fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card ([email protected])
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller ([email protected]), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/vfs.h>
  30 #include <linux/seq_file.h>
  31 #include <linux/mount.h>
  32 #include <linux/log2.h>
  33 #include <linux/quotaops.h>
  34 #include <linux/uaccess.h>
  35 #include <linux/dax.h>
  36 #include <linux/iversion.h>
  37 #include "ext2.h"
  38 #include "xattr.h"
  39 #include "acl.h"
  40
  41 static void ext2_write_super(struct super_block *sb);
  42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  44 static int ext2_sync_fs(struct super_block *sb, int wait);
  45 static int ext2_freeze(struct super_block *sb);
  46 static int ext2_unfreeze(struct super_block *sb);
  47
  48 void ext2_error(struct super_block *sb, const char *function,
  49                 const char *fmt, ...)
  50 {
  51         struct va_format vaf;
  52         va_list args;
  53         struct ext2_sb_info *sbi = EXT2_SB(sb);
  54         struct ext2_super_block *es = sbi->s_es;
  55
  56         if (!sb_rdonly(sb)) {
  57                 spin_lock(&sbi->s_lock);
  58                 sbi->s_mount_state |= EXT2_ERROR_FS;
  59                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  60                 spin_unlock(&sbi->s_lock);
  61                 ext2_sync_super(sb, es, 1);
  62         }
  63
  64         va_start(args, fmt);
  65
  66         vaf.fmt = fmt;
  67         vaf.va = &args;
  68
  69         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  70                sb->s_id, function, &vaf);
  71
  72         va_end(args);
  73
  74         if (test_opt(sb, ERRORS_PANIC))
  75                 panic("EXT2-fs: panic from previous error\n");
  76         if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
  77                 ext2_msg(sb, KERN_CRIT,
  78                              "error: remounting filesystem read-only");
  79                 sb->s_flags |= SB_RDONLY;
  80         }
  81 }
  82
  83 void ext2_msg(struct super_block *sb, const char *prefix,
  84                 const char *fmt, ...)
  85 {
  86         struct va_format vaf;
  87         va_list args;
  88
  89         va_start(args, fmt);
  90
  91         vaf.fmt = fmt;
  92         vaf.va = &args;
  93
  94         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  95
  96         va_end(args);
  97 }
  98
  99 /*
 100  * This must be called with sbi->s_lock held.
 101  */
 102 void ext2_update_dynamic_rev(struct super_block *sb)
 103 {
 104         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 105
 106         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 107                 return;
 108
 109         ext2_msg(sb, KERN_WARNING,
 110                      "warning: updating to rev %d because of "
 111                      "new feature flag, running e2fsck is recommended",
 112                      EXT2_DYNAMIC_REV);
 113
 114         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 115         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 116         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 117         /* leave es->s_feature_*compat flags alone */
 118         /* es->s_uuid will be set by e2fsck if empty */
 119
 120         /*
 121          * The rest of the superblock fields should be zero, and if not it
 122          * means they are likely already in use, so leave them alone.  We
 123          * can leave it up to e2fsck to clean up any inconsistencies there.
 124          */
 125 }
 126
 127 #ifdef CONFIG_QUOTA
 128 static int ext2_quota_off(struct super_block *sb, int type);
 129
 130 static void ext2_quota_off_umount(struct super_block *sb)
 131 {
 132         int type;
 133
 134         for (type = 0; type < MAXQUOTAS; type++)
 135                 ext2_quota_off(sb, type);
 136 }
 137 #else
 138 static inline void ext2_quota_off_umount(struct super_block *sb)
 139 {
 140 }
 141 #endif
 142
 143 static void ext2_put_super (struct super_block * sb)
 144 {
 145         int db_count;
 146         int i;
 147         struct ext2_sb_info *sbi = EXT2_SB(sb);
 148
 149         ext2_quota_off_umount(sb);
 150
 151         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
 152         sbi->s_ea_block_cache = NULL;
 153
 154         if (!sb_rdonly(sb)) {
 155                 struct ext2_super_block *es = sbi->s_es;
 156
 157                 spin_lock(&sbi->s_lock);
 158                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 159                 spin_unlock(&sbi->s_lock);
 160                 ext2_sync_super(sb, es, 1);
 161         }
 162         db_count = sbi->s_gdb_count;
 163         for (i = 0; i < db_count; i++)
 164                 if (sbi->s_group_desc[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 = kmem_cache_alloc(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_i_callback(struct rcu_head *head)
 196 {
 197         struct inode *inode = container_of(head, struct inode, i_rcu);
 198         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 199 }
 200
 201 static void ext2_destroy_inode(struct inode *inode)
 202 {
 203         call_rcu(&inode->i_rcu, ext2_i_callback);
 204 }
 205
 206 static void init_once(void *foo)
 207 {
 208         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 209
 210         rwlock_init(&ei->i_meta_lock);
 211 #ifdef CONFIG_EXT2_FS_XATTR
 212         init_rwsem(&ei->xattr_sem);
 213 #endif
 214         mutex_init(&ei->truncate_mutex);
 215 #ifdef CONFIG_FS_DAX
 216         init_rwsem(&ei->dax_sem);
 217 #endif
 218         inode_init_once(&ei->vfs_inode);
 219 }
 220
 221 static int __init init_inodecache(void)
 222 {
 223         ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
 224                                 sizeof(struct ext2_inode_info), 0,
 225                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
 226                                         SLAB_ACCOUNT),
 227                                 offsetof(struct ext2_inode_info, i_data),
 228                                 sizeof_field(struct ext2_inode_info, i_data),
 229                                 init_once);
 230         if (ext2_inode_cachep == NULL)
 231                 return -ENOMEM;
 232         return 0;
 233 }
 234
 235 static void destroy_inodecache(void)
 236 {
 237         /*
 238          * Make sure all delayed rcu free inodes are flushed before we
 239          * destroy cache.
 240          */
 241         rcu_barrier();
 242         kmem_cache_destroy(ext2_inode_cachep);
 243 }
 244
 245 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 246 {
 247         struct super_block *sb = root->d_sb;
 248         struct ext2_sb_info *sbi = EXT2_SB(sb);
 249         struct ext2_super_block *es = sbi->s_es;
 250         unsigned long def_mount_opts;
 251
 252         spin_lock(&sbi->s_lock);
 253         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 254
 255         if (sbi->s_sb_block != 1)
 256                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 257         if (test_opt(sb, MINIX_DF))
 258                 seq_puts(seq, ",minixdf");
 259         if (test_opt(sb, GRPID))
 260                 seq_puts(seq, ",grpid");
 261         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 262                 seq_puts(seq, ",nogrpid");
 263         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 264             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 265                 seq_printf(seq, ",resuid=%u",
 266                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 267         }
 268         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 269             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 270                 seq_printf(seq, ",resgid=%u",
 271                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 272         }
 273         if (test_opt(sb, ERRORS_RO)) {
 274                 int def_errors = le16_to_cpu(es->s_errors);
 275
 276                 if (def_errors == EXT2_ERRORS_PANIC ||
 277                     def_errors == EXT2_ERRORS_CONTINUE) {
 278                         seq_puts(seq, ",errors=remount-ro");
 279                 }
 280         }
 281         if (test_opt(sb, ERRORS_CONT))
 282                 seq_puts(seq, ",errors=continue");
 283         if (test_opt(sb, ERRORS_PANIC))
 284                 seq_puts(seq, ",errors=panic");
 285         if (test_opt(sb, NO_UID32))
 286                 seq_puts(seq, ",nouid32");
 287         if (test_opt(sb, DEBUG))
 288                 seq_puts(seq, ",debug");
 289         if (test_opt(sb, OLDALLOC))
 290                 seq_puts(seq, ",oldalloc");
 291
 292 #ifdef CONFIG_EXT2_FS_XATTR
 293         if (test_opt(sb, XATTR_USER))
 294                 seq_puts(seq, ",user_xattr");
 295         if (!test_opt(sb, XATTR_USER) &&
 296             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 297                 seq_puts(seq, ",nouser_xattr");
 298         }
 299 #endif
 300
 301 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 302         if (test_opt(sb, POSIX_ACL))
 303                 seq_puts(seq, ",acl");
 304         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 305                 seq_puts(seq, ",noacl");
 306 #endif
 307
 308         if (test_opt(sb, NOBH))
 309                 seq_puts(seq, ",nobh");
 310
 311         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 312                 seq_puts(seq, ",usrquota");
 313
 314         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 315                 seq_puts(seq, ",grpquota");
 316
 317         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 318                 seq_puts(seq, ",xip");
 319
 320         if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
 321                 seq_puts(seq, ",dax");
 322
 323         if (!test_opt(sb, RESERVATION))
 324                 seq_puts(seq, ",noreservation");
 325
 326         spin_unlock(&sbi->s_lock);
 327         return 0;
 328 }
 329
 330 #ifdef CONFIG_QUOTA
 331 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 332 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 333 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
 334                          const struct path *path);
 335 static struct dquot **ext2_get_dquots(struct inode *inode)
 336 {
 337         return EXT2_I(inode)->i_dquot;
 338 }
 339
 340 static const struct quotactl_ops ext2_quotactl_ops = {
 341         .quota_on       = ext2_quota_on,
 342         .quota_off      = ext2_quota_off,
 343         .quota_sync     = dquot_quota_sync,
 344         .get_state      = dquot_get_state,
 345         .set_info       = dquot_set_dqinfo,
 346         .get_dqblk      = dquot_get_dqblk,
 347         .set_dqblk      = dquot_set_dqblk,
 348         .get_nextdqblk  = dquot_get_next_dqblk,
 349 };
 350 #endif
 351
 352 static const struct super_operations ext2_sops = {
 353         .alloc_inode    = ext2_alloc_inode,
 354         .destroy_inode  = ext2_destroy_inode,
 355         .write_inode    = ext2_write_inode,
 356         .evict_inode    = ext2_evict_inode,
 357         .put_super      = ext2_put_super,
 358         .sync_fs        = ext2_sync_fs,
 359         .freeze_fs      = ext2_freeze,
 360         .unfreeze_fs    = ext2_unfreeze,
 361         .statfs         = ext2_statfs,
 362         .remount_fs     = ext2_remount,
 363         .show_options   = ext2_show_options,
 364 #ifdef CONFIG_QUOTA
 365         .quota_read     = ext2_quota_read,
 366         .quota_write    = ext2_quota_write,
 367         .get_dquots     = ext2_get_dquots,
 368 #endif
 369 };
 370
 371 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 372                 u64 ino, u32 generation)
 373 {
 374         struct inode *inode;
 375
 376         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 377                 return ERR_PTR(-ESTALE);
 378         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 379                 return ERR_PTR(-ESTALE);
 380
 381         /*
 382          * ext2_iget isn't quite right if the inode is currently unallocated!
 383          * However ext2_iget currently does appropriate checks to handle stale
 384          * inodes so everything is OK.
 385          */
 386         inode = ext2_iget(sb, ino);
 387         if (IS_ERR(inode))
 388                 return ERR_CAST(inode);
 389         if (generation && inode->i_generation != generation) {
 390                 /* we didn't find the right inode.. */
 391                 iput(inode);
 392                 return ERR_PTR(-ESTALE);
 393         }
 394         return inode;
 395 }
 396
 397 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 398                 int fh_len, int fh_type)
 399 {
 400         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 401                                     ext2_nfs_get_inode);
 402 }
 403
 404 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 405                 int fh_len, int fh_type)
 406 {
 407         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 408                                     ext2_nfs_get_inode);
 409 }
 410
 411 static const struct export_operations ext2_export_ops = {
 412         .fh_to_dentry = ext2_fh_to_dentry,
 413         .fh_to_parent = ext2_fh_to_parent,
 414         .get_parent = ext2_get_parent,
 415 };
 416
 417 static unsigned long get_sb_block(void **data)
 418 {
 419         unsigned long   sb_block;
 420         char            *options = (char *) *data;
 421
 422         if (!options || strncmp(options, "sb=", 3) != 0)
 423                 return 1;       /* Default location */
 424         options += 3;
 425         sb_block = simple_strtoul(options, &options, 0);
 426         if (*options && *options != ',') {
 427                 printk("EXT2-fs: Invalid sb specification: %s\n",
 428                        (char *) *data);
 429                 return 1;
 430         }
 431         if (*options == ',')
 432                 options++;
 433         *data = (void *) options;
 434         return sb_block;
 435 }
 436
 437 enum {
 438         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 439         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 440         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 441         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 442         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 443         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 444 };
 445
 446 static const match_table_t tokens = {
 447         {Opt_bsd_df, "bsddf"},
 448         {Opt_minix_df, "minixdf"},
 449         {Opt_grpid, "grpid"},
 450         {Opt_grpid, "bsdgroups"},
 451         {Opt_nogrpid, "nogrpid"},
 452         {Opt_nogrpid, "sysvgroups"},
 453         {Opt_resgid, "resgid=%u"},
 454         {Opt_resuid, "resuid=%u"},
 455         {Opt_sb, "sb=%u"},
 456         {Opt_err_cont, "errors=continue"},
 457         {Opt_err_panic, "errors=panic"},
 458         {Opt_err_ro, "errors=remount-ro"},
 459         {Opt_nouid32, "nouid32"},
 460         {Opt_nocheck, "check=none"},
 461         {Opt_nocheck, "nocheck"},
 462         {Opt_debug, "debug"},
 463         {Opt_oldalloc, "oldalloc"},
 464         {Opt_orlov, "orlov"},
 465         {Opt_nobh, "nobh"},
 466         {Opt_user_xattr, "user_xattr"},
 467         {Opt_nouser_xattr, "nouser_xattr"},
 468         {Opt_acl, "acl"},
 469         {Opt_noacl, "noacl"},
 470         {Opt_xip, "xip"},
 471         {Opt_dax, "dax"},
 472         {Opt_grpquota, "grpquota"},
 473         {Opt_ignore, "noquota"},
 474         {Opt_quota, "quota"},
 475         {Opt_usrquota, "usrquota"},
 476         {Opt_reservation, "reservation"},
 477         {Opt_noreservation, "noreservation"},
 478         {Opt_err, NULL}
 479 };
 480
 481 static int parse_options(char *options, struct super_block *sb,
 482                          struct ext2_mount_options *opts)
 483 {
 484         char *p;
 485         substring_t args[MAX_OPT_ARGS];
 486         int option;
 487         kuid_t uid;
 488         kgid_t gid;
 489
 490         if (!options)
 491                 return 1;
 492
 493         while ((p = strsep (&options, ",")) != NULL) {
 494                 int token;
 495                 if (!*p)
 496                         continue;
 497
 498                 token = match_token(p, tokens, args);
 499                 switch (token) {
 500                 case Opt_bsd_df:
 501                         clear_opt (opts->s_mount_opt, MINIX_DF);
 502                         break;
 503                 case Opt_minix_df:
 504                         set_opt (opts->s_mount_opt, MINIX_DF);
 505                         break;
 506                 case Opt_grpid:
 507                         set_opt (opts->s_mount_opt, GRPID);
 508                         break;
 509                 case Opt_nogrpid:
 510                         clear_opt (opts->s_mount_opt, GRPID);
 511                         break;
 512                 case Opt_resuid:
 513                         if (match_int(&args[0], &option))
 514                                 return 0;
 515                         uid = make_kuid(current_user_ns(), option);
 516                         if (!uid_valid(uid)) {
 517                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 518                                 return 0;
 519
 520                         }
 521                         opts->s_resuid = uid;
 522                         break;
 523                 case Opt_resgid:
 524                         if (match_int(&args[0], &option))
 525                                 return 0;
 526                         gid = make_kgid(current_user_ns(), option);
 527                         if (!gid_valid(gid)) {
 528                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 529                                 return 0;
 530                         }
 531                         opts->s_resgid = gid;
 532                         break;
 533                 case Opt_sb:
 534                         /* handled by get_sb_block() instead of here */
 535                         /* *sb_block = match_int(&args[0]); */
 536                         break;
 537                 case Opt_err_panic:
 538                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 539                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 540                         set_opt (opts->s_mount_opt, ERRORS_PANIC);
 541                         break;
 542                 case Opt_err_ro:
 543                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 544                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 545                         set_opt (opts->s_mount_opt, ERRORS_RO);
 546                         break;
 547                 case Opt_err_cont:
 548                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 549                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 550                         set_opt (opts->s_mount_opt, ERRORS_CONT);
 551                         break;
 552                 case Opt_nouid32:
 553                         set_opt (opts->s_mount_opt, NO_UID32);
 554                         break;
 555                 case Opt_nocheck:
 556                         ext2_msg(sb, KERN_WARNING,
 557                                 "Option nocheck/check=none is deprecated and"
 558                                 " will be removed in June 2020.");
 559                         clear_opt (opts->s_mount_opt, CHECK);
 560                         break;
 561                 case Opt_debug:
 562                         set_opt (opts->s_mount_opt, DEBUG);
 563                         break;
 564                 case Opt_oldalloc:
 565                         set_opt (opts->s_mount_opt, OLDALLOC);
 566                         break;
 567                 case Opt_orlov:
 568                         clear_opt (opts->s_mount_opt, OLDALLOC);
 569                         break;
 570                 case Opt_nobh:
 571                         set_opt (opts->s_mount_opt, NOBH);
 572                         break;
 573 #ifdef CONFIG_EXT2_FS_XATTR
 574                 case Opt_user_xattr:
 575                         set_opt (opts->s_mount_opt, XATTR_USER);
 576                         break;
 577                 case Opt_nouser_xattr:
 578                         clear_opt (opts->s_mount_opt, XATTR_USER);
 579                         break;
 580 #else
 581                 case Opt_user_xattr:
 582                 case Opt_nouser_xattr:
 583                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 584                                 "not supported");
 585                         break;
 586 #endif
 587 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 588                 case Opt_acl:
 589                         set_opt(opts->s_mount_opt, POSIX_ACL);
 590                         break;
 591                 case Opt_noacl:
 592                         clear_opt(opts->s_mount_opt, POSIX_ACL);
 593                         break;
 594 #else
 595                 case Opt_acl:
 596                 case Opt_noacl:
 597                         ext2_msg(sb, KERN_INFO,
 598                                 "(no)acl options not supported");
 599                         break;
 600 #endif
 601                 case Opt_xip:
 602                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 603                         set_opt(opts->s_mount_opt, XIP);
 604                         /* Fall through */
 605                 case Opt_dax:
 606 #ifdef CONFIG_FS_DAX
 607                         ext2_msg(sb, KERN_WARNING,
 608                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 609                         set_opt(opts->s_mount_opt, DAX);
 610 #else
 611                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 612 #endif
 613                         break;
 614
 615 #if defined(CONFIG_QUOTA)
 616                 case Opt_quota:
 617                 case Opt_usrquota:
 618                         set_opt(opts->s_mount_opt, USRQUOTA);
 619                         break;
 620
 621                 case Opt_grpquota:
 622                         set_opt(opts->s_mount_opt, GRPQUOTA);
 623                         break;
 624 #else
 625                 case Opt_quota:
 626                 case Opt_usrquota:
 627                 case Opt_grpquota:
 628                         ext2_msg(sb, KERN_INFO,
 629                                 "quota operations not supported");
 630                         break;
 631 #endif
 632
 633                 case Opt_reservation:
 634                         set_opt(opts->s_mount_opt, RESERVATION);
 635                         ext2_msg(sb, KERN_INFO, "reservations ON");
 636                         break;
 637                 case Opt_noreservation:
 638                         clear_opt(opts->s_mount_opt, RESERVATION);
 639                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 640                         break;
 641                 case Opt_ignore:
 642                         break;
 643                 default:
 644                         return 0;
 645                 }
 646         }
 647         return 1;
 648 }
 649
 650 static int ext2_setup_super (struct super_block * sb,
 651                               struct ext2_super_block * es,
 652                               int read_only)
 653 {
 654         int res = 0;
 655         struct ext2_sb_info *sbi = EXT2_SB(sb);
 656
 657         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 658                 ext2_msg(sb, KERN_ERR,
 659                         "error: revision level too high, "
 660                         "forcing read-only mode");
 661                 res = SB_RDONLY;
 662         }
 663         if (read_only)
 664                 return res;
 665         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 666                 ext2_msg(sb, KERN_WARNING,
 667                         "warning: mounting unchecked fs, "
 668                         "running e2fsck is recommended");
 669         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 670                 ext2_msg(sb, KERN_WARNING,
 671                         "warning: mounting fs with errors, "
 672                         "running e2fsck is recommended");
 673         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 674                  le16_to_cpu(es->s_mnt_count) >=
 675                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 676                 ext2_msg(sb, KERN_WARNING,
 677                         "warning: maximal mount count reached, "
 678                         "running e2fsck is recommended");
 679         else if (le32_to_cpu(es->s_checkinterval) &&
 680                 (le32_to_cpu(es->s_lastcheck) +
 681                         le32_to_cpu(es->s_checkinterval) <=
 682                         ktime_get_real_seconds()))
 683                 ext2_msg(sb, KERN_WARNING,
 684                         "warning: checktime reached, "
 685                         "running e2fsck is recommended");
 686         if (!le16_to_cpu(es->s_max_mnt_count))
 687                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 688         le16_add_cpu(&es->s_mnt_count, 1);
 689         if (test_opt (sb, DEBUG))
 690                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 691                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 692                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 693                         sbi->s_frag_size,
 694                         sbi->s_groups_count,
 695                         EXT2_BLOCKS_PER_GROUP(sb),
 696                         EXT2_INODES_PER_GROUP(sb),
 697                         sbi->s_mount_opt);
 698         return res;
 699 }
 700
 701 static int ext2_check_descriptors(struct super_block *sb)
 702 {
 703         int i;
 704         struct ext2_sb_info *sbi = EXT2_SB(sb);
 705
 706         ext2_debug ("Checking group descriptors");
 707
 708         for (i = 0; i < sbi->s_groups_count; i++) {
 709                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 710                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 711                 ext2_fsblk_t last_block;
 712
 713                 if (i == sbi->s_groups_count - 1)
 714                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 715                 else
 716                         last_block = first_block +
 717                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 718
 719                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 720                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 721                 {
 722                         ext2_error (sb, "ext2_check_descriptors",
 723                                     "Block bitmap for group %d"
 724                                     " not in group (block %lu)!",
 725                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 726                         return 0;
 727                 }
 728                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 729                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 730                 {
 731                         ext2_error (sb, "ext2_check_descriptors",
 732                                     "Inode bitmap for group %d"
 733                                     " not in group (block %lu)!",
 734                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 735                         return 0;
 736                 }
 737                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 738                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 739                     last_block)
 740                 {
 741                         ext2_error (sb, "ext2_check_descriptors",
 742                                     "Inode table for group %d"
 743                                     " not in group (block %lu)!",
 744                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 745                         return 0;
 746                 }
 747         }
 748         return 1;
 749 }
 750
 751 /*
 752  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 753  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 754  * We need to be 1 filesystem block less than the 2^32 sector limit.
 755  */
 756 static loff_t ext2_max_size(int bits)
 757 {
 758         loff_t res = EXT2_NDIR_BLOCKS;
 759         int meta_blocks;
 760         loff_t upper_limit;
 761
 762         /* This is calculated to be the largest file size for a
 763          * dense, file such that the total number of
 764          * sectors in the file, including data and all indirect blocks,
 765          * does not exceed 2^32 -1
 766          * __u32 i_blocks representing the total number of
 767          * 512 bytes blocks of the file
 768          */
 769         upper_limit = (1LL << 32) - 1;
 770
 771         /* total blocks in file system block size */
 772         upper_limit >>= (bits - 9);
 773
 774
 775         /* indirect blocks */
 776         meta_blocks = 1;
 777         /* double indirect blocks */
 778         meta_blocks += 1 + (1LL << (bits-2));
 779         /* tripple indirect blocks */
 780         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 781
 782         upper_limit -= meta_blocks;
 783         upper_limit <<= bits;
 784
 785         res += 1LL << (bits-2);
 786         res += 1LL << (2*(bits-2));
 787         res += 1LL << (3*(bits-2));
 788         res <<= bits;
 789         if (res > upper_limit)
 790                 res = upper_limit;
 791
 792         if (res > MAX_LFS_FILESIZE)
 793                 res = MAX_LFS_FILESIZE;
 794
 795         return res;
 796 }
 797
 798 static unsigned long descriptor_loc(struct super_block *sb,
 799                                     unsigned long logic_sb_block,
 800                                     int nr)
 801 {
 802         struct ext2_sb_info *sbi = EXT2_SB(sb);
 803         unsigned long bg, first_meta_bg;
 804         int has_super = 0;
 805
 806         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 807
 808         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 809             nr < first_meta_bg)
 810                 return (logic_sb_block + nr + 1);
 811         bg = sbi->s_desc_per_block * nr;
 812         if (ext2_bg_has_super(sb, bg))
 813                 has_super = 1;
 814
 815         return ext2_group_first_block_no(sb, bg) + has_super;
 816 }
 817
 818 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 819 {
 820         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 821         struct buffer_head * bh;
 822         struct ext2_sb_info * sbi;
 823         struct ext2_super_block * es;
 824         struct inode *root;
 825         unsigned long block;
 826         unsigned long sb_block = get_sb_block(&data);
 827         unsigned long logic_sb_block;
 828         unsigned long offset = 0;
 829         unsigned long def_mount_opts;
 830         long ret = -ENOMEM;
 831         int blocksize = BLOCK_SIZE;
 832         int db_count;
 833         int i, j;
 834         __le32 features;
 835         int err;
 836         struct ext2_mount_options opts;
 837
 838         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 839         if (!sbi)
 840                 goto failed;
 841
 842         sbi->s_blockgroup_lock =
 843                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 844         if (!sbi->s_blockgroup_lock) {
 845                 kfree(sbi);
 846                 goto failed;
 847         }
 848         sb->s_fs_info = sbi;
 849         sbi->s_sb_block = sb_block;
 850         sbi->s_daxdev = dax_dev;
 851
 852         spin_lock_init(&sbi->s_lock);
 853         ret = -EINVAL;
 854
 855         /*
 856          * See what the current blocksize for the device is, and
 857          * use that as the blocksize.  Otherwise (or if the blocksize
 858          * is smaller than the default) use the default.
 859          * This is important for devices that have a hardware
 860          * sectorsize that is larger than the default.
 861          */
 862         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 863         if (!blocksize) {
 864                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 865                 goto failed_sbi;
 866         }
 867
 868         /*
 869          * If the superblock doesn't start on a hardware sector boundary,
 870          * calculate the offset.
 871          */
 872         if (blocksize != BLOCK_SIZE) {
 873                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 874                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 875         } else {
 876                 logic_sb_block = sb_block;
 877         }
 878
 879         if (!(bh = sb_bread(sb, logic_sb_block))) {
 880                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 881                 goto failed_sbi;
 882         }
 883         /*
 884          * Note: s_es must be initialized as soon as possible because
 885          *       some ext2 macro-instructions depend on its value
 886          */
 887         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 888         sbi->s_es = es;
 889         sb->s_magic = le16_to_cpu(es->s_magic);
 890
 891         if (sb->s_magic != EXT2_SUPER_MAGIC)
 892                 goto cantfind_ext2;
 893
 894         opts.s_mount_opt = 0;
 895         /* Set defaults before we parse the mount options */
 896         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 897         if (def_mount_opts & EXT2_DEFM_DEBUG)
 898                 set_opt(opts.s_mount_opt, DEBUG);
 899         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 900                 set_opt(opts.s_mount_opt, GRPID);
 901         if (def_mount_opts & EXT2_DEFM_UID16)
 902                 set_opt(opts.s_mount_opt, NO_UID32);
 903 #ifdef CONFIG_EXT2_FS_XATTR
 904         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 905                 set_opt(opts.s_mount_opt, XATTR_USER);
 906 #endif
 907 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 908         if (def_mount_opts & EXT2_DEFM_ACL)
 909                 set_opt(opts.s_mount_opt, POSIX_ACL);
 910 #endif
 911
 912         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 913                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 914         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 915                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 916         else
 917                 set_opt(opts.s_mount_opt, ERRORS_RO);
 918
 919         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 920         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 921
 922         set_opt(opts.s_mount_opt, RESERVATION);
 923
 924         if (!parse_options((char *) data, sb, &opts))
 925                 goto failed_mount;
 926
 927         sbi->s_mount_opt = opts.s_mount_opt;
 928         sbi->s_resuid = opts.s_resuid;
 929         sbi->s_resgid = opts.s_resgid;
 930
 931         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 932                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 933                  SB_POSIXACL : 0);
 934         sb->s_iflags |= SB_I_CGROUPWB;
 935
 936         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 937             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 938              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 939              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 940                 ext2_msg(sb, KERN_WARNING,
 941                         "warning: feature flags set on rev 0 fs, "
 942                         "running e2fsck is recommended");
 943         /*
 944          * Check feature flags regardless of the revision level, since we
 945          * previously didn't change the revision level when setting the flags,
 946          * so there is a chance incompat flags are set on a rev 0 filesystem.
 947          */
 948         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 949         if (features) {
 950                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 951                        "unsupported optional features (%x)",
 952                         le32_to_cpu(features));
 953                 goto failed_mount;
 954         }
 955         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 956                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 957                        "unsupported optional features (%x)",
 958                        le32_to_cpu(features));
 959                 goto failed_mount;
 960         }
 961
 962         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 963
 964         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
 965                 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
 966                         ext2_msg(sb, KERN_ERR,
 967                                 "DAX unsupported by block device. Turning off DAX.");
 968                         sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
 969                 }
 970         }
 971
 972         /* If the blocksize doesn't match, re-read the thing.. */
 973         if (sb->s_blocksize != blocksize) {
 974                 brelse(bh);
 975
 976                 if (!sb_set_blocksize(sb, blocksize)) {
 977                         ext2_msg(sb, KERN_ERR,
 978                                 "error: bad blocksize %d", blocksize);
 979                         goto failed_sbi;
 980                 }
 981
 982                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 983                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 984                 bh = sb_bread(sb, logic_sb_block);
 985                 if(!bh) {
 986                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 987                                 "superblock on 2nd try");
 988                         goto failed_sbi;
 989                 }
 990                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 991                 sbi->s_es = es;
 992                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 993                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 994                         goto failed_mount;
 995                 }
 996         }
 997
 998         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 999         sb->s_max_links = EXT2_LINK_MAX;
1000
1001         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1002                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1003                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1004         } else {
1005                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1006                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1007                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1008                     !is_power_of_2(sbi->s_inode_size) ||
1009                     (sbi->s_inode_size > blocksize)) {
1010                         ext2_msg(sb, KERN_ERR,
1011                                 "error: unsupported inode size: %d",
1012                                 sbi->s_inode_size);
1013                         goto failed_mount;
1014                 }
1015         }
1016
1017         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1018                                    le32_to_cpu(es->s_log_frag_size);
1019         if (sbi->s_frag_size == 0)
1020                 goto cantfind_ext2;
1021         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1022
1023         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1024         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1025         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1026
1027         if (EXT2_INODE_SIZE(sb) == 0)
1028                 goto cantfind_ext2;
1029         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1030         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1031                 goto cantfind_ext2;
1032         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1033                                         sbi->s_inodes_per_block;
1034         sbi->s_desc_per_block = sb->s_blocksize /
1035                                         sizeof (struct ext2_group_desc);
1036         sbi->s_sbh = bh;
1037         sbi->s_mount_state = le16_to_cpu(es->s_state);
1038         sbi->s_addr_per_block_bits =
1039                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1040         sbi->s_desc_per_block_bits =
1041                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1042
1043         if (sb->s_magic != EXT2_SUPER_MAGIC)
1044                 goto cantfind_ext2;
1045
1046         if (sb->s_blocksize != bh->b_size) {
1047                 if (!silent)
1048                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1049                 goto failed_mount;
1050         }
1051
1052         if (sb->s_blocksize != sbi->s_frag_size) {
1053                 ext2_msg(sb, KERN_ERR,
1054                         "error: fragsize %lu != blocksize %lu"
1055                         "(not supported yet)",
1056                         sbi->s_frag_size, sb->s_blocksize);
1057                 goto failed_mount;
1058         }
1059
1060         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1061                 ext2_msg(sb, KERN_ERR,
1062                         "error: #blocks per group too big: %lu",
1063                         sbi->s_blocks_per_group);
1064                 goto failed_mount;
1065         }
1066         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1067                 ext2_msg(sb, KERN_ERR,
1068                         "error: #fragments per group too big: %lu",
1069                         sbi->s_frags_per_group);
1070                 goto failed_mount;
1071         }
1072         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1073                 ext2_msg(sb, KERN_ERR,
1074                         "error: #inodes per group too big: %lu",
1075                         sbi->s_inodes_per_group);
1076                 goto failed_mount;
1077         }
1078
1079         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1080                 goto cantfind_ext2;
1081         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1082                                 le32_to_cpu(es->s_first_data_block) - 1)
1083                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1084         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1085                    EXT2_DESC_PER_BLOCK(sb);
1086         sbi->s_group_desc = kmalloc_array (db_count,
1087                                            sizeof(struct buffer_head *),
1088                                            GFP_KERNEL);
1089         if (sbi->s_group_desc == NULL) {
1090                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1091                 goto failed_mount;
1092         }
1093         bgl_lock_init(sbi->s_blockgroup_lock);
1094         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1095         if (!sbi->s_debts) {
1096                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1097                 goto failed_mount_group_desc;
1098         }
1099         for (i = 0; i < db_count; i++) {
1100                 block = descriptor_loc(sb, logic_sb_block, i);
1101                 sbi->s_group_desc[i] = sb_bread(sb, block);
1102                 if (!sbi->s_group_desc[i]) {
1103                         for (j = 0; j < i; j++)
1104                                 brelse (sbi->s_group_desc[j]);
1105                         ext2_msg(sb, KERN_ERR,
1106                                 "error: unable to read group descriptors");
1107                         goto failed_mount_group_desc;
1108                 }
1109         }
1110         if (!ext2_check_descriptors (sb)) {
1111                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1112                 goto failed_mount2;
1113         }
1114         sbi->s_gdb_count = db_count;
1115         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1116         spin_lock_init(&sbi->s_next_gen_lock);
1117
1118         /* per fileystem reservation list head & lock */
1119         spin_lock_init(&sbi->s_rsv_window_lock);
1120         sbi->s_rsv_window_root = RB_ROOT;
1121         /*
1122          * Add a single, static dummy reservation to the start of the
1123          * reservation window list --- it gives us a placeholder for
1124          * append-at-start-of-list which makes the allocation logic
1125          * _much_ simpler.
1126          */
1127         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1128         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1129         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1130         sbi->s_rsv_window_head.rsv_goal_size = 0;
1131         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1132
1133         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1134                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1135         if (!err) {
1136                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1137                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1138         }
1139         if (!err) {
1140                 err = percpu_counter_init(&sbi->s_dirs_counter,
1141                                 ext2_count_dirs(sb), GFP_KERNEL);
1142         }
1143         if (err) {
1144                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1145                 goto failed_mount3;
1146         }
1147
1148 #ifdef CONFIG_EXT2_FS_XATTR
1149         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1150         if (!sbi->s_ea_block_cache) {
1151                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1152                 goto failed_mount3;
1153         }
1154 #endif
1155         /*
1156          * set up enough so that it can read an inode
1157          */
1158         sb->s_op = &ext2_sops;
1159         sb->s_export_op = &ext2_export_ops;
1160         sb->s_xattr = ext2_xattr_handlers;
1161
1162 #ifdef CONFIG_QUOTA
1163         sb->dq_op = &dquot_operations;
1164         sb->s_qcop = &ext2_quotactl_ops;
1165         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1166 #endif
1167
1168         root = ext2_iget(sb, EXT2_ROOT_INO);
1169         if (IS_ERR(root)) {
1170                 ret = PTR_ERR(root);
1171                 goto failed_mount3;
1172         }
1173         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1174                 iput(root);
1175                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1176                 goto failed_mount3;
1177         }
1178
1179         sb->s_root = d_make_root(root);
1180         if (!sb->s_root) {
1181                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1182                 ret = -ENOMEM;
1183                 goto failed_mount3;
1184         }
1185         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1186                 ext2_msg(sb, KERN_WARNING,
1187                         "warning: mounting ext3 filesystem as ext2");
1188         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1189                 sb->s_flags |= SB_RDONLY;
1190         ext2_write_super(sb);
1191         return 0;
1192
1193 cantfind_ext2:
1194         if (!silent)
1195                 ext2_msg(sb, KERN_ERR,
1196                         "error: can't find an ext2 filesystem on dev %s.",
1197                         sb->s_id);
1198         goto failed_mount;
1199 failed_mount3:
1200         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1201         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1202         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1203         percpu_counter_destroy(&sbi->s_dirs_counter);
1204 failed_mount2:
1205         for (i = 0; i < db_count; i++)
1206                 brelse(sbi->s_group_desc[i]);
1207 failed_mount_group_desc:
1208         kfree(sbi->s_group_desc);
1209         kfree(sbi->s_debts);
1210 failed_mount:
1211         brelse(bh);
1212 failed_sbi:
1213         sb->s_fs_info = NULL;
1214         kfree(sbi->s_blockgroup_lock);
1215         kfree(sbi);
1216 failed:
1217         fs_put_dax(dax_dev);
1218         return ret;
1219 }
1220
1221 static void ext2_clear_super_error(struct super_block *sb)
1222 {
1223         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1224
1225         if (buffer_write_io_error(sbh)) {
1226                 /*
1227                  * Oh, dear.  A previous attempt to write the
1228                  * superblock failed.  This could happen because the
1229                  * USB device was yanked out.  Or it could happen to
1230                  * be a transient write error and maybe the block will
1231                  * be remapped.  Nothing we can do but to retry the
1232                  * write and hope for the best.
1233                  */
1234                 ext2_msg(sb, KERN_ERR,
1235                        "previous I/O error to superblock detected");
1236                 clear_buffer_write_io_error(sbh);
1237                 set_buffer_uptodate(sbh);
1238         }
1239 }
1240
1241 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1242                      int wait)
1243 {
1244         ext2_clear_super_error(sb);
1245         spin_lock(&EXT2_SB(sb)->s_lock);
1246         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1247         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1248         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1249         /* unlock before we do IO */
1250         spin_unlock(&EXT2_SB(sb)->s_lock);
1251         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1252         if (wait)
1253                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1254 }
1255
1256 /*
1257  * In the second extended file system, it is not necessary to
1258  * write the super block since we use a mapping of the
1259  * disk super block in a buffer.
1260  *
1261  * However, this function is still used to set the fs valid
1262  * flags to 0.  We need to set this flag to 0 since the fs
1263  * may have been checked while mounted and e2fsck may have
1264  * set s_state to EXT2_VALID_FS after some corrections.
1265  */
1266 static int ext2_sync_fs(struct super_block *sb, int wait)
1267 {
1268         struct ext2_sb_info *sbi = EXT2_SB(sb);
1269         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1270
1271         /*
1272          * Write quota structures to quota file, sync_blockdev() will write
1273          * them to disk later
1274          */
1275         dquot_writeback_dquots(sb, -1);
1276
1277         spin_lock(&sbi->s_lock);
1278         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1279                 ext2_debug("setting valid to 0\n");
1280                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1281         }
1282         spin_unlock(&sbi->s_lock);
1283         ext2_sync_super(sb, es, wait);
1284         return 0;
1285 }
1286
1287 static int ext2_freeze(struct super_block *sb)
1288 {
1289         struct ext2_sb_info *sbi = EXT2_SB(sb);
1290
1291         /*
1292          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1293          * because we have unattached inodes and thus filesystem is not fully
1294          * consistent.
1295          */
1296         if (atomic_long_read(&sb->s_remove_count)) {
1297                 ext2_sync_fs(sb, 1);
1298                 return 0;
1299         }
1300         /* Set EXT2_FS_VALID flag */
1301         spin_lock(&sbi->s_lock);
1302         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1303         spin_unlock(&sbi->s_lock);
1304         ext2_sync_super(sb, sbi->s_es, 1);
1305
1306         return 0;
1307 }
1308
1309 static int ext2_unfreeze(struct super_block *sb)
1310 {
1311         /* Just write sb to clear EXT2_VALID_FS flag */
1312         ext2_write_super(sb);
1313
1314         return 0;
1315 }
1316
1317 static void ext2_write_super(struct super_block *sb)
1318 {
1319         if (!sb_rdonly(sb))
1320                 ext2_sync_fs(sb, 1);
1321 }
1322
1323 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1324 {
1325         struct ext2_sb_info * sbi = EXT2_SB(sb);
1326         struct ext2_super_block * es;
1327         struct ext2_mount_options new_opts;
1328         int err;
1329
1330         sync_filesystem(sb);
1331
1332         spin_lock(&sbi->s_lock);
1333         new_opts.s_mount_opt = sbi->s_mount_opt;
1334         new_opts.s_resuid = sbi->s_resuid;
1335         new_opts.s_resgid = sbi->s_resgid;
1336         spin_unlock(&sbi->s_lock);
1337
1338         if (!parse_options(data, sb, &new_opts))
1339                 return -EINVAL;
1340
1341         spin_lock(&sbi->s_lock);
1342         es = sbi->s_es;
1343         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1344                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1345                          "dax flag with busy inodes while remounting");
1346                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1347         }
1348         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1349                 goto out_set;
1350         if (*flags & SB_RDONLY) {
1351                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1352                     !(sbi->s_mount_state & EXT2_VALID_FS))
1353                         goto out_set;
1354
1355                 /*
1356                  * OK, we are remounting a valid rw partition rdonly, so set
1357                  * the rdonly flag and then mark the partition as valid again.
1358                  */
1359                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1360                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1361                 spin_unlock(&sbi->s_lock);
1362
1363                 err = dquot_suspend(sb, -1);
1364                 if (err < 0)
1365                         return err;
1366
1367                 ext2_sync_super(sb, es, 1);
1368         } else {
1369                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1370                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1371                 if (ret) {
1372                         spin_unlock(&sbi->s_lock);
1373                         ext2_msg(sb, KERN_WARNING,
1374                                 "warning: couldn't remount RDWR because of "
1375                                 "unsupported optional features (%x).",
1376                                 le32_to_cpu(ret));
1377                         return -EROFS;
1378                 }
1379                 /*
1380                  * Mounting a RDONLY partition read-write, so reread and
1381                  * store the current valid flag.  (It may have been changed
1382                  * by e2fsck since we originally mounted the partition.)
1383                  */
1384                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1385                 if (!ext2_setup_super (sb, es, 0))
1386                         sb->s_flags &= ~SB_RDONLY;
1387                 spin_unlock(&sbi->s_lock);
1388
1389                 ext2_write_super(sb);
1390
1391                 dquot_resume(sb, -1);
1392         }
1393
1394         spin_lock(&sbi->s_lock);
1395 out_set:
1396         sbi->s_mount_opt = new_opts.s_mount_opt;
1397         sbi->s_resuid = new_opts.s_resuid;
1398         sbi->s_resgid = new_opts.s_resgid;
1399         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1400                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1401         spin_unlock(&sbi->s_lock);
1402
1403         return 0;
1404 }
1405
1406 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1407 {
1408         struct super_block *sb = dentry->d_sb;
1409         struct ext2_sb_info *sbi = EXT2_SB(sb);
1410         struct ext2_super_block *es = sbi->s_es;
1411         u64 fsid;
1412
1413         spin_lock(&sbi->s_lock);
1414
1415         if (test_opt (sb, MINIX_DF))
1416                 sbi->s_overhead_last = 0;
1417         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1418                 unsigned long i, overhead = 0;
1419                 smp_rmb();
1420
1421                 /*
1422                  * Compute the overhead (FS structures). This is constant
1423                  * for a given filesystem unless the number of block groups
1424                  * changes so we cache the previous value until it does.
1425                  */
1426
1427                 /*
1428                  * All of the blocks before first_data_block are
1429                  * overhead
1430                  */
1431                 overhead = le32_to_cpu(es->s_first_data_block);
1432
1433                 /*
1434                  * Add the overhead attributed to the superblock and
1435                  * block group descriptors.  If the sparse superblocks
1436                  * feature is turned on, then not all groups have this.
1437                  */
1438                 for (i = 0; i < sbi->s_groups_count; i++)
1439                         overhead += ext2_bg_has_super(sb, i) +
1440                                 ext2_bg_num_gdb(sb, i);
1441
1442                 /*
1443                  * Every block group has an inode bitmap, a block
1444                  * bitmap, and an inode table.
1445                  */
1446                 overhead += (sbi->s_groups_count *
1447                              (2 + sbi->s_itb_per_group));
1448                 sbi->s_overhead_last = overhead;
1449                 smp_wmb();
1450                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1451         }
1452
1453         buf->f_type = EXT2_SUPER_MAGIC;
1454         buf->f_bsize = sb->s_blocksize;
1455         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1456         buf->f_bfree = ext2_count_free_blocks(sb);
1457         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1458         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1459         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1460                 buf->f_bavail = 0;
1461         buf->f_files = le32_to_cpu(es->s_inodes_count);
1462         buf->f_ffree = ext2_count_free_inodes(sb);
1463         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1464         buf->f_namelen = EXT2_NAME_LEN;
1465         fsid = le64_to_cpup((void *)es->s_uuid) ^
1466                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1467         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1468         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1469         spin_unlock(&sbi->s_lock);
1470         return 0;
1471 }
1472
1473 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1474         int flags, const char *dev_name, void *data)
1475 {
1476         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1477 }
1478
1479 #ifdef CONFIG_QUOTA
1480
1481 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1482  * acquiring the locks... As quota files are never truncated and quota code
1483  * itself serializes the operations (and no one else should touch the files)
1484  * we don't have to be afraid of races */
1485 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1486                                size_t len, loff_t off)
1487 {
1488         struct inode *inode = sb_dqopt(sb)->files[type];
1489         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1490         int err = 0;
1491         int offset = off & (sb->s_blocksize - 1);
1492         int tocopy;
1493         size_t toread;
1494         struct buffer_head tmp_bh;
1495         struct buffer_head *bh;
1496         loff_t i_size = i_size_read(inode);
1497
1498         if (off > i_size)
1499                 return 0;
1500         if (off+len > i_size)
1501                 len = i_size-off;
1502         toread = len;
1503         while (toread > 0) {
1504                 tocopy = sb->s_blocksize - offset < toread ?
1505                                 sb->s_blocksize - offset : toread;
1506
1507                 tmp_bh.b_state = 0;
1508                 tmp_bh.b_size = sb->s_blocksize;
1509                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1510                 if (err < 0)
1511                         return err;
1512                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1513                         memset(data, 0, tocopy);
1514                 else {
1515                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1516                         if (!bh)
1517                                 return -EIO;
1518                         memcpy(data, bh->b_data+offset, tocopy);
1519                         brelse(bh);
1520                 }
1521                 offset = 0;
1522                 toread -= tocopy;
1523                 data += tocopy;
1524                 blk++;
1525         }
1526         return len;
1527 }
1528
1529 /* Write to quotafile */
1530 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1531                                 const char *data, size_t len, loff_t off)
1532 {
1533         struct inode *inode = sb_dqopt(sb)->files[type];
1534         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1535         int err = 0;
1536         int offset = off & (sb->s_blocksize - 1);
1537         int tocopy;
1538         size_t towrite = len;
1539         struct buffer_head tmp_bh;
1540         struct buffer_head *bh;
1541
1542         while (towrite > 0) {
1543                 tocopy = sb->s_blocksize - offset < towrite ?
1544                                 sb->s_blocksize - offset : towrite;
1545
1546                 tmp_bh.b_state = 0;
1547                 tmp_bh.b_size = sb->s_blocksize;
1548                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1549                 if (err < 0)
1550                         goto out;
1551                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1552                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1553                 else
1554                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1555                 if (unlikely(!bh)) {
1556                         err = -EIO;
1557                         goto out;
1558                 }
1559                 lock_buffer(bh);
1560                 memcpy(bh->b_data+offset, data, tocopy);
1561                 flush_dcache_page(bh->b_page);
1562                 set_buffer_uptodate(bh);
1563                 mark_buffer_dirty(bh);
1564                 unlock_buffer(bh);
1565                 brelse(bh);
1566                 offset = 0;
1567                 towrite -= tocopy;
1568                 data += tocopy;
1569                 blk++;
1570         }
1571 out:
1572         if (len == towrite)
1573                 return err;
1574         if (inode->i_size < off+len-towrite)
1575                 i_size_write(inode, off+len-towrite);
1576         inode_inc_iversion(inode);
1577         inode->i_mtime = inode->i_ctime = current_time(inode);
1578         mark_inode_dirty(inode);
1579         return len - towrite;
1580 }
1581
1582 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1583                          const struct path *path)
1584 {
1585         int err;
1586         struct inode *inode;
1587
1588         err = dquot_quota_on(sb, type, format_id, path);
1589         if (err)
1590                 return err;
1591
1592         inode = d_inode(path->dentry);
1593         inode_lock(inode);
1594         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1595         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1596                         S_NOATIME | S_IMMUTABLE);
1597         inode_unlock(inode);
1598         mark_inode_dirty(inode);
1599
1600         return 0;
1601 }
1602
1603 static int ext2_quota_off(struct super_block *sb, int type)
1604 {
1605         struct inode *inode = sb_dqopt(sb)->files[type];
1606         int err;
1607
1608         if (!inode || !igrab(inode))
1609                 goto out;
1610
1611         err = dquot_quota_off(sb, type);
1612         if (err)
1613                 goto out_put;
1614
1615         inode_lock(inode);
1616         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1617         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1618         inode_unlock(inode);
1619         mark_inode_dirty(inode);
1620 out_put:
1621         iput(inode);
1622         return err;
1623 out:
1624         return dquot_quota_off(sb, type);
1625 }
1626
1627 #endif
1628
1629 static struct file_system_type ext2_fs_type = {
1630         .owner          = THIS_MODULE,
1631         .name           = "ext2",
1632         .mount          = ext2_mount,
1633         .kill_sb        = kill_block_super,
1634         .fs_flags       = FS_REQUIRES_DEV,
1635 };
1636 MODULE_ALIAS_FS("ext2");
1637
1638 static int __init init_ext2_fs(void)
1639 {
1640         int err;
1641
1642         err = init_inodecache();
1643         if (err)
1644                 return err;
1645         err = register_filesystem(&ext2_fs_type);
1646         if (err)
1647                 goto out;
1648         return 0;
1649 out:
1650         destroy_inodecache();
1651         return err;
1652 }
1653
1654 static void __exit exit_ext2_fs(void)
1655 {
1656         unregister_filesystem(&ext2_fs_type);
1657         destroy_inodecache();
1658 }
1659
1660 MODULE_AUTHOR("Remy Card and others");
1661 MODULE_DESCRIPTION("Second Extended Filesystem");
1662 MODULE_LICENSE("GPL");
1663 module_init(init_ext2_fs)
1664 module_exit(exit_ext2_fs)