg(sb, KERN_WARNING,
 591                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 592                         set_opt(opts->s_mount_opt, DAX);
 593 #else
 594                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 595 #endif
 596                         break;
 597
 598 #if defined(CONFIG_QUOTA)
 599                 case Opt_quota:
 600                 case Opt_usrquota:
 601                         set_opt(opts->s_mount_opt, USRQUOTA);
 602                         break;
 603
 604                 case Opt_grpquota:
 605                         set_opt(opts->s_mount_opt, GRPQUOTA);
 606                         break;
 607 #else
 608                 case Opt_quota:
 609                 case Opt_usrquota:
 610                 case Opt_grpquota:
 611                         ext2_msg(sb, KERN_INFO,
 612                                 "quota operations not supported");
 613                         break;
 614 #endif
 615
 616                 case Opt_reservation:
 617                         set_opt(opts->s_mount_opt, RESERVATION);
 618                         ext2_msg(sb, KERN_INFO, "reservations ON");
 619                         break;
 620                 case Opt_noreservation:
 621                         clear_opt(opts->s_mount_opt, RESERVATION);
 622                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 623                         break;
 624                 case Opt_ignore:
 625                         break;
 626                 default:
 627                         return 0;
 628                 }
 629         }
 630         return 1;
 631 }
 632
 633 static int ext2_setup_super (struct super_block * sb,
 634                               struct ext2_super_block * es,
 635                               int read_only)
 636 {
 637         int res = 0;
 638         struct ext2_sb_info *sbi = EXT2_SB(sb);
 639
 640         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 641                 ext2_msg(sb, KERN_ERR,
 642                         "error: revision level too high, "
 643                         "forcing read-only mode");
 644                 res = SB_RDONLY;
 645         }
 646         if (read_only)
 647                 return res;
 648         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 649                 ext2_msg(sb, KERN_WARNING,
 650                         "warning: mounting unchecked fs, "
 651                         "running e2fsck is recommended");
 652         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 653                 ext2_msg(sb, KERN_WARNING,
 654                         "warning: mounting fs with errors, "
 655                         "running e2fsck is recommended");
 656         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 657                  le16_to_cpu(es->s_mnt_count) >=
 658                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 659                 ext2_msg(sb, KERN_WARNING,
 660                         "warning: maximal mount count reached, "
 661                         "running e2fsck is recommended");
 662         else if (le32_to_cpu(es->s_checkinterval) &&
 663                 (le32_to_cpu(es->s_lastcheck) +
 664                         le32_to_cpu(es->s_checkinterval) <=
 665                         ktime_get_real_seconds()))
 666                 ext2_msg(sb, KERN_WARNING,
 667                         "warning: checktime reached, "
 668                         "running e2fsck is recommended");
 669         if (!le16_to_cpu(es->s_max_mnt_count))
 670                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 671         le16_add_cpu(&es->s_mnt_count, 1);
 672         if (test_opt (sb, DEBUG))
 673                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 674                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 675                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 676                         sbi->s_frag_size,
 677                         sbi->s_groups_count,
 678                         EXT2_BLOCKS_PER_GROUP(sb),
 679                         EXT2_INODES_PER_GROUP(sb),
 680                         sbi->s_mount_opt);
 681         return res;
 682 }
 683
 684 static int ext2_check_descriptors(struct super_block *sb)
 685 {
 686         int i;
 687         struct ext2_sb_info *sbi = EXT2_SB(sb);
 688
 689         ext2_debug ("Checking group descriptors");
 690
 691         for (i = 0; i < sbi->s_groups_count; i++) {
 692                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 693                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 694                 ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
 695
 696                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 697                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 698                 {
 699                         ext2_error (sb, "ext2_check_descriptors",
 700                                     "Block bitmap for group %d"
 701                                     " not in group (block %lu)!",
 702                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 703                         return 0;
 704                 }
 705                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 706                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 707                 {
 708                         ext2_error (sb, "ext2_check_descriptors",
 709                                     "Inode bitmap for group %d"
 710                                     " not in group (block %lu)!",
 711                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 712                         return 0;
 713                 }
 714                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 715                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 716                     last_block)
 717                 {
 718                         ext2_error (sb, "ext2_check_descriptors",
 719                                     "Inode table for group %d"
 720                                     " not in group (block %lu)!",
 721                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 722                         return 0;
 723                 }
 724         }
 725         return 1;
 726 }
 727
 728 /*
 729  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 730  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 731  * We need to be 1 filesystem block less than the 2^32 sector limit.
 732  */
 733 static loff_t ext2_max_size(int bits)
 734 {
 735         loff_t res = EXT2_NDIR_BLOCKS;
 736         int meta_blocks;
 737         unsigned int upper_limit;
 738         unsigned int ppb = 1 << (bits-2);
 739
 740         /* This is calculated to be the largest file size for a
 741          * dense, file such that the total number of
 742          * sectors in the file, including data and all indirect blocks,
 743          * does not exceed 2^32 -1
 744          * __u32 i_blocks representing the total number of
 745          * 512 bytes blocks of the file
 746          */
 747         upper_limit = (1LL << 32) - 1;
 748
 749         /* total blocks in file system block size */
 750         upper_limit >>= (bits - 9);
 751
 752         /* Compute how many blocks we can address by block tree */
 753         res += 1LL << (bits-2);
 754         res += 1LL << (2*(bits-2));
 755         res += 1LL << (3*(bits-2));
 756         /* Compute how many metadata blocks are needed */
 757         meta_blocks = 1;
 758         meta_blocks += 1 + ppb;
 759         meta_blocks += 1 + ppb + ppb * ppb;
 760         /* Does block tree limit file size? */
 761         if (res + meta_blocks <= upper_limit)
 762                 goto check_lfs;
 763
 764         res = upper_limit;
 765         /* How many metadata blocks are needed for addressing upper_limit? */
 766         upper_limit -= EXT2_NDIR_BLOCKS;
 767         /* indirect blocks */
 768         meta_blocks = 1;
 769         upper_limit -= ppb;
 770         /* double indirect blocks */
 771         if (upper_limit < ppb * ppb) {
 772                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 773                 res -= meta_blocks;
 774                 goto check_lfs;
 775         }
 776         meta_blocks += 1 + ppb;
 777         upper_limit -= ppb * ppb;
 778         /* tripple indirect blocks for the rest */
 779         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 780                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 781         res -= meta_blocks;
 782 check_lfs:
 783         res <<= bits;
 784         if (res > MAX_LFS_FILESIZE)
 785                 res = MAX_LFS_FILESIZE;
 786
 787         return res;
 788 }
 789
 790 static unsigned long descriptor_loc(struct super_block *sb,
 791                                     unsigned long logic_sb_block,
 792                                     int nr)
 793 {
 794         struct ext2_sb_info *sbi = EXT2_SB(sb);
 795         unsigned long bg, first_meta_bg;
 796
 797         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 798
 799         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 800             nr < first_meta_bg)
 801                 return (logic_sb_block + nr + 1);
 802         bg = sbi->s_desc_per_block * nr;
 803
 804         return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
 805 }
 806
 807 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 808 {
 809         struct buffer_head * bh;
 810         struct ext2_sb_info * sbi;
 811         struct ext2_super_block * es;
 812         struct inode *root;
 813         unsigned long block;
 814         unsigned long sb_block = get_sb_block(&data);
 815         unsigned long logic_sb_block;
 816         unsigned long offset = 0;
 817         unsigned long def_mount_opts;
 818         long ret = -ENOMEM;
 819         int blocksize = BLOCK_SIZE;
 820         int db_count;
 821         int i, j;
 822         __le32 features;
 823         int err;
 824         struct ext2_mount_options opts;
 825
 826         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 827         if (!sbi)
 828                 return -ENOMEM;
 829
 830         sbi->s_blockgroup_lock =
 831                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 832         if (!sbi->s_blockgroup_lock) {
 833                 kfree(sbi);
 834                 return -ENOMEM;
 835         }
 836         sb->s_fs_info = sbi;
 837         sbi->s_sb_block = sb_block;
 838         sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off);
 839
 840         spin_lock_init(&sbi->s_lock);
 841         ret = -EINVAL;
 842
 843         /*
 844          * See what the current blocksize for the device is, and
 845          * use that as the blocksize.  Otherwise (or if the blocksize
 846          * is smaller than the default) use the default.
 847          * This is important for devices that have a hardware
 848          * sectorsize that is larger than the default.
 849          */
 850         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 851         if (!blocksize) {
 852                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 853                 goto failed_sbi;
 854         }
 855
 856         /*
 857          * If the superblock doesn't start on a hardware sector boundary,
 858          * calculate the offset.
 859          */
 860         if (blocksize != BLOCK_SIZE) {
 861                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 862                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 863         } else {
 864                 logic_sb_block = sb_block;
 865         }
 866
 867         if (!(bh = sb_bread(sb, logic_sb_block))) {
 868                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 869                 goto failed_sbi;
 870         }
 871         /*
 872          * Note: s_es must be initialized as soon as possible because
 873          *       some ext2 macro-instructions depend on its value
 874          */
 875         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 876         sbi->s_es = es;
 877         sb->s_magic = le16_to_cpu(es->s_magic);
 878
 879         if (sb->s_magic != EXT2_SUPER_MAGIC)
 880                 goto cantfind_ext2;
 881
 882         opts.s_mount_opt = 0;
 883         /* Set defaults before we parse the mount options */
 884         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 885         if (def_mount_opts & EXT2_DEFM_DEBUG)
 886                 set_opt(opts.s_mount_opt, DEBUG);
 887         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 888                 set_opt(opts.s_mount_opt, GRPID);
 889         if (def_mount_opts & EXT2_DEFM_UID16)
 890                 set_opt(opts.s_mount_opt, NO_UID32);
 891 #ifdef CONFIG_EXT2_FS_XATTR
 892         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 893                 set_opt(opts.s_mount_opt, XATTR_USER);
 894 #endif
 895 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 896         if (def_mount_opts & EXT2_DEFM_ACL)
 897                 set_opt(opts.s_mount_opt, POSIX_ACL);
 898 #endif
 899
 900         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 901                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 902         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 903                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 904         else
 905                 set_opt(opts.s_mount_opt, ERRORS_RO);
 906
 907         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 908         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 909
 910         set_opt(opts.s_mount_opt, RESERVATION);
 911
 912         if (!parse_options((char *) data, sb, &opts))
 913                 goto failed_mount;
 914
 915         sbi->s_mount_opt = opts.s_mount_opt;
 916         sbi->s_resuid = opts.s_resuid;
 917         sbi->s_resgid = opts.s_resgid;
 918
 919         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 920                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
 921         sb->s_iflags |= SB_I_CGROUPWB;
 922
 923         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 924             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 925              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 926              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 927                 ext2_msg(sb, KERN_WARNING,
 928                         "warning: feature flags set on rev 0 fs, "
 929                         "running e2fsck is recommended");
 930         /*
 931          * Check feature flags regardless of the revision level, since we
 932          * previously didn't change the revision level when setting the flags,
 933          * so there is a chance incompat flags are set on a rev 0 filesystem.
 934          */
 935         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 936         if (features) {
 937                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 938                        "unsupported optional features (%x)",
 939                         le32_to_cpu(features));
 940                 goto failed_mount;
 941         }
 942         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 943                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 944                        "unsupported optional features (%x)",
 945                        le32_to_cpu(features));
 946                 goto failed_mount;
 947         }
 948
 949         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 950
 951         if (test_opt(sb, DAX)) {
 952                 if (!sbi->s_daxdev) {
 953                         ext2_msg(sb, KERN_ERR,
 954                                 "DAX unsupported by block device. Turning off DAX.");
 955                         clear_opt(sbi->s_mount_opt, DAX);
 956                 } else if (blocksize != PAGE_SIZE) {
 957                         ext2_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
 958                         clear_opt(sbi->s_mount_opt, DAX);
 959                 }
 960         }
 961
 962         /* If the blocksize doesn't match, re-read the thing.. */
 963         if (sb->s_blocksize != blocksize) {
 964                 brelse(bh);
 965
 966                 if (!sb_set_blocksize(sb, blocksize)) {
 967                         ext2_msg(sb, KERN_ERR,
 968                                 "error: bad blocksize %d", blocksize);
 969                         goto failed_sbi;
 970                 }
 971
 972                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 973                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 974                 bh = sb_bread(sb, logic_sb_block);
 975                 if(!bh) {
 976                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 977                                 "superblock on 2nd try");
 978                         goto failed_sbi;
 979                 }
 980                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 981                 sbi->s_es = es;
 982                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 983                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 984                         goto failed_mount;
 985                 }
 986         }
 987
 988         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 989         sb->s_max_links = EXT2_LINK_MAX;
 990         sb->s_time_min = S32_MIN;
 991         sb->s_time_max = S32_MAX;
 992
 993         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 994                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 995                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 996         } else {
 997                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 998                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 999                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1000                     !is_power_of_2(sbi->s_inode_size) ||
1001                     (sbi->s_inode_size > blocksize)) {
1002                         ext2_msg(sb, KERN_ERR,
1003                                 "error: unsupported inode size: %d",
1004                                 sbi->s_inode_size);
1005                         goto failed_mount;
1006                 }
1007         }
1008
1009         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1010                                    le32_to_cpu(es->s_log_frag_size);
1011         if (sbi->s_frag_size == 0)
1012                 goto cantfind_ext2;
1013         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1014
1015         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1016         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1017         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1018
1019         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1020         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1021                 goto cantfind_ext2;
1022         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1023                                         sbi->s_inodes_per_block;
1024         sbi->s_desc_per_block = sb->s_blocksize /
1025                                         sizeof (struct ext2_group_desc);
1026         sbi->s_sbh = bh;
1027         sbi->s_mount_state = le16_to_cpu(es->s_state);
1028         sbi->s_addr_per_block_bits =
1029                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1030         sbi->s_desc_per_block_bits =
1031                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1032
1033         if (sb->s_magic != EXT2_SUPER_MAGIC)
1034                 goto cantfind_ext2;
1035
1036         if (sb->s_blocksize != bh->b_size) {
1037                 if (!silent)
1038                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1039                 goto failed_mount;
1040         }
1041
1042         if (sb->s_blocksize != sbi->s_frag_size) {
1043                 ext2_msg(sb, KERN_ERR,
1044                         "error: fragsize %lu != blocksize %lu"
1045                         "(not supported yet)",
1046                         sbi->s_frag_size, sb->s_blocksize);
1047                 goto failed_mount;
1048         }
1049
1050         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1051                 ext2_msg(sb, KERN_ERR,
1052                         "error: #blocks per group too big: %lu",
1053                         sbi->s_blocks_per_group);
1054                 goto failed_mount;
1055         }
1056         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1057                 ext2_msg(sb, KERN_ERR,
1058                         "error: #fragments per group too big: %lu",
1059                         sbi->s_frags_per_group);
1060                 goto failed_mount;
1061         }
1062         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1063                 ext2_msg(sb, KERN_ERR,
1064                         "error: #inodes per group too big: %lu",
1065                         sbi->s_inodes_per_group);
1066                 goto failed_mount;
1067         }
1068
1069         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1070                 goto cantfind_ext2;
1071         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1072                                 le32_to_cpu(es->s_first_data_block) - 1)
1073                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1074         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1075                    EXT2_DESC_PER_BLOCK(sb);
1076         sbi->s_group_desc = kmalloc_array(db_count,
1077                                            sizeof(struct buffer_head *),
1078                                            GFP_KERNEL);
1079         if (sbi->s_group_desc == NULL) {
1080                 ret = -ENOMEM;
1081                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1082                 goto failed_mount;
1083         }
1084         bgl_lock_init(sbi->s_blockgroup_lock);
1085         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1086         if (!sbi->s_debts) {
1087                 ret = -ENOMEM;
1088                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1089                 goto failed_mount_group_desc;
1090         }
1091         for (i = 0; i < db_count; i++) {
1092                 block = descriptor_loc(sb, logic_sb_block, i);
1093                 sbi->s_group_desc[i] = sb_bread(sb, block);
1094                 if (!sbi->s_group_desc[i]) {
1095                         for (j = 0; j < i; j++)
1096                                 brelse (sbi->s_group_desc[j]);
1097                         ext2_msg(sb, KERN_ERR,
1098                                 "error: unable to read group descriptors");
1099                         goto failed_mount_group_desc;
1100                 }
1101         }
1102         if (!ext2_check_descriptors (sb)) {
1103                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1104                 goto failed_mount2;
1105         }
1106         sbi->s_gdb_count = db_count;
1107         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1108         spin_lock_init(&sbi->s_next_gen_lock);
1109
1110         /* per filesystem reservation list head & lock */
1111         spin_lock_init(&sbi->s_rsv_window_lock);
1112         sbi->s_rsv_window_root = RB_ROOT;
1113         /*
1114          * Add a single, static dummy reservation to the start of the
1115          * reservation window list --- it gives us a placeholder for
1116          * append-at-start-of-list which makes the allocation logic
1117          * _much_ simpler.
1118          */
1119         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1120         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1121         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1122         sbi->s_rsv_window_head.rsv_goal_size = 0;
1123         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1124
1125         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1126                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1127         if (!err) {
1128                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1129                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1130         }
1131         if (!err) {
1132                 err = percpu_counter_init(&sbi->s_dirs_counter,
1133                                 ext2_count_dirs(sb), GFP_KERNEL);
1134         }
1135         if (err) {
1136                 ret = err;
1137                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1138                 goto failed_mount3;
1139         }
1140
1141 #ifdef CONFIG_EXT2_FS_XATTR
1142         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1143         if (!sbi->s_ea_block_cache) {
1144                 ret = -ENOMEM;
1145                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1146                 goto failed_mount3;
1147         }
1148 #endif
1149         /*
1150          * set up enough so that it can read an inode
1151          */
1152         sb->s_op = &ext2_sops;
1153         sb->s_export_op = &ext2_export_ops;
1154         sb->s_xattr = ext2_xattr_handlers;
1155
1156 #ifdef CONFIG_QUOTA
1157         sb->dq_op = &dquot_operations;
1158         sb->s_qcop = &ext2_quotactl_ops;
1159         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1160 #endif
1161
1162         root = ext2_iget(sb, EXT2_ROOT_INO);
1163         if (IS_ERR(root)) {
1164                 ret = PTR_ERR(root);
1165                 goto failed_mount3;
1166         }
1167         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1168                 iput(root);
1169                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1170                 goto failed_mount3;
1171         }
1172
1173         sb->s_root = d_make_root(root);
1174         if (!sb->s_root) {
1175                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1176                 ret = -ENOMEM;
1177                 goto failed_mount3;
1178         }
1179         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1180                 ext2_msg(sb, KERN_WARNING,
1181                         "warning: mounting ext3 filesystem as ext2");
1182         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1183                 sb->s_flags |= SB_RDONLY;
1184         ext2_write_super(sb);
1185         return 0;
1186
1187 cantfind_ext2:
1188         if (!silent)
1189                 ext2_msg(sb, KERN_ERR,
1190                         "error: can't find an ext2 filesystem on dev %s.",
1191                         sb->s_id);
1192         goto failed_mount;
1193 failed_mount3:
1194         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1195         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1196         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1197         percpu_counter_destroy(&sbi->s_dirs_counter);
1198 failed_mount2:
1199         for (i = 0; i < db_count; i++)
1200                 brelse(sbi->s_group_desc[i]);
1201 failed_mount_group_desc:
1202         kfree(sbi->s_group_desc);
1203         kfree(sbi->s_debts);
1204 failed_mount:
1205         brelse(bh);
1206 failed_sbi:
1207         fs_put_dax(sbi->s_daxdev);
1208         sb->s_fs_info = NULL;
1209         kfree(sbi->s_blockgroup_lock);
1210         kfree(sbi);
1211         return ret;
1212 }
1213
1214 static void ext2_clear_super_error(struct super_block *sb)
1215 {
1216         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1217
1218         if (buffer_write_io_error(sbh)) {
1219                 /*
1220                  * Oh, dear.  A previous attempt to write the
1221                  * superblock failed.  This could happen because the
1222                  * USB device was yanked out.  Or it could happen to
1223                  * be a transient write error and maybe the block will
1224                  * be remapped.  Nothing we can do but to retry the
1225                  * write and hope for the best.
1226                  */
1227                 ext2_msg(sb, KERN_ERR,
1228                        "previous I/O error to superblock detected");
1229                 clear_buffer_write_io_error(sbh);
1230                 set_buffer_uptodate(sbh);
1231         }
1232 }
1233
1234 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1235                      int wait)
1236 {
1237         ext2_clear_super_error(sb);
1238         spin_lock(&EXT2_SB(sb)->s_lock);
1239         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1240         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1241         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1242         /* unlock before we do IO */
1243         spin_unlock(&EXT2_SB(sb)->s_lock);
1244         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1245         if (wait)
1246                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1247 }
1248
1249 /*
1250  * In the second extended file system, it is not necessary to
1251  * write the super block since we use a mapping of the
1252  * disk super block in a buffer.
1253  *
1254  * However, this function is still used to set the fs valid
1255  * flags to 0.  We need to set this flag to 0 since the fs
1256  * may have been checked while mounted and e2fsck may have
1257  * set s_state to EXT2_VALID_FS after some corrections.
1258  */
1259 static int ext2_sync_fs(struct super_block *sb, int wait)
1260 {
1261         struct ext2_sb_info *sbi = EXT2_SB(sb);
1262         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1263
1264         /*
1265          * Write quota structures to quota file, sync_blockdev() will write
1266          * them to disk later
1267          */
1268         dquot_writeback_dquots(sb, -1);
1269
1270         spin_lock(&sbi->s_lock);
1271         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1272                 ext2_debug("setting valid to 0\n");
1273                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1274         }
1275         spin_unlock(&sbi->s_lock);
1276         ext2_sync_super(sb, es, wait);
1277         return 0;
1278 }
1279
1280 static int ext2_freeze(struct super_block *sb)
1281 {
1282         struct ext2_sb_info *sbi = EXT2_SB(sb);
1283
1284         /*
1285          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1286          * because we have unattached inodes and thus filesystem is not fully
1287          * consistent.
1288          */
1289         if (atomic_long_read(&sb->s_remove_count)) {
1290                 ext2_sync_fs(sb, 1);
1291                 return 0;
1292         }
1293         /* Set EXT2_FS_VALID flag */
1294         spin_lock(&sbi->s_lock);
1295         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1296         spin_unlock(&sbi->s_lock);
1297         ext2_sync_super(sb, sbi->s_es, 1);
1298
1299         return 0;
1300 }
1301
1302 static int ext2_unfreeze(struct super_block *sb)
1303 {
1304         /* Just write sb to clear EXT2_VALID_FS flag */
1305         ext2_write_super(sb);
1306
1307         return 0;
1308 }
1309
1310 static void ext2_write_super(struct super_block *sb)
1311 {
1312         if (!sb_rdonly(sb))
1313                 ext2_sync_fs(sb, 1);
1314 }
1315
1316 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1317 {
1318         struct ext2_sb_info * sbi = EXT2_SB(sb);
1319         struct ext2_super_block * es;
1320         struct ext2_mount_options new_opts;
1321         int err;
1322
1323         sync_filesystem(sb);
1324
1325         spin_lock(&sbi->s_lock);
1326         new_opts.s_mount_opt = sbi->s_mount_opt;
1327         new_opts.s_resuid = sbi->s_resuid;
1328         new_opts.s_resgid = sbi->s_resgid;
1329         spin_unlock(&sbi->s_lock);
1330
1331         if (!parse_options(data, sb, &new_opts))
1332                 return -EINVAL;
1333
1334         spin_lock(&sbi->s_lock);
1335         es = sbi->s_es;
1336         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1337                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1338                          "dax flag with busy inodes while remounting");
1339                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1340         }
1341         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1342                 goto out_set;
1343         if (*flags & SB_RDONLY) {
1344                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1345                     !(sbi->s_mount_state & EXT2_VALID_FS))
1346                         goto out_set;
1347
1348                 /*
1349                  * OK, we are remounting a valid rw partition rdonly, so set
1350                  * the rdonly flag and then mark the partition as valid again.
1351                  */
1352                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1353                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1354                 spin_unlock(&sbi->s_lock);
1355
1356                 err = dquot_suspend(sb, -1);
1357                 if (err < 0)
1358                         return err;
1359
1360                 ext2_sync_super(sb, es, 1);
1361         } else {
1362                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1363                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1364                 if (ret) {
1365                         spin_unlock(&sbi->s_lock);
1366                         ext2_msg(sb, KERN_WARNING,
1367                                 "warning: couldn't remount RDWR because of "
1368                                 "unsupported optional features (%x).",
1369                                 le32_to_cpu(ret));
1370                         return -EROFS;
1371                 }
1372                 /*
1373                  * Mounting a RDONLY partition read-write, so reread and
1374                  * store the current valid flag.  (It may have been changed
1375                  * by e2fsck since we originally mounted the partition.)
1376                  */
1377                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1378                 if (!ext2_setup_super (sb, es, 0))
1379                         sb->s_flags &= ~SB_RDONLY;
1380                 spin_unlock(&sbi->s_lock);
1381
1382                 ext2_write_super(sb);
1383
1384                 dquot_resume(sb, -1);
1385         }
1386
1387         spin_lock(&sbi->s_lock);
1388 out_set:
1389         sbi->s_mount_opt = new_opts.s_mount_opt;
1390         sbi->s_resuid = new_opts.s_resuid;
1391         sbi->s_resgid = new_opts.s_resgid;
1392         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1393                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1394         spin_unlock(&sbi->s_lock);
1395
1396         return 0;
1397 }
1398
1399 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1400 {
1401         struct super_block *sb = dentry->d_sb;
1402         struct ext2_sb_info *sbi = EXT2_SB(sb);
1403         struct ext2_super_block *es = sbi->s_es;
1404
1405         spin_lock(&sbi->s_lock);
1406
1407         if (test_opt (sb, MINIX_DF))
1408                 sbi->s_overhead_last = 0;
1409         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1410                 unsigned long i, overhead = 0;
1411                 smp_rmb();
1412
1413                 /*
1414                  * Compute the overhead (FS structures). This is constant
1415                  * for a given filesystem unless the number of block groups
1416                  * changes so we cache the previous value until it does.
1417                  */
1418
1419                 /*
1420                  * All of the blocks before first_data_block are
1421                  * overhead
1422                  */
1423                 overhead = le32_to_cpu(es->s_first_data_block);
1424
1425                 /*
1426                  * Add the overhead attributed to the superblock and
1427                  * block group descriptors.  If the sparse superblocks
1428                  * feature is turned on, then not all groups have this.
1429                  */
1430                 for (i = 0; i < sbi->s_groups_count; i++)
1431                         overhead += ext2_bg_has_super(sb, i) +
1432                                 ext2_bg_num_gdb(sb, i);
1433
1434                 /*
1435                  * Every block group has an inode bitmap, a block
1436                  * bitmap, and an inode table.
1437                  */
1438                 overhead += (sbi->s_groups_count *
1439                              (2 + sbi->s_itb_per_group));
1440                 sbi->s_overhead_last = overhead;
1441                 smp_wmb();
1442                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1443         }
1444
1445         buf->f_type = EXT2_SUPER_MAGIC;
1446         buf->f_bsize = sb->s_blocksize;
1447         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1448         buf->f_bfree = ext2_count_free_blocks(sb);
1449         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1450         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1451         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1452                 buf->f_bavail = 0;
1453         buf->f_files = le32_to_cpu(es->s_inodes_count);
1454         buf->f_ffree = ext2_count_free_inodes(sb);
1455         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1456         buf->f_namelen = EXT2_NAME_LEN;
1457         buf->f_fsid = uuid_to_fsid(es->s_uuid);
1458         spin_unlock(&sbi->s_lock);
1459         return 0;
1460 }
1461
1462 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1463         int flags, const char *dev_name, void *data)
1464 {
1465         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1466 }
1467
1468 #ifdef CONFIG_QUOTA
1469
1470 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1471  * acquiring the locks... As quota files are never truncated and quota code
1472  * itself serializes the operations (and no one else should touch the files)
1473  * we don't have to be afraid of races */
1474 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1475                                size_t len, loff_t off)
1476 {
1477         struct inode *inode = sb_dqopt(sb)->files[type];
1478         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1479         int err = 0;
1480         int offset = off & (sb->s_blocksize - 1);
1481         int tocopy;
1482         size_t toread;
1483         struct buffer_head tmp_bh;
1484         struct buffer_head *bh;
1485         loff_t i_size = i_size_read(inode);
1486
1487         if (off > i_size)
1488                 return 0;
1489         if (off+len > i_size)
1490                 len = i_size-off;
1491         toread = len;
1492         while (toread > 0) {
1493                 tocopy = sb->s_blocksize - offset < toread ?
1494                                 sb->s_blocksize - offset : toread;
1495
1496                 tmp_bh.b_state = 0;
1497                 tmp_bh.b_size = sb->s_blocksize;
1498                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1499                 if (err < 0)
1500                         return err;
1501                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1502                         memset(data, 0, tocopy);
1503                 else {
1504                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1505                         if (!bh)
1506                                 return -EIO;
1507                         memcpy(data, bh->b_data+offset, tocopy);
1508                         brelse(bh);
1509                 }
1510                 offset = 0;
1511                 toread -= tocopy;
1512                 data += tocopy;
1513                 blk++;
1514         }
1515         return len;
1516 }
1517
1518 /* Write to quotafile */
1519 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1520                                 const char *data, size_t len, loff_t off)
1521 {
1522         struct inode *inode = sb_dqopt(sb)->files[type];
1523         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1524         int err = 0;
1525         int offset = off & (sb->s_blocksize - 1);
1526         int tocopy;
1527         size_t towrite = len;
1528         struct buffer_head tmp_bh;
1529         struct buffer_head *bh;
1530
1531         while (towrite > 0) {
1532                 tocopy = sb->s_blocksize - offset < towrite ?
1533                                 sb->s_blocksize - offset : towrite;
1534
1535                 tmp_bh.b_state = 0;
1536                 tmp_bh.b_size = sb->s_blocksize;
1537                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1538                 if (err < 0)
1539                         goto out;
1540                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1541                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1542                 else
1543                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1544                 if (unlikely(!bh)) {
1545                         err = -EIO;
1546                         goto out;
1547                 }
1548                 lock_buffer(bh);
1549                 memcpy(bh->b_data+offset, data, tocopy);
1550                 flush_dcache_page(bh->b_page);
1551                 set_buffer_uptodate(bh);
1552                 mark_buffer_dirty(bh);
1553                 unlock_buffer(bh);
1554                 brelse(bh);
1555                 offset = 0;
1556                 towrite -= tocopy;
1557                 data += tocopy;
1558                 blk++;
1559         }
1560 out:
1561         if (len == towrite)
1562                 return err;
1563         if (inode->i_size < off+len-towrite)
1564                 i_size_write(inode, off+len-towrite);
1565         inode_inc_iversion(inode);
1566         inode->i_mtime = inode->i_ctime = current_time(inode);
1567         mark_inode_dirty(inode);
1568         return len - towrite;
1569 }
1570
1571 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1572                          const struct path *path)
1573 {
1574         int err;
1575         struct inode *inode;
1576
1577         err = dquot_quota_on(sb, type, format_id, path);
1578         if (err)
1579                 return err;
1580
1581         inode = d_inode(path->dentry);
1582         inode_lock(inode);
1583         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1584         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1585                         S_NOATIME | S_IMMUTABLE);
1586         inode_unlock(inode);
1587         mark_inode_dirty(inode);
1588
1589         return 0;
1590 }
1591
1592 static int ext2_quota_off(struct super_block *sb, int type)
1593 {
1594         struct inode *inode = sb_dqopt(sb)->files[type];
1595         int err;
1596
1597         if (!inode || !igrab(inode))
1598                 goto out;
1599
1600         err = dquot_quota_off(sb, type);
1601         if (err)
1602                 goto out_put;
1603
1604         inode_lock(inode);
1605         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1606         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1607         inode_unlock(inode);
1608         mark_inode_dirty(inode);
1609 out_put:
1610         iput(inode);
1611         return err;
1612 out:
1613         return dquot_quota_off(sb, type);
1614 }
1615
1616 #endif
1617
1618 static struct file_system_type ext2_fs_type = {
1619         .owner          = THIS_MODULE,
1620         .name           = "ext2",
1621         .mount          = ext2_mount,
1622         .kill_sb        = kill_block_super,
1623         .fs_flags       = FS_REQUIRES_DEV,
1624 };
1625 MODULE_ALIAS_FS("ext2");
1626
1627 static int __init init_ext2_fs(void)
1628 {
1629         int err;
1630
1631         err = init_inodecache();
1632         if (err)
1633                 return err;
1634         err = register_filesystem(&ext2_fs_type);
1635         if (err)
1636                 goto out;
1637         return 0;
1638 out:
1639         destroy_inodecache();
1640         return err;
1641 }
1642
1643 static void __exit exit_ext2_fs(void)
1644 {
1645         unregister_filesystem(&ext2_fs_type);
1646         destroy_inodecache();
1647 }
1648
1649 MODULE_AUTHOR("Remy Card and others");
1650 MODULE_DESCRIPTION("Second Extended Filesystem");
1651 MODULE_LICENSE("GPL");
1652 module_init(init_ext2_fs)
1653 module_exit(exit_ext2_fs)