fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   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 (davem@caip.rutgers.edu), 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 "ext2.h"
  37 #include "xattr.h"
  38 #include "acl.h"
  39
  40 static void ext2_write_super(struct super_block *sb);
  41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  43 static int ext2_sync_fs(struct super_block *sb, int wait);
  44 static int ext2_freeze(struct super_block *sb);
  45 static int ext2_unfreeze(struct super_block *sb);
  46
  47 void ext2_error(struct super_block *sb, const char *function,
  48                 const char *fmt, ...)
  49 {
  50         struct va_format vaf;
  51         va_list args;
  52         struct ext2_sb_info *sbi = EXT2_SB(sb);
  53         struct ext2_super_block *es = sbi->s_es;
  54
  55         if (!(sb->s_flags & MS_RDONLY)) {
  56                 spin_lock(&sbi->s_lock);
  57                 sbi->s_mount_state |= EXT2_ERROR_FS;
  58                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  59                 spin_unlock(&sbi->s_lock);
  60                 ext2_sync_super(sb, es, 1);
  61         }
  62
  63         va_start(args, fmt);
  64
  65         vaf.fmt = fmt;
  66         vaf.va = &args;
  67
  68         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  69                sb->s_id, function, &vaf);
  70
  71         va_end(args);
  72
  73         if (test_opt(sb, ERRORS_PANIC))
  74                 panic("EXT2-fs: panic from previous error\n");
  75         if (test_opt(sb, ERRORS_RO)) {
  76                 ext2_msg(sb, KERN_CRIT,
  77                              "error: remounting filesystem read-only");
  78                 sb->s_flags |= MS_RDONLY;
  79         }
  80 }
  81
  82 void ext2_msg(struct super_block *sb, const char *prefix,
  83                 const char *fmt, ...)
  84 {
  85         struct va_format vaf;
  86         va_list args;
  87
  88         va_start(args, fmt);
  89
  90         vaf.fmt = fmt;
  91         vaf.va = &args;
  92
  93         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  94
  95         va_end(args);
  96 }
  97
  98 /*
  99  * This must be called with sbi->s_lock held.
 100  */
 101 void ext2_update_dynamic_rev(struct super_block *sb)
 102 {
 103         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 104
 105         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 106                 return;
 107
 108         ext2_msg(sb, KERN_WARNING,
 109                      "warning: updating to rev %d because of "
 110                      "new feature flag, running e2fsck is recommended",
 111                      EXT2_DYNAMIC_REV);
 112
 113         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 114         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 115         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 116         /* leave es->s_feature_*compat flags alone */
 117         /* es->s_uuid will be set by e2fsck if empty */
 118
 119         /*
 120          * The rest of the superblock fields should be zero, and if not it
 121          * means they are likely already in use, so leave them alone.  We
 122          * can leave it up to e2fsck to clean up any inconsistencies there.
 123          */
 124 }
 125
 126 #ifdef CONFIG_QUOTA
 127 static int ext2_quota_off(struct super_block *sb, int type);
 128
 129 static void ext2_quota_off_umount(struct super_block *sb)
 130 {
 131         int type;
 132
 133         for (type = 0; type < MAXQUOTAS; type++)
 134                 ext2_quota_off(sb, type);
 135 }
 136 #else
 137 static inline void ext2_quota_off_umount(struct super_block *sb)
 138 {
 139 }
 140 #endif
 141
 142 static void ext2_put_super (struct super_block * sb)
 143 {
 144         int db_count;
 145         int i;
 146         struct ext2_sb_info *sbi = EXT2_SB(sb);
 147
 148         ext2_quota_off_umount(sb);
 149
 150         if (sbi->s_ea_block_cache) {
 151                 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
 152                 sbi->s_ea_block_cache = NULL;
 153         }
 154         if (!(sb->s_flags & MS_RDONLY)) {
 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         ei->vfs_inode.i_version = 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("ext2_inode_cache",
 224                                              sizeof(struct ext2_inode_info),
 225                                              0, (SLAB_RECLAIM_ACCOUNT|
 226                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
 227                                              init_once);
 228         if (ext2_inode_cachep == NULL)
 229                 return -ENOMEM;
 230         return 0;
 231 }
 232
 233 static void destroy_inodecache(void)
 234 {
 235         /*
 236          * Make sure all delayed rcu free inodes are flushed before we
 237          * destroy cache.
 238          */
 239         rcu_barrier();
 240         kmem_cache_destroy(ext2_inode_cachep);
 241 }
 242
 243 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 244 {
 245         struct super_block *sb = root->d_sb;
 246         struct ext2_sb_info *sbi = EXT2_SB(sb);
 247         struct ext2_super_block *es = sbi->s_es;
 248         unsigned long def_mount_opts;
 249
 250         spin_lock(&sbi->s_lock);
 251         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 252
 253         if (sbi->s_sb_block != 1)
 254                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 255         if (test_opt(sb, MINIX_DF))
 256                 seq_puts(seq, ",minixdf");
 257         if (test_opt(sb, GRPID))
 258                 seq_puts(seq, ",grpid");
 259         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 260                 seq_puts(seq, ",nogrpid");
 261         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 262             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 263                 seq_printf(seq, ",resuid=%u",
 264                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 265         }
 266         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 267             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 268                 seq_printf(seq, ",resgid=%u",
 269                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 270         }
 271         if (test_opt(sb, ERRORS_RO)) {
 272                 int def_errors = le16_to_cpu(es->s_errors);
 273
 274                 if (def_errors == EXT2_ERRORS_PANIC ||
 275                     def_errors == EXT2_ERRORS_CONTINUE) {
 276                         seq_puts(seq, ",errors=remount-ro");
 277                 }
 278         }
 279         if (test_opt(sb, ERRORS_CONT))
 280                 seq_puts(seq, ",errors=continue");
 281         if (test_opt(sb, ERRORS_PANIC))
 282                 seq_puts(seq, ",errors=panic");
 283         if (test_opt(sb, NO_UID32))
 284                 seq_puts(seq, ",nouid32");
 285         if (test_opt(sb, DEBUG))
 286                 seq_puts(seq, ",debug");
 287         if (test_opt(sb, OLDALLOC))
 288                 seq_puts(seq, ",oldalloc");
 289
 290 #ifdef CONFIG_EXT2_FS_XATTR
 291         if (test_opt(sb, XATTR_USER))
 292                 seq_puts(seq, ",user_xattr");
 293         if (!test_opt(sb, XATTR_USER) &&
 294             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 295                 seq_puts(seq, ",nouser_xattr");
 296         }
 297 #endif
 298
 299 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 300         if (test_opt(sb, POSIX_ACL))
 301                 seq_puts(seq, ",acl");
 302         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 303                 seq_puts(seq, ",noacl");
 304 #endif
 305
 306         if (test_opt(sb, NOBH))
 307                 seq_puts(seq, ",nobh");
 308
 309 #if defined(CONFIG_QUOTA)
 310         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 311                 seq_puts(seq, ",usrquota");
 312
 313         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 314                 seq_puts(seq, ",grpquota");
 315 #endif
 316
 317 #ifdef CONFIG_FS_DAX
 318         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 319                 seq_puts(seq, ",xip");
 320         if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
 321                 seq_puts(seq, ",dax");
 322 #endif
 323
 324         if (!test_opt(sb, RESERVATION))
 325                 seq_puts(seq, ",noreservation");
 326
 327         spin_unlock(&sbi->s_lock);
 328         return 0;
 329 }
 330
 331 #ifdef CONFIG_QUOTA
 332 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 333 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 334 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
 335                          const struct path *path);
 336 static struct dquot **ext2_get_dquots(struct inode *inode)
 337 {
 338         return EXT2_I(inode)->i_dquot;
 339 }
 340
 341 static const struct quotactl_ops ext2_quotactl_ops = {
 342         .quota_on       = ext2_quota_on,
 343         .quota_off      = ext2_quota_off,
 344         .quota_sync     = dquot_quota_sync,
 345         .get_state      = dquot_get_state,
 346         .set_info       = dquot_set_dqinfo,
 347         .get_dqblk      = dquot_get_dqblk,
 348         .set_dqblk      = dquot_set_dqblk,
 349         .get_nextdqblk  = dquot_get_next_dqblk,
 350 };
 351 #endif
 352
 353 static const struct super_operations ext2_sops = {
 354         .alloc_inode    = ext2_alloc_inode,
 355         .destroy_inode  = ext2_destroy_inode,
 356         .write_inode    = ext2_write_inode,
 357         .evict_inode    = ext2_evict_inode,
 358         .put_super      = ext2_put_super,
 359         .sync_fs        = ext2_sync_fs,
 360         .freeze_fs      = ext2_freeze,
 361         .unfreeze_fs    = ext2_unfreeze,
 362         .statfs         = ext2_statfs,
 363         .remount_fs     = ext2_remount,
 364         .show_options   = ext2_show_options,
 365 #ifdef CONFIG_QUOTA
 366         .quota_read     = ext2_quota_read,
 367         .quota_write    = ext2_quota_write,
 368         .get_dquots     = ext2_get_dquots,
 369 #endif
 370 };
 371
 372 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 373                 u64 ino, u32 generation)
 374 {
 375         struct inode *inode;
 376
 377         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 378                 return ERR_PTR(-ESTALE);
 379         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 380                 return ERR_PTR(-ESTALE);
 381
 382         /*
 383          * ext2_iget isn't quite right if the inode is currently unallocated!
 384          * However ext2_iget currently does appropriate checks to handle stale
 385          * inodes so everything is OK.
 386          */
 387         inode = ext2_iget(sb, ino);
 388         if (IS_ERR(inode))
 389                 return ERR_CAST(inode);
 390         if (generation && inode->i_generation != generation) {
 391                 /* we didn't find the right inode.. */
 392                 iput(inode);
 393                 return ERR_PTR(-ESTALE);
 394         }
 395         return inode;
 396 }
 397
 398 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 399                 int fh_len, int fh_type)
 400 {
 401         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 402                                     ext2_nfs_get_inode);
 403 }
 404
 405 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 406                 int fh_len, int fh_type)
 407 {
 408         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 409                                     ext2_nfs_get_inode);
 410 }
 411
 412 static const struct export_operations ext2_export_ops = {
 413         .fh_to_dentry = ext2_fh_to_dentry,
 414         .fh_to_parent = ext2_fh_to_parent,
 415         .get_parent = ext2_get_parent,
 416 };
 417
 418 static unsigned long get_sb_block(void **data)
 419 {
 420         unsigned long   sb_block;
 421         char            *options = (char *) *data;
 422
 423         if (!options || strncmp(options, "sb=", 3) != 0)
 424                 return 1;       /* Default location */
 425         options += 3;
 426         sb_block = simple_strtoul(options, &options, 0);
 427         if (*options && *options != ',') {
 428                 printk("EXT2-fs: Invalid sb specification: %s\n",
 429                        (char *) *data);
 430                 return 1;
 431         }
 432         if (*options == ',')
 433                 options++;
 434         *data = (void *) options;
 435         return sb_block;
 436 }
 437
 438 enum {
 439         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 440         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 441         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 442         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 443         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 444         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 445 };
 446
 447 static const match_table_t tokens = {
 448         {Opt_bsd_df, "bsddf"},
 449         {Opt_minix_df, "minixdf"},
 450         {Opt_grpid, "grpid"},
 451         {Opt_grpid, "bsdgroups"},
 452         {Opt_nogrpid, "nogrpid"},
 453         {Opt_nogrpid, "sysvgroups"},
 454         {Opt_resgid, "resgid=%u"},
 455         {Opt_resuid, "resuid=%u"},
 456         {Opt_sb, "sb=%u"},
 457         {Opt_err_cont, "errors=continue"},
 458         {Opt_err_panic, "errors=panic"},
 459         {Opt_err_ro, "errors=remount-ro"},
 460         {Opt_nouid32, "nouid32"},
 461         {Opt_nocheck, "check=none"},
 462         {Opt_nocheck, "nocheck"},
 463         {Opt_debug, "debug"},
 464         {Opt_oldalloc, "oldalloc"},
 465         {Opt_orlov, "orlov"},
 466         {Opt_nobh, "nobh"},
 467         {Opt_user_xattr, "user_xattr"},
 468         {Opt_nouser_xattr, "nouser_xattr"},
 469         {Opt_acl, "acl"},
 470         {Opt_noacl, "noacl"},
 471         {Opt_xip, "xip"},
 472         {Opt_dax, "dax"},
 473         {Opt_grpquota, "grpquota"},
 474         {Opt_ignore, "noquota"},
 475         {Opt_quota, "quota"},
 476         {Opt_usrquota, "usrquota"},
 477         {Opt_reservation, "reservation"},
 478         {Opt_noreservation, "noreservation"},
 479         {Opt_err, NULL}
 480 };
 481
 482 static int parse_options(char *options, struct super_block *sb)
 483 {
 484         char *p;
 485         struct ext2_sb_info *sbi = EXT2_SB(sb);
 486         substring_t args[MAX_OPT_ARGS];
 487         int option;
 488         kuid_t uid;
 489         kgid_t gid;
 490
 491         if (!options)
 492                 return 1;
 493
 494         while ((p = strsep (&options, ",")) != NULL) {
 495                 int token;
 496                 if (!*p)
 497                         continue;
 498
 499                 token = match_token(p, tokens, args);
 500                 switch (token) {
 501                 case Opt_bsd_df:
 502                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 503                         break;
 504                 case Opt_minix_df:
 505                         set_opt (sbi->s_mount_opt, MINIX_DF);
 506                         break;
 507                 case Opt_grpid:
 508                         set_opt (sbi->s_mount_opt, GRPID);
 509                         break;
 510                 case Opt_nogrpid:
 511                         clear_opt (sbi->s_mount_opt, GRPID);
 512                         break;
 513                 case Opt_resuid:
 514                         if (match_int(&args[0], &option))
 515                                 return 0;
 516                         uid = make_kuid(current_user_ns(), option);
 517                         if (!uid_valid(uid)) {
 518                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 519                                 return 0;
 520
 521                         }
 522                         sbi->s_resuid = uid;
 523                         break;
 524                 case Opt_resgid:
 525                         if (match_int(&args[0], &option))
 526                                 return 0;
 527                         gid = make_kgid(current_user_ns(), option);
 528                         if (!gid_valid(gid)) {
 529                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 530                                 return 0;
 531                         }
 532                         sbi->s_resgid = gid;
 533                         break;
 534                 case Opt_sb:
 535                         /* handled by get_sb_block() instead of here */
 536                         /* *sb_block = match_int(&args[0]); */
 537                         break;
 538                 case Opt_err_panic:
 539                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 540                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 541                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 542                         break;
 543                 case Opt_err_ro:
 544                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 545                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 546                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 547                         break;
 548                 case Opt_err_cont:
 549                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 550                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 551                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 552                         break;
 553                 case Opt_nouid32:
 554                         set_opt (sbi->s_mount_opt, NO_UID32);
 555                         break;
 556                 case Opt_nocheck:
 557                         clear_opt (sbi->s_mount_opt, CHECK);
 558                         break;
 559                 case Opt_debug:
 560                         set_opt (sbi->s_mount_opt, DEBUG);
 561                         break;
 562                 case Opt_oldalloc:
 563                         set_opt (sbi->s_mount_opt, OLDALLOC);
 564                         break;
 565                 case Opt_orlov:
 566                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 567                         break;
 568                 case Opt_nobh:
 569                         set_opt (sbi->s_mount_opt, NOBH);
 570                         break;
 571 #ifdef CONFIG_EXT2_FS_XATTR
 572                 case Opt_user_xattr:
 573                         set_opt (sbi->s_mount_opt, XATTR_USER);
 574                         break;
 575                 case Opt_nouser_xattr:
 576                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 577                         break;
 578 #else
 579                 case Opt_user_xattr:
 580                 case Opt_nouser_xattr:
 581                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 582                                 "not supported");
 583                         break;
 584 #endif
 585 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 586                 case Opt_acl:
 587                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 588                         break;
 589                 case Opt_noacl:
 590                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 591                         break;
 592 #else
 593                 case Opt_acl:
 594                 case Opt_noacl:
 595                         ext2_msg(sb, KERN_INFO,
 596                                 "(no)acl options not supported");
 597                         break;
 598 #endif
 599                 case Opt_xip:
 600                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 601                         set_opt(sbi->s_mount_opt, XIP);
 602                         /* Fall through */
 603                 case Opt_dax:
 604 #ifdef CONFIG_FS_DAX
 605                         ext2_msg(sb, KERN_WARNING,
 606                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 607                         set_opt(sbi->s_mount_opt, DAX);
 608 #else
 609                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 610 #endif
 611                         break;
 612
 613 #if defined(CONFIG_QUOTA)
 614                 case Opt_quota:
 615                 case Opt_usrquota:
 616                         set_opt(sbi->s_mount_opt, USRQUOTA);
 617                         break;
 618
 619                 case Opt_grpquota:
 620                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 621                         break;
 622 #else
 623                 case Opt_quota:
 624                 case Opt_usrquota:
 625                 case Opt_grpquota:
 626                         ext2_msg(sb, KERN_INFO,
 627                                 "quota operations not supported");
 628                         break;
 629 #endif
 630
 631                 case Opt_reservation:
 632                         set_opt(sbi->s_mount_opt, RESERVATION);
 633                         ext2_msg(sb, KERN_INFO, "reservations ON");
 634                         break;
 635                 case Opt_noreservation:
 636                         clear_opt(sbi->s_mount_opt, RESERVATION);
 637                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 638                         break;
 639                 case Opt_ignore:
 640                         break;
 641                 default:
 642                         return 0;
 643                 }
 644         }
 645         return 1;
 646 }
 647
 648 static int ext2_setup_super (struct super_block * sb,
 649                               struct ext2_super_block * es,
 650                               int read_only)
 651 {
 652         int res = 0;
 653         struct ext2_sb_info *sbi = EXT2_SB(sb);
 654
 655         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 656                 ext2_msg(sb, KERN_ERR,
 657                         "error: revision level too high, "
 658                         "forcing read-only mode");
 659                 res = MS_RDONLY;
 660         }
 661         if (read_only)
 662                 return res;
 663         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 664                 ext2_msg(sb, KERN_WARNING,
 665                         "warning: mounting unchecked fs, "
 666                         "running e2fsck is recommended");
 667         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 668                 ext2_msg(sb, KERN_WARNING,
 669                         "warning: mounting fs with errors, "
 670                         "running e2fsck is recommended");
 671         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 672                  le16_to_cpu(es->s_mnt_count) >=
 673                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 674                 ext2_msg(sb, KERN_WARNING,
 675                         "warning: maximal mount count reached, "
 676                         "running e2fsck is recommended");
 677         else if (le32_to_cpu(es->s_checkinterval) &&
 678                 (le32_to_cpu(es->s_lastcheck) +
 679                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 680                 ext2_msg(sb, KERN_WARNING,
 681                         "warning: checktime reached, "
 682                         "running e2fsck is recommended");
 683         if (!le16_to_cpu(es->s_max_mnt_count))
 684                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 685         le16_add_cpu(&es->s_mnt_count, 1);
 686         if (test_opt (sb, DEBUG))
 687                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 688                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 689                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 690                         sbi->s_frag_size,
 691                         sbi->s_groups_count,
 692                         EXT2_BLOCKS_PER_GROUP(sb),
 693                         EXT2_INODES_PER_GROUP(sb),
 694                         sbi->s_mount_opt);
 695         return res;
 696 }
 697
 698 static int ext2_check_descriptors(struct super_block *sb)
 699 {
 700         int i;
 701         struct ext2_sb_info *sbi = EXT2_SB(sb);
 702
 703         ext2_debug ("Checking group descriptors");
 704
 705         for (i = 0; i < sbi->s_groups_count; i++) {
 706                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 707                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 708                 ext2_fsblk_t last_block;
 709
 710                 if (i == sbi->s_groups_count - 1)
 711                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 712                 else
 713                         last_block = first_block +
 714                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 715
 716                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 717                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 718                 {
 719                         ext2_error (sb, "ext2_check_descriptors",
 720                                     "Block bitmap for group %d"
 721                                     " not in group (block %lu)!",
 722                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 723                         return 0;
 724                 }
 725                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 726                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 727                 {
 728                         ext2_error (sb, "ext2_check_descriptors",
 729                                     "Inode bitmap for group %d"
 730                                     " not in group (block %lu)!",
 731                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 732                         return 0;
 733                 }
 734                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 735                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 736                     last_block)
 737                 {
 738                         ext2_error (sb, "ext2_check_descriptors",
 739                                     "Inode table for group %d"
 740                                     " not in group (block %lu)!",
 741                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 742                         return 0;
 743                 }
 744         }
 745         return 1;
 746 }
 747
 748 /*
 749  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 750  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 751  * We need to be 1 filesystem block less than the 2^32 sector limit.
 752  */
 753 static loff_t ext2_max_size(int bits)
 754 {
 755         loff_t res = EXT2_NDIR_BLOCKS;
 756         int meta_blocks;
 757         loff_t upper_limit;
 758
 759         /* This is calculated to be the largest file size for a
 760          * dense, file such that the total number of
 761          * sectors in the file, including data and all indirect blocks,
 762          * does not exceed 2^32 -1
 763          * __u32 i_blocks representing the total number of
 764          * 512 bytes blocks of the file
 765          */
 766         upper_limit = (1LL << 32) - 1;
 767
 768         /* total blocks in file system block size */
 769         upper_limit >>= (bits - 9);
 770
 771
 772         /* indirect blocks */
 773         meta_blocks = 1;
 774         /* double indirect blocks */
 775         meta_blocks += 1 + (1LL << (bits-2));
 776         /* tripple indirect blocks */
 777         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 778
 779         upper_limit -= meta_blocks;
 780         upper_limit <<= bits;
 781
 782         res += 1LL << (bits-2);
 783         res += 1LL << (2*(bits-2));
 784         res += 1LL << (3*(bits-2));
 785         res <<= bits;
 786         if (res > upper_limit)
 787                 res = upper_limit;
 788
 789         if (res > MAX_LFS_FILESIZE)
 790                 res = MAX_LFS_FILESIZE;
 791
 792         return res;
 793 }
 794
 795 static unsigned long descriptor_loc(struct super_block *sb,
 796                                     unsigned long logic_sb_block,
 797                                     int nr)
 798 {
 799         struct ext2_sb_info *sbi = EXT2_SB(sb);
 800         unsigned long bg, first_meta_bg;
 801         int has_super = 0;
 802
 803         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 804
 805         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 806             nr < first_meta_bg)
 807                 return (logic_sb_block + nr + 1);
 808         bg = sbi->s_desc_per_block * nr;
 809         if (ext2_bg_has_super(sb, bg))
 810                 has_super = 1;
 811
 812         return ext2_group_first_block_no(sb, bg) + has_super;
 813 }
 814
 815 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 816 {
 817         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 818         struct buffer_head * bh;
 819         struct ext2_sb_info * sbi;
 820         struct ext2_super_block * es;
 821         struct inode *root;
 822         unsigned long block;
 823         unsigned long sb_block = get_sb_block(&data);
 824         unsigned long logic_sb_block;
 825         unsigned long offset = 0;
 826         unsigned long def_mount_opts;
 827         long ret = -EINVAL;
 828         int blocksize = BLOCK_SIZE;
 829         int db_count;
 830         int i, j;
 831         __le32 features;
 832         int err;
 833
 834         err = -ENOMEM;
 835         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 836         if (!sbi)
 837                 goto failed;
 838
 839         sbi->s_blockgroup_lock =
 840                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 841         if (!sbi->s_blockgroup_lock) {
 842                 kfree(sbi);
 843                 goto failed;
 844         }
 845         sb->s_fs_info = sbi;
 846         sbi->s_sb_block = sb_block;
 847         sbi->s_daxdev = dax_dev;
 848
 849         spin_lock_init(&sbi->s_lock);
 850
 851         /*
 852          * See what the current blocksize for the device is, and
 853          * use that as the blocksize.  Otherwise (or if the blocksize
 854          * is smaller than the default) use the default.
 855          * This is important for devices that have a hardware
 856          * sectorsize that is larger than the default.
 857          */
 858         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 859         if (!blocksize) {
 860                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 861                 goto failed_sbi;
 862         }
 863
 864         /*
 865          * If the superblock doesn't start on a hardware sector boundary,
 866          * calculate the offset.
 867          */
 868         if (blocksize != BLOCK_SIZE) {
 869                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 870                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 871         } else {
 872                 logic_sb_block = sb_block;
 873         }
 874
 875         if (!(bh = sb_bread(sb, logic_sb_block))) {
 876                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 877                 goto failed_sbi;
 878         }
 879         /*
 880          * Note: s_es must be initialized as soon as possible because
 881          *       some ext2 macro-instructions depend on its value
 882          */
 883         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 884         sbi->s_es = es;
 885         sb->s_magic = le16_to_cpu(es->s_magic);
 886
 887         if (sb->s_magic != EXT2_SUPER_MAGIC)
 888                 goto cantfind_ext2;
 889
 890         /* Set defaults before we parse the mount options */
 891         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 892         if (def_mount_opts & EXT2_DEFM_DEBUG)
 893                 set_opt(sbi->s_mount_opt, DEBUG);
 894         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 895                 set_opt(sbi->s_mount_opt, GRPID);
 896         if (def_mount_opts & EXT2_DEFM_UID16)
 897                 set_opt(sbi->s_mount_opt, NO_UID32);
 898 #ifdef CONFIG_EXT2_FS_XATTR
 899         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 900                 set_opt(sbi->s_mount_opt, XATTR_USER);
 901 #endif
 902 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 903         if (def_mount_opts & EXT2_DEFM_ACL)
 904                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 905 #endif
 906
 907         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 908                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 909         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 910                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 911         else
 912                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 913
 914         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 915         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 916
 917         set_opt(sbi->s_mount_opt, RESERVATION);
 918
 919         if (!parse_options((char *) data, sb))
 920                 goto failed_mount;
 921
 922         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 923                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 924                  MS_POSIXACL : 0);
 925         sb->s_iflags |= SB_I_CGROUPWB;
 926
 927         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 928             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 929              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 930              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 931                 ext2_msg(sb, KERN_WARNING,
 932                         "warning: feature flags set on rev 0 fs, "
 933                         "running e2fsck is recommended");
 934         /*
 935          * Check feature flags regardless of the revision level, since we
 936          * previously didn't change the revision level when setting the flags,
 937          * so there is a chance incompat flags are set on a rev 0 filesystem.
 938          */
 939         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 940         if (features) {
 941                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 942                        "unsupported optional features (%x)",
 943                         le32_to_cpu(features));
 944                 goto failed_mount;
 945         }
 946         if (!(sb->s_flags & MS_RDONLY) &&
 947             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 948                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 949                        "unsupported optional features (%x)",
 950                        le32_to_cpu(features));
 951                 goto failed_mount;
 952         }
 953
 954         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 955
 956         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
 957                 err = bdev_dax_supported(sb, blocksize);
 958                 if (err)
 959                         goto failed_mount;
 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
 991         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 992                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 993                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 994         } else {
 995                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 996                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 997                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 998                     !is_power_of_2(sbi->s_inode_size) ||
 999                     (sbi->s_inode_size > blocksize)) {
1000                         ext2_msg(sb, KERN_ERR,
1001                                 "error: unsupported inode size: %d",
1002                                 sbi->s_inode_size);
1003                         goto failed_mount;
1004                 }
1005         }
1006
1007         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1008                                    le32_to_cpu(es->s_log_frag_size);
1009         if (sbi->s_frag_size == 0)
1010                 goto cantfind_ext2;
1011         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1012
1013         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1014         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1015         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1016
1017         if (EXT2_INODE_SIZE(sb) == 0)
1018                 goto cantfind_ext2;
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 (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1077         if (sbi->s_group_desc == NULL) {
1078                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1079                 goto failed_mount;
1080         }
1081         bgl_lock_init(sbi->s_blockgroup_lock);
1082         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1083         if (!sbi->s_debts) {
1084                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1085                 goto failed_mount_group_desc;
1086         }
1087         for (i = 0; i < db_count; i++) {
1088                 block = descriptor_loc(sb, logic_sb_block, i);
1089                 sbi->s_group_desc[i] = sb_bread(sb, block);
1090                 if (!sbi->s_group_desc[i]) {
1091                         for (j = 0; j < i; j++)
1092                                 brelse (sbi->s_group_desc[j]);
1093                         ext2_msg(sb, KERN_ERR,
1094                                 "error: unable to read group descriptors");
1095                         goto failed_mount_group_desc;
1096                 }
1097         }
1098         if (!ext2_check_descriptors (sb)) {
1099                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1100                 goto failed_mount2;
1101         }
1102         sbi->s_gdb_count = db_count;
1103         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1104         spin_lock_init(&sbi->s_next_gen_lock);
1105
1106         /* per fileystem reservation list head & lock */
1107         spin_lock_init(&sbi->s_rsv_window_lock);
1108         sbi->s_rsv_window_root = RB_ROOT;
1109         /*
1110          * Add a single, static dummy reservation to the start of the
1111          * reservation window list --- it gives us a placeholder for
1112          * append-at-start-of-list which makes the allocation logic
1113          * _much_ simpler.
1114          */
1115         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1116         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1117         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1118         sbi->s_rsv_window_head.rsv_goal_size = 0;
1119         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1120
1121         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1122                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1123         if (!err) {
1124                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1125                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1126         }
1127         if (!err) {
1128                 err = percpu_counter_init(&sbi->s_dirs_counter,
1129                                 ext2_count_dirs(sb), GFP_KERNEL);
1130         }
1131         if (err) {
1132                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1133                 goto failed_mount3;
1134         }
1135
1136 #ifdef CONFIG_EXT2_FS_XATTR
1137         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1138         if (!sbi->s_ea_block_cache) {
1139                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1140                 goto failed_mount3;
1141         }
1142 #endif
1143         /*
1144          * set up enough so that it can read an inode
1145          */
1146         sb->s_op = &ext2_sops;
1147         sb->s_export_op = &ext2_export_ops;
1148         sb->s_xattr = ext2_xattr_handlers;
1149
1150 #ifdef CONFIG_QUOTA
1151         sb->dq_op = &dquot_operations;
1152         sb->s_qcop = &ext2_quotactl_ops;
1153         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1154 #endif
1155
1156         root = ext2_iget(sb, EXT2_ROOT_INO);
1157         if (IS_ERR(root)) {
1158                 ret = PTR_ERR(root);
1159                 goto failed_mount3;
1160         }
1161         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1162                 iput(root);
1163                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1164                 goto failed_mount3;
1165         }
1166
1167         sb->s_root = d_make_root(root);
1168         if (!sb->s_root) {
1169                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1170                 ret = -ENOMEM;
1171                 goto failed_mount3;
1172         }
1173         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1174                 ext2_msg(sb, KERN_WARNING,
1175                         "warning: mounting ext3 filesystem as ext2");
1176         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1177                 sb->s_flags |= MS_RDONLY;
1178         ext2_write_super(sb);
1179         return 0;
1180
1181 cantfind_ext2:
1182         if (!silent)
1183                 ext2_msg(sb, KERN_ERR,
1184                         "error: can't find an ext2 filesystem on dev %s.",
1185                         sb->s_id);
1186         goto failed_mount;
1187 failed_mount3:
1188         if (sbi->s_ea_block_cache)
1189                 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1190         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1191         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1192         percpu_counter_destroy(&sbi->s_dirs_counter);
1193 failed_mount2:
1194         for (i = 0; i < db_count; i++)
1195                 brelse(sbi->s_group_desc[i]);
1196 failed_mount_group_desc:
1197         kfree(sbi->s_group_desc);
1198         kfree(sbi->s_debts);
1199 failed_mount:
1200         brelse(bh);
1201 failed_sbi:
1202         sb->s_fs_info = NULL;
1203         kfree(sbi->s_blockgroup_lock);
1204         kfree(sbi);
1205 failed:
1206         fs_put_dax(dax_dev);
1207         return ret;
1208 }
1209
1210 static void ext2_clear_super_error(struct super_block *sb)
1211 {
1212         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1213
1214         if (buffer_write_io_error(sbh)) {
1215                 /*
1216                  * Oh, dear.  A previous attempt to write the
1217                  * superblock failed.  This could happen because the
1218                  * USB device was yanked out.  Or it could happen to
1219                  * be a transient write error and maybe the block will
1220                  * be remapped.  Nothing we can do but to retry the
1221                  * write and hope for the best.
1222                  */
1223                 ext2_msg(sb, KERN_ERR,
1224                        "previous I/O error to superblock detected\n");
1225                 clear_buffer_write_io_error(sbh);
1226                 set_buffer_uptodate(sbh);
1227         }
1228 }
1229
1230 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1231                      int wait)
1232 {
1233         ext2_clear_super_error(sb);
1234         spin_lock(&EXT2_SB(sb)->s_lock);
1235         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1236         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1237         es->s_wtime = cpu_to_le32(get_seconds());
1238         /* unlock before we do IO */
1239         spin_unlock(&EXT2_SB(sb)->s_lock);
1240         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1241         if (wait)
1242                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1243 }
1244
1245 /*
1246  * In the second extended file system, it is not necessary to
1247  * write the super block since we use a mapping of the
1248  * disk super block in a buffer.
1249  *
1250  * However, this function is still used to set the fs valid
1251  * flags to 0.  We need to set this flag to 0 since the fs
1252  * may have been checked while mounted and e2fsck may have
1253  * set s_state to EXT2_VALID_FS after some corrections.
1254  */
1255 static int ext2_sync_fs(struct super_block *sb, int wait)
1256 {
1257         struct ext2_sb_info *sbi = EXT2_SB(sb);
1258         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1259
1260         /*
1261          * Write quota structures to quota file, sync_blockdev() will write
1262          * them to disk later
1263          */
1264         dquot_writeback_dquots(sb, -1);
1265
1266         spin_lock(&sbi->s_lock);
1267         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1268                 ext2_debug("setting valid to 0\n");
1269                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1270         }
1271         spin_unlock(&sbi->s_lock);
1272         ext2_sync_super(sb, es, wait);
1273         return 0;
1274 }
1275
1276 static int ext2_freeze(struct super_block *sb)
1277 {
1278         struct ext2_sb_info *sbi = EXT2_SB(sb);
1279
1280         /*
1281          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1282          * because we have unattached inodes and thus filesystem is not fully
1283          * consistent.
1284          */
1285         if (atomic_long_read(&sb->s_remove_count)) {
1286                 ext2_sync_fs(sb, 1);
1287                 return 0;
1288         }
1289         /* Set EXT2_FS_VALID flag */
1290         spin_lock(&sbi->s_lock);
1291         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1292         spin_unlock(&sbi->s_lock);
1293         ext2_sync_super(sb, sbi->s_es, 1);
1294
1295         return 0;
1296 }
1297
1298 static int ext2_unfreeze(struct super_block *sb)
1299 {
1300         /* Just write sb to clear EXT2_VALID_FS flag */
1301         ext2_write_super(sb);
1302
1303         return 0;
1304 }
1305
1306 static void ext2_write_super(struct super_block *sb)
1307 {
1308         if (!(sb->s_flags & MS_RDONLY))
1309                 ext2_sync_fs(sb, 1);
1310 }
1311
1312 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1313 {
1314         struct ext2_sb_info * sbi = EXT2_SB(sb);
1315         struct ext2_super_block * es;
1316         struct ext2_mount_options old_opts;
1317         unsigned long old_sb_flags;
1318         int err;
1319
1320         sync_filesystem(sb);
1321         spin_lock(&sbi->s_lock);
1322
1323         /* Store the old options */
1324         old_sb_flags = sb->s_flags;
1325         old_opts.s_mount_opt = sbi->s_mount_opt;
1326         old_opts.s_resuid = sbi->s_resuid;
1327         old_opts.s_resgid = sbi->s_resgid;
1328
1329         /*
1330          * Allow the "check" option to be passed as a remount option.
1331          */
1332         if (!parse_options(data, sb)) {
1333                 err = -EINVAL;
1334                 goto restore_opts;
1335         }
1336
1337         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1338                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1339
1340         es = sbi->s_es;
1341         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1342                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1343                          "dax flag with busy inodes while remounting");
1344                 sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1345         }
1346         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1347                 spin_unlock(&sbi->s_lock);
1348                 return 0;
1349         }
1350         if (*flags & MS_RDONLY) {
1351                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1352                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1353                         spin_unlock(&sbi->s_lock);
1354                         return 0;
1355                 }
1356
1357                 /*
1358                  * OK, we are remounting a valid rw partition rdonly, so set
1359                  * the rdonly flag and then mark the partition as valid again.
1360                  */
1361                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1362                 es->s_mtime = cpu_to_le32(get_seconds());
1363                 spin_unlock(&sbi->s_lock);
1364
1365                 err = dquot_suspend(sb, -1);
1366                 if (err < 0) {
1367                         spin_lock(&sbi->s_lock);
1368                         goto restore_opts;
1369                 }
1370
1371                 ext2_sync_super(sb, es, 1);
1372         } else {
1373                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1374                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1375                 if (ret) {
1376                         ext2_msg(sb, KERN_WARNING,
1377                                 "warning: couldn't remount RDWR because of "
1378                                 "unsupported optional features (%x).",
1379                                 le32_to_cpu(ret));
1380                         err = -EROFS;
1381                         goto restore_opts;
1382                 }
1383                 /*
1384                  * Mounting a RDONLY partition read-write, so reread and
1385                  * store the current valid flag.  (It may have been changed
1386                  * by e2fsck since we originally mounted the partition.)
1387                  */
1388                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1389                 if (!ext2_setup_super (sb, es, 0))
1390                         sb->s_flags &= ~MS_RDONLY;
1391                 spin_unlock(&sbi->s_lock);
1392
1393                 ext2_write_super(sb);
1394
1395                 dquot_resume(sb, -1);
1396         }
1397
1398         return 0;
1399 restore_opts:
1400         sbi->s_mount_opt = old_opts.s_mount_opt;
1401         sbi->s_resuid = old_opts.s_resuid;
1402         sbi->s_resgid = old_opts.s_resgid;
1403         sb->s_flags = old_sb_flags;
1404         spin_unlock(&sbi->s_lock);
1405         return err;
1406 }
1407
1408 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1409 {
1410         struct super_block *sb = dentry->d_sb;
1411         struct ext2_sb_info *sbi = EXT2_SB(sb);
1412         struct ext2_super_block *es = sbi->s_es;
1413         u64 fsid;
1414
1415         spin_lock(&sbi->s_lock);
1416
1417         if (test_opt (sb, MINIX_DF))
1418                 sbi->s_overhead_last = 0;
1419         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1420                 unsigned long i, overhead = 0;
1421                 smp_rmb();
1422
1423                 /*
1424                  * Compute the overhead (FS structures). This is constant
1425                  * for a given filesystem unless the number of block groups
1426                  * changes so we cache the previous value until it does.
1427                  */
1428
1429                 /*
1430                  * All of the blocks before first_data_block are
1431                  * overhead
1432                  */
1433                 overhead = le32_to_cpu(es->s_first_data_block);
1434
1435                 /*
1436                  * Add the overhead attributed to the superblock and
1437                  * block group descriptors.  If the sparse superblocks
1438                  * feature is turned on, then not all groups have this.
1439                  */
1440                 for (i = 0; i < sbi->s_groups_count; i++)
1441                         overhead += ext2_bg_has_super(sb, i) +
1442                                 ext2_bg_num_gdb(sb, i);
1443
1444                 /*
1445                  * Every block group has an inode bitmap, a block
1446                  * bitmap, and an inode table.
1447                  */
1448                 overhead += (sbi->s_groups_count *
1449                              (2 + sbi->s_itb_per_group));
1450                 sbi->s_overhead_last = overhead;
1451                 smp_wmb();
1452                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1453         }
1454
1455         buf->f_type = EXT2_SUPER_MAGIC;
1456         buf->f_bsize = sb->s_blocksize;
1457         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1458         buf->f_bfree = ext2_count_free_blocks(sb);
1459         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1460         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1461         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1462                 buf->f_bavail = 0;
1463         buf->f_files = le32_to_cpu(es->s_inodes_count);
1464         buf->f_ffree = ext2_count_free_inodes(sb);
1465         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1466         buf->f_namelen = EXT2_NAME_LEN;
1467         fsid = le64_to_cpup((void *)es->s_uuid) ^
1468                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1469         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1470         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1471         spin_unlock(&sbi->s_lock);
1472         return 0;
1473 }
1474
1475 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1476         int flags, const char *dev_name, void *data)
1477 {
1478         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1479 }
1480
1481 #ifdef CONFIG_QUOTA
1482
1483 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1484  * acquiring the locks... As quota files are never truncated and quota code
1485  * itself serializes the operations (and no one else should touch the files)
1486  * we don't have to be afraid of races */
1487 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1488                                size_t len, loff_t off)
1489 {
1490         struct inode *inode = sb_dqopt(sb)->files[type];
1491         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1492         int err = 0;
1493         int offset = off & (sb->s_blocksize - 1);
1494         int tocopy;
1495         size_t toread;
1496         struct buffer_head tmp_bh;
1497         struct buffer_head *bh;
1498         loff_t i_size = i_size_read(inode);
1499
1500         if (off > i_size)
1501                 return 0;
1502         if (off+len > i_size)
1503                 len = i_size-off;
1504         toread = len;
1505         while (toread > 0) {
1506                 tocopy = sb->s_blocksize - offset < toread ?
1507                                 sb->s_blocksize - offset : toread;
1508
1509                 tmp_bh.b_state = 0;
1510                 tmp_bh.b_size = sb->s_blocksize;
1511                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1512                 if (err < 0)
1513                         return err;
1514                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1515                         memset(data, 0, tocopy);
1516                 else {
1517                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1518                         if (!bh)
1519                                 return -EIO;
1520                         memcpy(data, bh->b_data+offset, tocopy);
1521                         brelse(bh);
1522                 }
1523                 offset = 0;
1524                 toread -= tocopy;
1525                 data += tocopy;
1526                 blk++;
1527         }
1528         return len;
1529 }
1530
1531 /* Write to quotafile */
1532 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1533                                 const char *data, size_t len, loff_t off)
1534 {
1535         struct inode *inode = sb_dqopt(sb)->files[type];
1536         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1537         int err = 0;
1538         int offset = off & (sb->s_blocksize - 1);
1539         int tocopy;
1540         size_t towrite = len;
1541         struct buffer_head tmp_bh;
1542         struct buffer_head *bh;
1543
1544         while (towrite > 0) {
1545                 tocopy = sb->s_blocksize - offset < towrite ?
1546                                 sb->s_blocksize - offset : towrite;
1547
1548                 tmp_bh.b_state = 0;
1549                 tmp_bh.b_size = sb->s_blocksize;
1550                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1551                 if (err < 0)
1552                         goto out;
1553                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1554                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1555                 else
1556                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1557                 if (unlikely(!bh)) {
1558                         err = -EIO;
1559                         goto out;
1560                 }
1561                 lock_buffer(bh);
1562                 memcpy(bh->b_data+offset, data, tocopy);
1563                 flush_dcache_page(bh->b_page);
1564                 set_buffer_uptodate(bh);
1565                 mark_buffer_dirty(bh);
1566                 unlock_buffer(bh);
1567                 brelse(bh);
1568                 offset = 0;
1569                 towrite -= tocopy;
1570                 data += tocopy;
1571                 blk++;
1572         }
1573 out:
1574         if (len == towrite)
1575                 return err;
1576         if (inode->i_size < off+len-towrite)
1577                 i_size_write(inode, off+len-towrite);
1578         inode->i_version++;
1579         inode->i_mtime = inode->i_ctime = current_time(inode);
1580         mark_inode_dirty(inode);
1581         return len - towrite;
1582 }
1583
1584 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1585                          const struct path *path)
1586 {
1587         int err;
1588         struct inode *inode;
1589
1590         err = dquot_quota_on(sb, type, format_id, path);
1591         if (err)
1592                 return err;
1593
1594         inode = d_inode(path->dentry);
1595         inode_lock(inode);
1596         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1597         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1598                         S_NOATIME | S_IMMUTABLE);
1599         inode_unlock(inode);
1600         mark_inode_dirty(inode);
1601
1602         return 0;
1603 }
1604
1605 static int ext2_quota_off(struct super_block *sb, int type)
1606 {
1607         struct inode *inode = sb_dqopt(sb)->files[type];
1608         int err;
1609
1610         if (!inode || !igrab(inode))
1611                 goto out;
1612
1613         err = dquot_quota_off(sb, type);
1614         if (err)
1615                 goto out_put;
1616
1617         inode_lock(inode);
1618         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1619         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1620         inode_unlock(inode);
1621         mark_inode_dirty(inode);
1622 out_put:
1623         iput(inode);
1624         return err;
1625 out:
1626         return dquot_quota_off(sb, type);
1627 }
1628
1629 #endif
1630
1631 static struct file_system_type ext2_fs_type = {
1632         .owner          = THIS_MODULE,
1633         .name           = "ext2",
1634         .mount          = ext2_mount,
1635         .kill_sb        = kill_block_super,
1636         .fs_flags       = FS_REQUIRES_DEV,
1637 };
1638 MODULE_ALIAS_FS("ext2");
1639
1640 static int __init init_ext2_fs(void)
1641 {
1642         int err;
1643
1644         err = init_inodecache();
1645         if (err)
1646                 return err;
1647         err = register_filesystem(&ext2_fs_type);
1648         if (err)
1649                 goto out;
1650         return 0;
1651 out:
1652         destroy_inodecache();
1653         return err;
1654 }
1655
1656 static void __exit exit_ext2_fs(void)
1657 {
1658         unregister_filesystem(&ext2_fs_type);
1659         destroy_inodecache();
1660 }
1661
1662 MODULE_AUTHOR("Remy Card and others");
1663 MODULE_DESCRIPTION("Second Extended Filesystem");
1664 MODULE_LICENSE("GPL");
1665 module_init(init_ext2_fs)
1666 module_exit(exit_ext2_fs)