fs/ext2/super.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/fs/ext2/super.c
   4  *
   5  * Copyright (C) 1992, 1993, 1994, 1995
   6  * Remy Card (card@masi.ibp.fr)
   7  * Laboratoire MASI - Institut Blaise Pascal
   8  * Universite Pierre et Marie Curie (Paris VI)
   9  *
  10  *  from
  11  *
  12  *  linux/fs/minix/inode.c
  13  *
  14  *  Copyright (C) 1991, 1992  Linus Torvalds
  15  *
  16  *  Big-endian to little-endian byte-swapping/bitmaps by
  17  *        David S. Miller (davem@caip.rutgers.edu), 1995
  18  */
  19
  20 #include <linux/module.h>
  21 #include <linux/string.h>
  22 #include <linux/fs.h>
  23 #include <linux/slab.h>
  24 #include <linux/init.h>
  25 #include <linux/blkdev.h>
  26 #include <linux/parser.h>
  27 #include <linux/random.h>
  28 #include <linux/buffer_head.h>
  29 #include <linux/exportfs.h>
  30 #include <linux/vfs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/mount.h>
  33 #include <linux/log2.h>
  34 #include <linux/quotaops.h>
  35 #include <linux/uaccess.h>
  36 #include <linux/dax.h>
  37 #include <linux/iversion.h>
  38 #include "ext2.h"
  39 #include "xattr.h"
  40 #include "acl.h"
  41
  42 static void ext2_write_super(struct super_block *sb);
  43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  45 static int ext2_sync_fs(struct super_block *sb, int wait);
  46 static int ext2_freeze(struct super_block *sb);
  47 static int ext2_unfreeze(struct super_block *sb);
  48
  49 void ext2_error(struct super_block *sb, const char *function,
  50                 const char *fmt, ...)
  51 {
  52         struct va_format vaf;
  53         va_list args;
  54         struct ext2_sb_info *sbi = EXT2_SB(sb);
  55         struct ext2_super_block *es = sbi->s_es;
  56
  57         if (!sb_rdonly(sb)) {
  58                 spin_lock(&sbi->s_lock);
  59                 sbi->s_mount_state |= EXT2_ERROR_FS;
  60                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  61                 spin_unlock(&sbi->s_lock);
  62                 ext2_sync_super(sb, es, 1);
  63         }
  64
  65         va_start(args, fmt);
  66
  67         vaf.fmt = fmt;
  68         vaf.va = &args;
  69
  70         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  71                sb->s_id, function, &vaf);
  72
  73         va_end(args);
  74
  75         if (test_opt(sb, ERRORS_PANIC))
  76                 panic("EXT2-fs: panic from previous error\n");
  77         if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
  78                 ext2_msg(sb, KERN_CRIT,
  79                              "error: remounting filesystem read-only");
  80                 sb->s_flags |= SB_RDONLY;
  81         }
  82 }
  83
  84 void ext2_msg(struct super_block *sb, const char *prefix,
  85                 const char *fmt, ...)
  86 {
  87         struct va_format vaf;
  88         va_list args;
  89
  90         va_start(args, fmt);
  91
  92         vaf.fmt = fmt;
  93         vaf.va = &args;
  94
  95         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  96
  97         va_end(args);
  98 }
  99
 100 /*
 101  * This must be called with sbi->s_lock held.
 102  */
 103 void ext2_update_dynamic_rev(struct super_block *sb)
 104 {
 105         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 106
 107         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 108                 return;
 109
 110         ext2_msg(sb, KERN_WARNING,
 111                      "warning: updating to rev %d because of "
 112                      "new feature flag, running e2fsck is recommended",
 113                      EXT2_DYNAMIC_REV);
 114
 115         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 116         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 117         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 118         /* leave es->s_feature_*compat flags alone */
 119         /* es->s_uuid will be set by e2fsck if empty */
 120
 121         /*
 122          * The rest of the superblock fields should be zero, and if not it
 123          * means they are likely already in use, so leave them alone.  We
 124          * can leave it up to e2fsck to clean up any inconsistencies there.
 125          */
 126 }
 127
 128 #ifdef CONFIG_QUOTA
 129 static int ext2_quota_off(struct super_block *sb, int type);
 130
 131 static void ext2_quota_off_umount(struct super_block *sb)
 132 {
 133         int type;
 134
 135         for (type = 0; type < MAXQUOTAS; type++)
 136                 ext2_quota_off(sb, type);
 137 }
 138 #else
 139 static inline void ext2_quota_off_umount(struct super_block *sb)
 140 {
 141 }
 142 #endif
 143
 144 static void ext2_put_super (struct super_block * sb)
 145 {
 146         int db_count;
 147         int i;
 148         struct ext2_sb_info *sbi = EXT2_SB(sb);
 149
 150         ext2_quota_off_umount(sb);
 151
 152         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
 153         sbi->s_ea_block_cache = NULL;
 154
 155         if (!sb_rdonly(sb)) {
 156                 struct ext2_super_block *es = sbi->s_es;
 157
 158                 spin_lock(&sbi->s_lock);
 159                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 160                 spin_unlock(&sbi->s_lock);
 161                 ext2_sync_super(sb, es, 1);
 162         }
 163         db_count = sbi->s_gdb_count;
 164         for (i = 0; i < db_count; i++)
 165                 brelse(sbi->s_group_desc[i]);
 166         kfree(sbi->s_group_desc);
 167         kfree(sbi->s_debts);
 168         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 169         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 170         percpu_counter_destroy(&sbi->s_dirs_counter);
 171         brelse (sbi->s_sbh);
 172         sb->s_fs_info = NULL;
 173         kfree(sbi->s_blockgroup_lock);
 174         fs_put_dax(sbi->s_daxdev);
 175         kfree(sbi);
 176 }
 177
 178 static struct kmem_cache * ext2_inode_cachep;
 179
 180 static struct inode *ext2_alloc_inode(struct super_block *sb)
 181 {
 182         struct ext2_inode_info *ei;
 183         ei = 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_free_in_core_inode(struct inode *inode)
 196 {
 197         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 198 }
 199
 200 static void init_once(void *foo)
 201 {
 202         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 203
 204         rwlock_init(&ei->i_meta_lock);
 205 #ifdef CONFIG_EXT2_FS_XATTR
 206         init_rwsem(&ei->xattr_sem);
 207 #endif
 208         mutex_init(&ei->truncate_mutex);
 209 #ifdef CONFIG_FS_DAX
 210         init_rwsem(&ei->dax_sem);
 211 #endif
 212         inode_init_once(&ei->vfs_inode);
 213 }
 214
 215 static int __init init_inodecache(void)
 216 {
 217         ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
 218                                 sizeof(struct ext2_inode_info), 0,
 219                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
 220                                         SLAB_ACCOUNT),
 221                                 offsetof(struct ext2_inode_info, i_data),
 222                                 sizeof_field(struct ext2_inode_info, i_data),
 223                                 init_once);
 224         if (ext2_inode_cachep == NULL)
 225                 return -ENOMEM;
 226         return 0;
 227 }
 228
 229 static void destroy_inodecache(void)
 230 {
 231         /*
 232          * Make sure all delayed rcu free inodes are flushed before we
 233          * destroy cache.
 234          */
 235         rcu_barrier();
 236         kmem_cache_destroy(ext2_inode_cachep);
 237 }
 238
 239 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 240 {
 241         struct super_block *sb = root->d_sb;
 242         struct ext2_sb_info *sbi = EXT2_SB(sb);
 243         struct ext2_super_block *es = sbi->s_es;
 244         unsigned long def_mount_opts;
 245
 246         spin_lock(&sbi->s_lock);
 247         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 248
 249         if (sbi->s_sb_block != 1)
 250                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 251         if (test_opt(sb, MINIX_DF))
 252                 seq_puts(seq, ",minixdf");
 253         if (test_opt(sb, GRPID))
 254                 seq_puts(seq, ",grpid");
 255         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 256                 seq_puts(seq, ",nogrpid");
 257         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 258             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 259                 seq_printf(seq, ",resuid=%u",
 260                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 261         }
 262         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 263             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 264                 seq_printf(seq, ",resgid=%u",
 265                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 266         }
 267         if (test_opt(sb, ERRORS_RO)) {
 268                 int def_errors = le16_to_cpu(es->s_errors);
 269
 270                 if (def_errors == EXT2_ERRORS_PANIC ||
 271                     def_errors == EXT2_ERRORS_CONTINUE) {
 272                         seq_puts(seq, ",errors=remount-ro");
 273                 }
 274         }
 275         if (test_opt(sb, ERRORS_CONT))
 276                 seq_puts(seq, ",errors=continue");
 277         if (test_opt(sb, ERRORS_PANIC))
 278                 seq_puts(seq, ",errors=panic");
 279         if (test_opt(sb, NO_UID32))
 280                 seq_puts(seq, ",nouid32");
 281         if (test_opt(sb, DEBUG))
 282                 seq_puts(seq, ",debug");
 283         if (test_opt(sb, OLDALLOC))
 284                 seq_puts(seq, ",oldalloc");
 285
 286 #ifdef CONFIG_EXT2_FS_XATTR
 287         if (test_opt(sb, XATTR_USER))
 288                 seq_puts(seq, ",user_xattr");
 289         if (!test_opt(sb, XATTR_USER) &&
 290             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 291                 seq_puts(seq, ",nouser_xattr");
 292         }
 293 #endif
 294
 295 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 296         if (test_opt(sb, POSIX_ACL))
 297                 seq_puts(seq, ",acl");
 298         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 299                 seq_puts(seq, ",noacl");
 300 #endif
 301
 302         if (test_opt(sb, NOBH))
 303                 seq_puts(seq, ",nobh");
 304
 305         if (test_opt(sb, USRQUOTA))
 306                 seq_puts(seq, ",usrquota");
 307
 308         if (test_opt(sb, GRPQUOTA))
 309                 seq_puts(seq, ",grpquota");
 310
 311         if (test_opt(sb, XIP))
 312                 seq_puts(seq, ",xip");
 313
 314         if (test_opt(sb, DAX))
 315                 seq_puts(seq, ",dax");
 316
 317         if (!test_opt(sb, RESERVATION))
 318                 seq_puts(seq, ",noreservation");
 319
 320         spin_unlock(&sbi->s_lock);
 321         return 0;
 322 }
 323
 324 #ifdef CONFIG_QUOTA
 325 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 326 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 327 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
 328                          const struct path *path);
 329 static struct dquot **ext2_get_dquots(struct inode *inode)
 330 {
 331         return EXT2_I(inode)->i_dquot;
 332 }
 333
 334 static const struct quotactl_ops ext2_quotactl_ops = {
 335         .quota_on       = ext2_quota_on,
 336         .quota_off      = ext2_quota_off,
 337         .quota_sync     = dquot_quota_sync,
 338         .get_state      = dquot_get_state,
 339         .set_info       = dquot_set_dqinfo,
 340         .get_dqblk      = dquot_get_dqblk,
 341         .set_dqblk      = dquot_set_dqblk,
 342         .get_nextdqblk  = dquot_get_next_dqblk,
 343 };
 344 #endif
 345
 346 static const struct super_operations ext2_sops = {
 347         .alloc_inode    = ext2_alloc_inode,
 348         .free_inode     = ext2_free_in_core_inode,
 349         .write_inode    = ext2_write_inode,
 350         .evict_inode    = ext2_evict_inode,
 351         .put_super      = ext2_put_super,
 352         .sync_fs        = ext2_sync_fs,
 353         .freeze_fs      = ext2_freeze,
 354         .unfreeze_fs    = ext2_unfreeze,
 355         .statfs         = ext2_statfs,
 356         .remount_fs     = ext2_remount,
 357         .show_options   = ext2_show_options,
 358 #ifdef CONFIG_QUOTA
 359         .quota_read     = ext2_quota_read,
 360         .quota_write    = ext2_quota_write,
 361         .get_dquots     = ext2_get_dquots,
 362 #endif
 363 };
 364
 365 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 366                 u64 ino, u32 generation)
 367 {
 368         struct inode *inode;
 369
 370         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 371                 return ERR_PTR(-ESTALE);
 372         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 373                 return ERR_PTR(-ESTALE);
 374
 375         /*
 376          * ext2_iget isn't quite right if the inode is currently unallocated!
 377          * However ext2_iget currently does appropriate checks to handle stale
 378          * inodes so everything is OK.
 379          */
 380         inode = ext2_iget(sb, ino);
 381         if (IS_ERR(inode))
 382                 return ERR_CAST(inode);
 383         if (generation && inode->i_generation != generation) {
 384                 /* we didn't find the right inode.. */
 385                 iput(inode);
 386                 return ERR_PTR(-ESTALE);
 387         }
 388         return inode;
 389 }
 390
 391 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 392                 int fh_len, int fh_type)
 393 {
 394         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 395                                     ext2_nfs_get_inode);
 396 }
 397
 398 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 399                 int fh_len, int fh_type)
 400 {
 401         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 402                                     ext2_nfs_get_inode);
 403 }
 404
 405 static const struct export_operations ext2_export_ops = {
 406         .fh_to_dentry = ext2_fh_to_dentry,
 407         .fh_to_parent = ext2_fh_to_parent,
 408         .get_parent = ext2_get_parent,
 409 };
 410
 411 static unsigned long get_sb_block(void **data)
 412 {
 413         unsigned long   sb_block;
 414         char            *options = (char *) *data;
 415
 416         if (!options || strncmp(options, "sb=", 3) != 0)
 417                 return 1;       /* Default location */
 418         options += 3;
 419         sb_block = simple_strtoul(options, &options, 0);
 420         if (*options && *options != ',') {
 421                 printk("EXT2-fs: Invalid sb specification: %s\n",
 422                        (char *) *data);
 423                 return 1;
 424         }
 425         if (*options == ',')
 426                 options++;
 427         *data = (void *) options;
 428         return sb_block;
 429 }
 430
 431 enum {
 432         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 433         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 434         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 435         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 436         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 437         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 438 };
 439
 440 static const match_table_t tokens = {
 441         {Opt_bsd_df, "bsddf"},
 442         {Opt_minix_df, "minixdf"},
 443         {Opt_grpid, "grpid"},
 444         {Opt_grpid, "bsdgroups"},
 445         {Opt_nogrpid, "nogrpid"},
 446         {Opt_nogrpid, "sysvgroups"},
 447         {Opt_resgid, "resgid=%u"},
 448         {Opt_resuid, "resuid=%u"},
 449         {Opt_sb, "sb=%u"},
 450         {Opt_err_cont, "errors=continue"},
 451         {Opt_err_panic, "errors=panic"},
 452         {Opt_err_ro, "errors=remount-ro"},
 453         {Opt_nouid32, "nouid32"},
 454         {Opt_nocheck, "check=none"},
 455         {Opt_nocheck, "nocheck"},
 456         {Opt_debug, "debug"},
 457         {Opt_oldalloc, "oldalloc"},
 458         {Opt_orlov, "orlov"},
 459         {Opt_nobh, "nobh"},
 460         {Opt_user_xattr, "user_xattr"},
 461         {Opt_nouser_xattr, "nouser_xattr"},
 462         {Opt_acl, "acl"},
 463         {Opt_noacl, "noacl"},
 464         {Opt_xip, "xip"},
 465         {Opt_dax, "dax"},
 466         {Opt_grpquota, "grpquota"},
 467         {Opt_ignore, "noquota"},
 468         {Opt_quota, "quota"},
 469         {Opt_usrquota, "usrquota"},
 470         {Opt_reservation, "reservation"},
 471         {Opt_noreservation, "noreservation"},
 472         {Opt_err, NULL}
 473 };
 474
 475 static int parse_options(char *options, struct super_block *sb,
 476                          struct ext2_mount_options *opts)
 477 {
 478         char *p;
 479         substring_t args[MAX_OPT_ARGS];
 480         int option;
 481         kuid_t uid;
 482         kgid_t gid;
 483
 484         if (!options)
 485                 return 1;
 486
 487         while ((p = strsep (&options, ",")) != NULL) {
 488                 int token;
 489                 if (!*p)
 490                         continue;
 491
 492                 token = match_token(p, tokens, args);
 493                 switch (token) {
 494                 case Opt_bsd_df:
 495                         clear_opt (opts->s_mount_opt, MINIX_DF);
 496                         break;
 497                 case Opt_minix_df:
 498                         set_opt (opts->s_mount_opt, MINIX_DF);
 499                         break;
 500                 case Opt_grpid:
 501                         set_opt (opts->s_mount_opt, GRPID);
 502                         break;
 503                 case Opt_nogrpid:
 504                         clear_opt (opts->s_mount_opt, GRPID);
 505                         break;
 506                 case Opt_resuid:
 507                         if (match_int(&args[0], &option))
 508                                 return 0;
 509                         uid = make_kuid(current_user_ns(), option);
 510                         if (!uid_valid(uid)) {
 511                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 512                                 return 0;
 513
 514                         }
 515                         opts->s_resuid = uid;
 516                         break;
 517                 case Opt_resgid:
 518                         if (match_int(&args[0], &option))
 519                                 return 0;
 520                         gid = make_kgid(current_user_ns(), option);
 521                         if (!gid_valid(gid)) {
 522                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 523                                 return 0;
 524                         }
 525                         opts->s_resgid = gid;
 526                         break;
 527                 case Opt_sb:
 528                         /* handled by get_sb_block() instead of here */
 529                         /* *sb_block = match_int(&args[0]); */
 530                         break;
 531                 case Opt_err_panic:
 532                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 533                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 534                         set_opt (opts->s_mount_opt, ERRORS_PANIC);
 535                         break;
 536                 case Opt_err_ro:
 537                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 538                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 539                         set_opt (opts->s_mount_opt, ERRORS_RO);
 540                         break;
 541                 case Opt_err_cont:
 542                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 543                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 544                         set_opt (opts->s_mount_opt, ERRORS_CONT);
 545                         break;
 546                 case Opt_nouid32:
 547                         set_opt (opts->s_mount_opt, NO_UID32);
 548                         break;
 549                 case Opt_nocheck:
 550                         ext2_msg(sb, KERN_WARNING,
 551                                 "Option nocheck/check=none is deprecated and"
 552                                 " will be removed in June 2020.");
 553                         clear_opt (opts->s_mount_opt, CHECK);
 554                         break;
 555                 case Opt_debug:
 556                         set_opt (opts->s_mount_opt, DEBUG);
 557                         break;
 558                 case Opt_oldalloc:
 559                         set_opt (opts->s_mount_opt, OLDALLOC);
 560                         break;
 561                 case Opt_orlov:
 562                         clear_opt (opts->s_mount_opt, OLDALLOC);
 563                         break;
 564                 case Opt_nobh:
 565                         set_opt (opts->s_mount_opt, NOBH);
 566                         break;
 567 #ifdef CONFIG_EXT2_FS_XATTR
 568                 case Opt_user_xattr:
 569                         set_opt (opts->s_mount_opt, XATTR_USER);
 570                         break;
 571                 case Opt_nouser_xattr:
 572                         clear_opt (opts->s_mount_opt, XATTR_USER);
 573                         break;
 574 #else
 575                 case Opt_user_xattr:
 576                 case Opt_nouser_xattr:
 577                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 578                                 "not supported");
 579                         break;
 580 #endif
 581 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 582                 case Opt_acl:
 583                         set_opt(opts->s_mount_opt, POSIX_ACL);
 584                         break;
 585                 case Opt_noacl:
 586                         clear_opt(opts->s_mount_opt, POSIX_ACL);
 587                         break;
 588 #else
 589                 case Opt_acl:
 590                 case Opt_noacl:
 591                         ext2_msg(sb, KERN_INFO,
 592                                 "(no)acl options not supported");
 593                         break;
 594 #endif
 595                 case Opt_xip:
 596                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 597                         set_opt(opts->s_mount_opt, XIP);
 598                         /* Fall through */
 599                 case Opt_dax:
 600 #ifdef CONFIG_FS_DAX
 601                         ext2_msg(sb, KERN_WARNING,
 602                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 603                         set_opt(opts->s_mount_opt, DAX);
 604 #else
 605                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 606 #endif
 607                         break;
 608
 609 #if defined(CONFIG_QUOTA)
 610                 case Opt_quota:
 611                 case Opt_usrquota:
 612                         set_opt(opts->s_mount_opt, USRQUOTA);
 613                         break;
 614
 615                 case Opt_grpquota:
 616                         set_opt(opts->s_mount_opt, GRPQUOTA);
 617                         break;
 618 #else
 619                 case Opt_quota:
 620                 case Opt_usrquota:
 621                 case Opt_grpquota:
 622                         ext2_msg(sb, KERN_INFO,
 623                                 "quota operations not supported");
 624                         break;
 625 #endif
 626
 627                 case Opt_reservation:
 628                         set_opt(opts->s_mount_opt, RESERVATION);
 629                         ext2_msg(sb, KERN_INFO, "reservations ON");
 630                         break;
 631                 case Opt_noreservation:
 632                         clear_opt(opts->s_mount_opt, RESERVATION);
 633                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 634                         break;
 635                 case Opt_ignore:
 636                         break;
 637                 default:
 638                         return 0;
 639                 }
 640         }
 641         return 1;
 642 }
 643
 644 static int ext2_setup_super (struct super_block * sb,
 645                               struct ext2_super_block * es,
 646                               int read_only)
 647 {
 648         int res = 0;
 649         struct ext2_sb_info *sbi = EXT2_SB(sb);
 650
 651         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 652                 ext2_msg(sb, KERN_ERR,
 653                         "error: revision level too high, "
 654                         "forcing read-only mode");
 655                 res = SB_RDONLY;
 656         }
 657         if (read_only)
 658                 return res;
 659         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 660                 ext2_msg(sb, KERN_WARNING,
 661                         "warning: mounting unchecked fs, "
 662                         "running e2fsck is recommended");
 663         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 664                 ext2_msg(sb, KERN_WARNING,
 665                         "warning: mounting fs with errors, "
 666                         "running e2fsck is recommended");
 667         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 668                  le16_to_cpu(es->s_mnt_count) >=
 669                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 670                 ext2_msg(sb, KERN_WARNING,
 671                         "warning: maximal mount count reached, "
 672                         "running e2fsck is recommended");
 673         else if (le32_to_cpu(es->s_checkinterval) &&
 674                 (le32_to_cpu(es->s_lastcheck) +
 675                         le32_to_cpu(es->s_checkinterval) <=
 676                         ktime_get_real_seconds()))
 677                 ext2_msg(sb, KERN_WARNING,
 678                         "warning: checktime reached, "
 679                         "running e2fsck is recommended");
 680         if (!le16_to_cpu(es->s_max_mnt_count))
 681                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 682         le16_add_cpu(&es->s_mnt_count, 1);
 683         if (test_opt (sb, DEBUG))
 684                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, gc=%lu, "
 685                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 686                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 687                         sbi->s_groups_count,
 688                         EXT2_BLOCKS_PER_GROUP(sb),
 689                         EXT2_INODES_PER_GROUP(sb),
 690                         sbi->s_mount_opt);
 691         return res;
 692 }
 693
 694 static int ext2_check_descriptors(struct super_block *sb)
 695 {
 696         int i;
 697         struct ext2_sb_info *sbi = EXT2_SB(sb);
 698
 699         ext2_debug ("Checking group descriptors");
 700
 701         for (i = 0; i < sbi->s_groups_count; i++) {
 702                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 703                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 704                 ext2_fsblk_t last_block;
 705
 706                 if (i == sbi->s_groups_count - 1)
 707                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 708                 else
 709                         last_block = first_block +
 710                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 711
 712                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 713                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 714                 {
 715                         ext2_error (sb, "ext2_check_descriptors",
 716                                     "Block bitmap for group %d"
 717                                     " not in group (block %lu)!",
 718                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 719                         return 0;
 720                 }
 721                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 722                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 723                 {
 724                         ext2_error (sb, "ext2_check_descriptors",
 725                                     "Inode bitmap for group %d"
 726                                     " not in group (block %lu)!",
 727                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 728                         return 0;
 729                 }
 730                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 731                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 732                     last_block)
 733                 {
 734                         ext2_error (sb, "ext2_check_descriptors",
 735                                     "Inode table for group %d"
 736                                     " not in group (block %lu)!",
 737                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 738                         return 0;
 739                 }
 740         }
 741         return 1;
 742 }
 743
 744 /*
 745  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 746  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 747  * We need to be 1 filesystem block less than the 2^32 sector limit.
 748  */
 749 static loff_t ext2_max_size(int bits)
 750 {
 751         loff_t res = EXT2_NDIR_BLOCKS;
 752         int meta_blocks;
 753         unsigned int upper_limit;
 754         unsigned int ppb = 1 << (bits-2);
 755
 756         /* This is calculated to be the largest file size for a
 757          * dense, file such that the total number of
 758          * sectors in the file, including data and all indirect blocks,
 759          * does not exceed 2^32 -1
 760          * __u32 i_blocks representing the total number of
 761          * 512 bytes blocks of the file
 762          */
 763         upper_limit = (1LL << 32) - 1;
 764
 765         /* total blocks in file system block size */
 766         upper_limit >>= (bits - 9);
 767
 768         /* Compute how many blocks we can address by block tree */
 769         res += 1LL << (bits-2);
 770         res += 1LL << (2*(bits-2));
 771         res += 1LL << (3*(bits-2));
 772         /* Compute how many metadata blocks are needed */
 773         meta_blocks = 1;
 774         meta_blocks += 1 + ppb;
 775         meta_blocks += 1 + ppb + ppb * ppb;
 776         /* Does block tree limit file size? */
 777         if (res + meta_blocks <= upper_limit)
 778                 goto check_lfs;
 779
 780         res = upper_limit;
 781         /* How many metadata blocks are needed for addressing upper_limit? */
 782         upper_limit -= EXT2_NDIR_BLOCKS;
 783         /* indirect blocks */
 784         meta_blocks = 1;
 785         upper_limit -= ppb;
 786         /* double indirect blocks */
 787         if (upper_limit < ppb * ppb) {
 788                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 789                 res -= meta_blocks;
 790                 goto check_lfs;
 791         }
 792         meta_blocks += 1 + ppb;
 793         upper_limit -= ppb * ppb;
 794         /* tripple indirect blocks for the rest */
 795         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 796                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 797         res -= meta_blocks;
 798 check_lfs:
 799         res <<= bits;
 800         if (res > MAX_LFS_FILESIZE)
 801                 res = MAX_LFS_FILESIZE;
 802
 803         return res;
 804 }
 805
 806 static unsigned long descriptor_loc(struct super_block *sb,
 807                                     unsigned long logic_sb_block,
 808                                     int nr)
 809 {
 810         struct ext2_sb_info *sbi = EXT2_SB(sb);
 811         unsigned long bg, first_meta_bg;
 812         int has_super = 0;
 813
 814         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 815
 816         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 817             nr < first_meta_bg)
 818                 return (logic_sb_block + nr + 1);
 819         bg = sbi->s_desc_per_block * nr;
 820         if (ext2_bg_has_super(sb, bg))
 821                 has_super = 1;
 822
 823         return ext2_group_first_block_no(sb, bg) + has_super;
 824 }
 825
 826 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 827 {
 828         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 829         struct buffer_head * bh;
 830         struct ext2_sb_info * sbi;
 831         struct ext2_super_block * es;
 832         struct inode *root;
 833         unsigned long block;
 834         unsigned long sb_block = get_sb_block(&data);
 835         unsigned long logic_sb_block;
 836         unsigned long offset = 0;
 837         unsigned long def_mount_opts;
 838         long ret = -ENOMEM;
 839         int blocksize = BLOCK_SIZE;
 840         int db_count;
 841         int i, j;
 842         __le32 features;
 843         int err;
 844         struct ext2_mount_options opts;
 845
 846         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 847         if (!sbi)
 848                 goto failed;
 849
 850         sbi->s_blockgroup_lock =
 851                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 852         if (!sbi->s_blockgroup_lock) {
 853                 kfree(sbi);
 854                 goto failed;
 855         }
 856         sb->s_fs_info = sbi;
 857         sbi->s_sb_block = sb_block;
 858         sbi->s_daxdev = dax_dev;
 859
 860         spin_lock_init(&sbi->s_lock);
 861         ret = -EINVAL;
 862
 863         /*
 864          * See what the current blocksize for the device is, and
 865          * use that as the blocksize.  Otherwise (or if the blocksize
 866          * is smaller than the default) use the default.
 867          * This is important for devices that have a hardware
 868          * sectorsize that is larger than the default.
 869          */
 870         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 871         if (!blocksize) {
 872                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 873                 goto failed_sbi;
 874         }
 875
 876         /*
 877          * If the superblock doesn't start on a hardware sector boundary,
 878          * calculate the offset.
 879          */
 880         if (blocksize != BLOCK_SIZE) {
 881                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 882                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 883         } else {
 884                 logic_sb_block = sb_block;
 885         }
 886
 887         if (!(bh = sb_bread(sb, logic_sb_block))) {
 888                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 889                 goto failed_sbi;
 890         }
 891         /*
 892          * Note: s_es must be initialized as soon as possible because
 893          *       some ext2 macro-instructions depend on its value
 894          */
 895         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 896         sbi->s_es = es;
 897         sb->s_magic = le16_to_cpu(es->s_magic);
 898
 899         if (sb->s_magic != EXT2_SUPER_MAGIC)
 900                 goto cantfind_ext2;
 901
 902         opts.s_mount_opt = 0;
 903         /* Set defaults before we parse the mount options */
 904         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 905         if (def_mount_opts & EXT2_DEFM_DEBUG)
 906                 set_opt(opts.s_mount_opt, DEBUG);
 907         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 908                 set_opt(opts.s_mount_opt, GRPID);
 909         if (def_mount_opts & EXT2_DEFM_UID16)
 910                 set_opt(opts.s_mount_opt, NO_UID32);
 911 #ifdef CONFIG_EXT2_FS_XATTR
 912         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 913                 set_opt(opts.s_mount_opt, XATTR_USER);
 914 #endif
 915 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 916         if (def_mount_opts & EXT2_DEFM_ACL)
 917                 set_opt(opts.s_mount_opt, POSIX_ACL);
 918 #endif
 919
 920         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 921                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 922         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 923                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 924         else
 925                 set_opt(opts.s_mount_opt, ERRORS_RO);
 926
 927         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 928         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 929
 930         set_opt(opts.s_mount_opt, RESERVATION);
 931
 932         if (!parse_options((char *) data, sb, &opts))
 933                 goto failed_mount;
 934
 935         sbi->s_mount_opt = opts.s_mount_opt;
 936         sbi->s_resuid = opts.s_resuid;
 937         sbi->s_resgid = opts.s_resgid;
 938
 939         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 940                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
 941         sb->s_iflags |= SB_I_CGROUPWB;
 942
 943         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 944             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 945              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 946              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 947                 ext2_msg(sb, KERN_WARNING,
 948                         "warning: feature flags set on rev 0 fs, "
 949                         "running e2fsck is recommended");
 950         /*
 951          * Check feature flags regardless of the revision level, since we
 952          * previously didn't change the revision level when setting the flags,
 953          * so there is a chance incompat flags are set on a rev 0 filesystem.
 954          */
 955         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 956         if (features) {
 957                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 958                        "unsupported optional features (%x)",
 959                         le32_to_cpu(features));
 960                 goto failed_mount;
 961         }
 962         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 963                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 964                        "unsupported optional features (%x)",
 965                        le32_to_cpu(features));
 966                 goto failed_mount;
 967         }
 968
 969         if (le32_to_cpu(es->s_log_block_size) >
 970             (EXT2_MAX_BLOCK_LOG_SIZE - BLOCK_SIZE_BITS)) {
 971                 ext2_msg(sb, KERN_ERR,
 972                          "Invalid log block size: %u",
 973                          le32_to_cpu(es->s_log_block_size));
 974                 goto failed_mount;
 975         }
 976         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 977
 978         if (test_opt(sb, DAX)) {
 979                 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
 980                         ext2_msg(sb, KERN_ERR,
 981                                 "DAX unsupported by block device. Turning off DAX.");
 982                         clear_opt(sbi->s_mount_opt, DAX);
 983                 }
 984         }
 985
 986         /* If the blocksize doesn't match, re-read the thing.. */
 987         if (sb->s_blocksize != blocksize) {
 988                 brelse(bh);
 989
 990                 if (!sb_set_blocksize(sb, blocksize)) {
 991                         ext2_msg(sb, KERN_ERR,
 992                                 "error: bad blocksize %d", blocksize);
 993                         goto failed_sbi;
 994                 }
 995
 996                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 997                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 998                 bh = sb_bread(sb, logic_sb_block);
 999                 if(!bh) {
1000                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
1001                                 "superblock on 2nd try");
1002                         goto failed_sbi;
1003                 }
1004                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
1005                 sbi->s_es = es;
1006                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
1007                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
1008                         goto failed_mount;
1009                 }
1010         }
1011
1012         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1013         sb->s_max_links = EXT2_LINK_MAX;
1014         sb->s_time_min = S32_MIN;
1015         sb->s_time_max = S32_MAX;
1016
1017         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1018                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1019                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1020         } else {
1021                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1022                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1023                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1024                     !is_power_of_2(sbi->s_inode_size) ||
1025                     (sbi->s_inode_size > blocksize)) {
1026                         ext2_msg(sb, KERN_ERR,
1027                                 "error: unsupported inode size: %d",
1028                                 sbi->s_inode_size);
1029                         goto failed_mount;
1030                 }
1031         }
1032
1033         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1034         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1035
1036         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1037         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1038                 goto cantfind_ext2;
1039         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1040                                         sbi->s_inodes_per_block;
1041         sbi->s_desc_per_block = sb->s_blocksize /
1042                                         sizeof (struct ext2_group_desc);
1043         sbi->s_sbh = bh;
1044         sbi->s_mount_state = le16_to_cpu(es->s_state);
1045         sbi->s_addr_per_block_bits =
1046                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1047         sbi->s_desc_per_block_bits =
1048                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1049
1050         if (sb->s_magic != EXT2_SUPER_MAGIC)
1051                 goto cantfind_ext2;
1052
1053         if (sb->s_blocksize != bh->b_size) {
1054                 if (!silent)
1055                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1056                 goto failed_mount;
1057         }
1058
1059         if (es->s_log_frag_size != es->s_log_block_size) {
1060                 ext2_msg(sb, KERN_ERR,
1061                         "error: fragsize log %u != blocksize log %u",
1062                         le32_to_cpu(es->s_log_frag_size), sb->s_blocksize_bits);
1063                 goto failed_mount;
1064         }
1065
1066         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1067                 ext2_msg(sb, KERN_ERR,
1068                         "error: #blocks per group too big: %lu",
1069                         sbi->s_blocks_per_group);
1070                 goto failed_mount;
1071         }
1072         if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
1073             sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1074                 ext2_msg(sb, KERN_ERR,
1075                         "error: invalid #inodes per group: %lu",
1076                         sbi->s_inodes_per_group);
1077                 goto failed_mount;
1078         }
1079
1080         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1081                 goto cantfind_ext2;
1082         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1083                                 le32_to_cpu(es->s_first_data_block) - 1)
1084                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1085         if ((u64)sbi->s_groups_count * sbi->s_inodes_per_group !=
1086             le32_to_cpu(es->s_inodes_count)) {
1087                 ext2_msg(sb, KERN_ERR, "error: invalid #inodes: %u vs computed %llu",
1088                          le32_to_cpu(es->s_inodes_count),
1089                          (u64)sbi->s_groups_count * sbi->s_inodes_per_group);
1090                 goto failed_mount;
1091         }
1092         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1093                    EXT2_DESC_PER_BLOCK(sb);
1094         sbi->s_group_desc = kmalloc_array (db_count,
1095                                            sizeof(struct buffer_head *),
1096                                            GFP_KERNEL);
1097         if (sbi->s_group_desc == NULL) {
1098                 ret = -ENOMEM;
1099                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1100                 goto failed_mount;
1101         }
1102         bgl_lock_init(sbi->s_blockgroup_lock);
1103         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1104         if (!sbi->s_debts) {
1105                 ret = -ENOMEM;
1106                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1107                 goto failed_mount_group_desc;
1108         }
1109         for (i = 0; i < db_count; i++) {
1110                 block = descriptor_loc(sb, logic_sb_block, i);
1111                 sbi->s_group_desc[i] = sb_bread(sb, block);
1112                 if (!sbi->s_group_desc[i]) {
1113                         for (j = 0; j < i; j++)
1114                                 brelse (sbi->s_group_desc[j]);
1115                         ext2_msg(sb, KERN_ERR,
1116                                 "error: unable to read group descriptors");
1117                         goto failed_mount_group_desc;
1118                 }
1119         }
1120         if (!ext2_check_descriptors (sb)) {
1121                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1122                 goto failed_mount2;
1123         }
1124         sbi->s_gdb_count = db_count;
1125         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1126         spin_lock_init(&sbi->s_next_gen_lock);
1127
1128         /* per fileystem reservation list head & lock */
1129         spin_lock_init(&sbi->s_rsv_window_lock);
1130         sbi->s_rsv_window_root = RB_ROOT;
1131         /*
1132          * Add a single, static dummy reservation to the start of the
1133          * reservation window list --- it gives us a placeholder for
1134          * append-at-start-of-list which makes the allocation logic
1135          * _much_ simpler.
1136          */
1137         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1138         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1139         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1140         sbi->s_rsv_window_head.rsv_goal_size = 0;
1141         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1142
1143         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1144                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1145         if (!err) {
1146                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1147                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1148         }
1149         if (!err) {
1150                 err = percpu_counter_init(&sbi->s_dirs_counter,
1151                                 ext2_count_dirs(sb), GFP_KERNEL);
1152         }
1153         if (err) {
1154                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1155                 goto failed_mount3;
1156         }
1157
1158 #ifdef CONFIG_EXT2_FS_XATTR
1159         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1160         if (!sbi->s_ea_block_cache) {
1161                 ret = -ENOMEM;
1162                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1163                 goto failed_mount3;
1164         }
1165 #endif
1166         /*
1167          * set up enough so that it can read an inode
1168          */
1169         sb->s_op = &ext2_sops;
1170         sb->s_export_op = &ext2_export_ops;
1171         sb->s_xattr = ext2_xattr_handlers;
1172
1173 #ifdef CONFIG_QUOTA
1174         sb->dq_op = &dquot_operations;
1175         sb->s_qcop = &ext2_quotactl_ops;
1176         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1177 #endif
1178
1179         root = ext2_iget(sb, EXT2_ROOT_INO);
1180         if (IS_ERR(root)) {
1181                 ret = PTR_ERR(root);
1182                 goto failed_mount3;
1183         }
1184         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1185                 iput(root);
1186                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1187                 goto failed_mount3;
1188         }
1189
1190         sb->s_root = d_make_root(root);
1191         if (!sb->s_root) {
1192                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1193                 ret = -ENOMEM;
1194                 goto failed_mount3;
1195         }
1196         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1197                 ext2_msg(sb, KERN_WARNING,
1198                         "warning: mounting ext3 filesystem as ext2");
1199         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1200                 sb->s_flags |= SB_RDONLY;
1201         ext2_write_super(sb);
1202         return 0;
1203
1204 cantfind_ext2:
1205         if (!silent)
1206                 ext2_msg(sb, KERN_ERR,
1207                         "error: can't find an ext2 filesystem on dev %s.",
1208                         sb->s_id);
1209         goto failed_mount;
1210 failed_mount3:
1211         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1212         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1213         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1214         percpu_counter_destroy(&sbi->s_dirs_counter);
1215 failed_mount2:
1216         for (i = 0; i < db_count; i++)
1217                 brelse(sbi->s_group_desc[i]);
1218 failed_mount_group_desc:
1219         kfree(sbi->s_group_desc);
1220         kfree(sbi->s_debts);
1221 failed_mount:
1222         brelse(bh);
1223 failed_sbi:
1224         sb->s_fs_info = NULL;
1225         kfree(sbi->s_blockgroup_lock);
1226         kfree(sbi);
1227 failed:
1228         fs_put_dax(dax_dev);
1229         return ret;
1230 }
1231
1232 static void ext2_clear_super_error(struct super_block *sb)
1233 {
1234         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1235
1236         if (buffer_write_io_error(sbh)) {
1237                 /*
1238                  * Oh, dear.  A previous attempt to write the
1239                  * superblock failed.  This could happen because the
1240                  * USB device was yanked out.  Or it could happen to
1241                  * be a transient write error and maybe the block will
1242                  * be remapped.  Nothing we can do but to retry the
1243                  * write and hope for the best.
1244                  */
1245                 ext2_msg(sb, KERN_ERR,
1246                        "previous I/O error to superblock detected");
1247                 clear_buffer_write_io_error(sbh);
1248                 set_buffer_uptodate(sbh);
1249         }
1250 }
1251
1252 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1253                      int wait)
1254 {
1255         ext2_clear_super_error(sb);
1256         spin_lock(&EXT2_SB(sb)->s_lock);
1257         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1258         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1259         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1260         /* unlock before we do IO */
1261         spin_unlock(&EXT2_SB(sb)->s_lock);
1262         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1263         if (wait)
1264                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1265 }
1266
1267 /*
1268  * In the second extended file system, it is not necessary to
1269  * write the super block since we use a mapping of the
1270  * disk super block in a buffer.
1271  *
1272  * However, this function is still used to set the fs valid
1273  * flags to 0.  We need to set this flag to 0 since the fs
1274  * may have been checked while mounted and e2fsck may have
1275  * set s_state to EXT2_VALID_FS after some corrections.
1276  */
1277 static int ext2_sync_fs(struct super_block *sb, int wait)
1278 {
1279         struct ext2_sb_info *sbi = EXT2_SB(sb);
1280         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1281
1282         /*
1283          * Write quota structures to quota file, sync_blockdev() will write
1284          * them to disk later
1285          */
1286         dquot_writeback_dquots(sb, -1);
1287
1288         spin_lock(&sbi->s_lock);
1289         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1290                 ext2_debug("setting valid to 0\n");
1291                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1292         }
1293         spin_unlock(&sbi->s_lock);
1294         ext2_sync_super(sb, es, wait);
1295         return 0;
1296 }
1297
1298 static int ext2_freeze(struct super_block *sb)
1299 {
1300         struct ext2_sb_info *sbi = EXT2_SB(sb);
1301
1302         /*
1303          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1304          * because we have unattached inodes and thus filesystem is not fully
1305          * consistent.
1306          */
1307         if (atomic_long_read(&sb->s_remove_count)) {
1308                 ext2_sync_fs(sb, 1);
1309                 return 0;
1310         }
1311         /* Set EXT2_FS_VALID flag */
1312         spin_lock(&sbi->s_lock);
1313         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1314         spin_unlock(&sbi->s_lock);
1315         ext2_sync_super(sb, sbi->s_es, 1);
1316
1317         return 0;
1318 }
1319
1320 static int ext2_unfreeze(struct super_block *sb)
1321 {
1322         /* Just write sb to clear EXT2_VALID_FS flag */
1323         ext2_write_super(sb);
1324
1325         return 0;
1326 }
1327
1328 static void ext2_write_super(struct super_block *sb)
1329 {
1330         if (!sb_rdonly(sb))
1331                 ext2_sync_fs(sb, 1);
1332 }
1333
1334 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1335 {
1336         struct ext2_sb_info * sbi = EXT2_SB(sb);
1337         struct ext2_super_block * es;
1338         struct ext2_mount_options new_opts;
1339         int err;
1340
1341         sync_filesystem(sb);
1342
1343         spin_lock(&sbi->s_lock);
1344         new_opts.s_mount_opt = sbi->s_mount_opt;
1345         new_opts.s_resuid = sbi->s_resuid;
1346         new_opts.s_resgid = sbi->s_resgid;
1347         spin_unlock(&sbi->s_lock);
1348
1349         if (!parse_options(data, sb, &new_opts))
1350                 return -EINVAL;
1351
1352         spin_lock(&sbi->s_lock);
1353         es = sbi->s_es;
1354         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1355                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1356                          "dax flag with busy inodes while remounting");
1357                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1358         }
1359         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1360                 goto out_set;
1361         if (*flags & SB_RDONLY) {
1362                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1363                     !(sbi->s_mount_state & EXT2_VALID_FS))
1364                         goto out_set;
1365
1366                 /*
1367                  * OK, we are remounting a valid rw partition rdonly, so set
1368                  * the rdonly flag and then mark the partition as valid again.
1369                  */
1370                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1371                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1372                 spin_unlock(&sbi->s_lock);
1373
1374                 err = dquot_suspend(sb, -1);
1375                 if (err < 0)
1376                         return err;
1377
1378                 ext2_sync_super(sb, es, 1);
1379         } else {
1380                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1381                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1382                 if (ret) {
1383                         spin_unlock(&sbi->s_lock);
1384                         ext2_msg(sb, KERN_WARNING,
1385                                 "warning: couldn't remount RDWR because of "
1386                                 "unsupported optional features (%x).",
1387                                 le32_to_cpu(ret));
1388                         return -EROFS;
1389                 }
1390                 /*
1391                  * Mounting a RDONLY partition read-write, so reread and
1392                  * store the current valid flag.  (It may have been changed
1393                  * by e2fsck since we originally mounted the partition.)
1394                  */
1395                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1396                 if (!ext2_setup_super (sb, es, 0))
1397                         sb->s_flags &= ~SB_RDONLY;
1398                 spin_unlock(&sbi->s_lock);
1399
1400                 ext2_write_super(sb);
1401
1402                 dquot_resume(sb, -1);
1403         }
1404
1405         spin_lock(&sbi->s_lock);
1406 out_set:
1407         sbi->s_mount_opt = new_opts.s_mount_opt;
1408         sbi->s_resuid = new_opts.s_resuid;
1409         sbi->s_resgid = new_opts.s_resgid;
1410         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1411                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1412         spin_unlock(&sbi->s_lock);
1413
1414         return 0;
1415 }
1416
1417 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1418 {
1419         struct super_block *sb = dentry->d_sb;
1420         struct ext2_sb_info *sbi = EXT2_SB(sb);
1421         struct ext2_super_block *es = sbi->s_es;
1422         u64 fsid;
1423
1424         spin_lock(&sbi->s_lock);
1425
1426         if (test_opt (sb, MINIX_DF))
1427                 sbi->s_overhead_last = 0;
1428         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1429                 unsigned long i, overhead = 0;
1430                 smp_rmb();
1431
1432                 /*
1433                  * Compute the overhead (FS structures). This is constant
1434                  * for a given filesystem unless the number of block groups
1435                  * changes so we cache the previous value until it does.
1436                  */
1437
1438                 /*
1439                  * All of the blocks before first_data_block are
1440                  * overhead
1441                  */
1442                 overhead = le32_to_cpu(es->s_first_data_block);
1443
1444                 /*
1445                  * Add the overhead attributed to the superblock and
1446                  * block group descriptors.  If the sparse superblocks
1447                  * feature is turned on, then not all groups have this.
1448                  */
1449                 for (i = 0; i < sbi->s_groups_count; i++)
1450                         overhead += ext2_bg_has_super(sb, i) +
1451                                 ext2_bg_num_gdb(sb, i);
1452
1453                 /*
1454                  * Every block group has an inode bitmap, a block
1455                  * bitmap, and an inode table.
1456                  */
1457                 overhead += (sbi->s_groups_count *
1458                              (2 + sbi->s_itb_per_group));
1459                 sbi->s_overhead_last = overhead;
1460                 smp_wmb();
1461                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1462         }
1463
1464         buf->f_type = EXT2_SUPER_MAGIC;
1465         buf->f_bsize = sb->s_blocksize;
1466         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1467         buf->f_bfree = ext2_count_free_blocks(sb);
1468         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1469         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1470         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1471                 buf->f_bavail = 0;
1472         buf->f_files = le32_to_cpu(es->s_inodes_count);
1473         buf->f_ffree = ext2_count_free_inodes(sb);
1474         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1475         buf->f_namelen = EXT2_NAME_LEN;
1476         fsid = le64_to_cpup((void *)es->s_uuid) ^
1477                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1478         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1479         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1480         spin_unlock(&sbi->s_lock);
1481         return 0;
1482 }
1483
1484 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1485         int flags, const char *dev_name, void *data)
1486 {
1487         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1488 }
1489
1490 #ifdef CONFIG_QUOTA
1491
1492 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1493  * acquiring the locks... As quota files are never truncated and quota code
1494  * itself serializes the operations (and no one else should touch the files)
1495  * we don't have to be afraid of races */
1496 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1497                                size_t len, loff_t off)
1498 {
1499         struct inode *inode = sb_dqopt(sb)->files[type];
1500         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1501         int err = 0;
1502         int offset = off & (sb->s_blocksize - 1);
1503         int tocopy;
1504         size_t toread;
1505         struct buffer_head tmp_bh;
1506         struct buffer_head *bh;
1507         loff_t i_size = i_size_read(inode);
1508
1509         if (off > i_size)
1510                 return 0;
1511         if (off+len > i_size)
1512                 len = i_size-off;
1513         toread = len;
1514         while (toread > 0) {
1515                 tocopy = sb->s_blocksize - offset < toread ?
1516                                 sb->s_blocksize - offset : toread;
1517
1518                 tmp_bh.b_state = 0;
1519                 tmp_bh.b_size = sb->s_blocksize;
1520                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1521                 if (err < 0)
1522                         return err;
1523                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1524                         memset(data, 0, tocopy);
1525                 else {
1526                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1527                         if (!bh)
1528                                 return -EIO;
1529                         memcpy(data, bh->b_data+offset, tocopy);
1530                         brelse(bh);
1531                 }
1532                 offset = 0;
1533                 toread -= tocopy;
1534                 data += tocopy;
1535                 blk++;
1536         }
1537         return len;
1538 }
1539
1540 /* Write to quotafile */
1541 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1542                                 const char *data, size_t len, loff_t off)
1543 {
1544         struct inode *inode = sb_dqopt(sb)->files[type];
1545         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1546         int err = 0;
1547         int offset = off & (sb->s_blocksize - 1);
1548         int tocopy;
1549         size_t towrite = len;
1550         struct buffer_head tmp_bh;
1551         struct buffer_head *bh;
1552
1553         while (towrite > 0) {
1554                 tocopy = sb->s_blocksize - offset < towrite ?
1555                                 sb->s_blocksize - offset : towrite;
1556
1557                 tmp_bh.b_state = 0;
1558                 tmp_bh.b_size = sb->s_blocksize;
1559                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1560                 if (err < 0)
1561                         goto out;
1562                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1563                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1564                 else
1565                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1566                 if (unlikely(!bh)) {
1567                         err = -EIO;
1568                         goto out;
1569                 }
1570                 lock_buffer(bh);
1571                 memcpy(bh->b_data+offset, data, tocopy);
1572                 flush_dcache_page(bh->b_page);
1573                 set_buffer_uptodate(bh);
1574                 mark_buffer_dirty(bh);
1575                 unlock_buffer(bh);
1576                 brelse(bh);
1577                 offset = 0;
1578                 towrite -= tocopy;
1579                 data += tocopy;
1580                 blk++;
1581         }
1582 out:
1583         if (len == towrite)
1584                 return err;
1585         if (inode->i_size < off+len-towrite)
1586                 i_size_write(inode, off+len-towrite);
1587         inode_inc_iversion(inode);
1588         inode->i_mtime = inode->i_ctime = current_time(inode);
1589         mark_inode_dirty(inode);
1590         return len - towrite;
1591 }
1592
1593 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1594                          const struct path *path)
1595 {
1596         int err;
1597         struct inode *inode;
1598
1599         err = dquot_quota_on(sb, type, format_id, path);
1600         if (err)
1601                 return err;
1602
1603         inode = d_inode(path->dentry);
1604         inode_lock(inode);
1605         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1606         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1607                         S_NOATIME | S_IMMUTABLE);
1608         inode_unlock(inode);
1609         mark_inode_dirty(inode);
1610
1611         return 0;
1612 }
1613
1614 static int ext2_quota_off(struct super_block *sb, int type)
1615 {
1616         struct inode *inode = sb_dqopt(sb)->files[type];
1617         int err;
1618
1619         if (!inode || !igrab(inode))
1620                 goto out;
1621
1622         err = dquot_quota_off(sb, type);
1623         if (err)
1624                 goto out_put;
1625
1626         inode_lock(inode);
1627         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1628         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1629         inode_unlock(inode);
1630         mark_inode_dirty(inode);
1631 out_put:
1632         iput(inode);
1633         return err;
1634 out:
1635         return dquot_quota_off(sb, type);
1636 }
1637
1638 #endif
1639
1640 static struct file_system_type ext2_fs_type = {
1641         .owner          = THIS_MODULE,
1642         .name           = "ext2",
1643         .mount          = ext2_mount,
1644         .kill_sb        = kill_block_super,
1645         .fs_flags       = FS_REQUIRES_DEV,
1646 };
1647 MODULE_ALIAS_FS("ext2");
1648
1649 static int __init init_ext2_fs(void)
1650 {
1651         int err;
1652
1653         err = init_inodecache();
1654         if (err)
1655                 return err;
1656         err = register_filesystem(&ext2_fs_type);
1657         if (err)
1658                 goto out;
1659         return 0;
1660 out:
1661         destroy_inodecache();
1662         return err;
1663 }
1664
1665 static void __exit exit_ext2_fs(void)
1666 {
1667         unregister_filesystem(&ext2_fs_type);
1668         destroy_inodecache();
1669 }
1670
1671 MODULE_AUTHOR("Remy Card and others");
1672 MODULE_DESCRIPTION("Second Extended Filesystem");
1673 MODULE_LICENSE("GPL");
1674 module_init(init_ext2_fs)
1675 module_exit(exit_ext2_fs)