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_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_debug, "debug"},
 455         {Opt_oldalloc, "oldalloc"},
 456         {Opt_orlov, "orlov"},
 457         {Opt_nobh, "nobh"},
 458         {Opt_user_xattr, "user_xattr"},
 459         {Opt_nouser_xattr, "nouser_xattr"},
 460         {Opt_acl, "acl"},
 461         {Opt_noacl, "noacl"},
 462         {Opt_xip, "xip"},
 463         {Opt_dax, "dax"},
 464         {Opt_grpquota, "grpquota"},
 465         {Opt_ignore, "noquota"},
 466         {Opt_quota, "quota"},
 467         {Opt_usrquota, "usrquota"},
 468         {Opt_reservation, "reservation"},
 469         {Opt_noreservation, "noreservation"},
 470         {Opt_err, NULL}
 471 };
 472
 473 static int parse_options(char *options, struct super_block *sb,
 474                          struct ext2_mount_options *opts)
 475 {
 476         char *p;
 477         substring_t args[MAX_OPT_ARGS];
 478         int option;
 479         kuid_t uid;
 480         kgid_t gid;
 481
 482         if (!options)
 483                 return 1;
 484
 485         while ((p = strsep (&options, ",")) != NULL) {
 486                 int token;
 487                 if (!*p)
 488                         continue;
 489
 490                 token = match_token(p, tokens, args);
 491                 switch (token) {
 492                 case Opt_bsd_df:
 493                         clear_opt (opts->s_mount_opt, MINIX_DF);
 494                         break;
 495                 case Opt_minix_df:
 496                         set_opt (opts->s_mount_opt, MINIX_DF);
 497                         break;
 498                 case Opt_grpid:
 499                         set_opt (opts->s_mount_opt, GRPID);
 500                         break;
 501                 case Opt_nogrpid:
 502                         clear_opt (opts->s_mount_opt, GRPID);
 503                         break;
 504                 case Opt_resuid:
 505                         if (match_int(&args[0], &option))
 506                                 return 0;
 507                         uid = make_kuid(current_user_ns(), option);
 508                         if (!uid_valid(uid)) {
 509                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 510                                 return 0;
 511
 512                         }
 513                         opts->s_resuid = uid;
 514                         break;
 515                 case Opt_resgid:
 516                         if (match_int(&args[0], &option))
 517                                 return 0;
 518                         gid = make_kgid(current_user_ns(), option);
 519                         if (!gid_valid(gid)) {
 520                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 521                                 return 0;
 522                         }
 523                         opts->s_resgid = gid;
 524                         break;
 525                 case Opt_sb:
 526                         /* handled by get_sb_block() instead of here */
 527                         /* *sb_block = match_int(&args[0]); */
 528                         break;
 529                 case Opt_err_panic:
 530                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 531                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 532                         set_opt (opts->s_mount_opt, ERRORS_PANIC);
 533                         break;
 534                 case Opt_err_ro:
 535                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 536                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 537                         set_opt (opts->s_mount_opt, ERRORS_RO);
 538                         break;
 539                 case Opt_err_cont:
 540                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 541                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 542                         set_opt (opts->s_mount_opt, ERRORS_CONT);
 543                         break;
 544                 case Opt_nouid32:
 545                         set_opt (opts->s_mount_opt, NO_UID32);
 546                         break;
 547                 case Opt_debug:
 548                         set_opt (opts->s_mount_opt, DEBUG);
 549                         break;
 550                 case Opt_oldalloc:
 551                         set_opt (opts->s_mount_opt, OLDALLOC);
 552                         break;
 553                 case Opt_orlov:
 554                         clear_opt (opts->s_mount_opt, OLDALLOC);
 555                         break;
 556                 case Opt_nobh:
 557                         set_opt (opts->s_mount_opt, NOBH);
 558                         break;
 559 #ifdef CONFIG_EXT2_FS_XATTR
 560                 case Opt_user_xattr:
 561                         set_opt (opts->s_mount_opt, XATTR_USER);
 562                         break;
 563                 case Opt_nouser_xattr:
 564                         clear_opt (opts->s_mount_opt, XATTR_USER);
 565                         break;
 566 #else
 567                 case Opt_user_xattr:
 568                 case Opt_nouser_xattr:
 569                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 570                                 "not supported");
 571                         break;
 572 #endif
 573 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 574                 case Opt_acl:
 575                         set_opt(opts->s_mount_opt, POSIX_ACL);
 576                         break;
 577                 case Opt_noacl:
 578                         clear_opt(opts->s_mount_opt, POSIX_ACL);
 579                         break;
 580 #else
 581                 case Opt_acl:
 582                 case Opt_noacl:
 583                         ext2_msg(sb, KERN_INFO,
 584                                 "(no)acl options not supported");
 585                         break;
 586 #endif
 587                 case Opt_xip:
 588                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 589                         set_opt(opts->s_mount_opt, XIP);
 590                         fallthrough;
 591                 case Opt_dax:
 592 #ifdef CONFIG_FS_DAX
 593                         ext2_msg(sb, KERN_WARNING,
 594                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 595                         set_opt(opts->s_mount_opt, DAX);
 596 #else
 597                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 598 #endif
 599                         break;
 600
 601 #if defined(CONFIG_QUOTA)
 602                 case Opt_quota:
 603                 case Opt_usrquota:
 604                         set_opt(opts->s_mount_opt, USRQUOTA);
 605                         break;
 606
 607                 case Opt_grpquota:
 608                         set_opt(opts->s_mount_opt, GRPQUOTA);
 609                         break;
 610 #else
 611                 case Opt_quota:
 612                 case Opt_usrquota:
 613                 case Opt_grpquota:
 614                         ext2_msg(sb, KERN_INFO,
 615                                 "quota operations not supported");
 616                         break;
 617 #endif
 618
 619                 case Opt_reservation:
 620                         set_opt(opts->s_mount_opt, RESERVATION);
 621                         ext2_msg(sb, KERN_INFO, "reservations ON");
 622                         break;
 623                 case Opt_noreservation:
 624                         clear_opt(opts->s_mount_opt, RESERVATION);
 625                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 626                         break;
 627                 case Opt_ignore:
 628                         break;
 629                 default:
 630                         return 0;
 631                 }
 632         }
 633         return 1;
 634 }
 635
 636 static int ext2_setup_super (struct super_block * sb,
 637                               struct ext2_super_block * es,
 638                               int read_only)
 639 {
 640         int res = 0;
 641         struct ext2_sb_info *sbi = EXT2_SB(sb);
 642
 643         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 644                 ext2_msg(sb, KERN_ERR,
 645                         "error: revision level too high, "
 646                         "forcing read-only mode");
 647                 res = SB_RDONLY;
 648         }
 649         if (read_only)
 650                 return res;
 651         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 652                 ext2_msg(sb, KERN_WARNING,
 653                         "warning: mounting unchecked fs, "
 654                         "running e2fsck is recommended");
 655         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 656                 ext2_msg(sb, KERN_WARNING,
 657                         "warning: mounting fs with errors, "
 658                         "running e2fsck is recommended");
 659         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 660                  le16_to_cpu(es->s_mnt_count) >=
 661                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 662                 ext2_msg(sb, KERN_WARNING,
 663                         "warning: maximal mount count reached, "
 664                         "running e2fsck is recommended");
 665         else if (le32_to_cpu(es->s_checkinterval) &&
 666                 (le32_to_cpu(es->s_lastcheck) +
 667                         le32_to_cpu(es->s_checkinterval) <=
 668                         ktime_get_real_seconds()))
 669                 ext2_msg(sb, KERN_WARNING,
 670                         "warning: checktime reached, "
 671                         "running e2fsck is recommended");
 672         if (!le16_to_cpu(es->s_max_mnt_count))
 673                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 674         le16_add_cpu(&es->s_mnt_count, 1);
 675         if (test_opt (sb, DEBUG))
 676                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 677                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 678                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 679                         sbi->s_frag_size,
 680                         sbi->s_groups_count,
 681                         EXT2_BLOCKS_PER_GROUP(sb),
 682                         EXT2_INODES_PER_GROUP(sb),
 683                         sbi->s_mount_opt);
 684         return res;
 685 }
 686
 687 static int ext2_check_descriptors(struct super_block *sb)
 688 {
 689         int i;
 690         struct ext2_sb_info *sbi = EXT2_SB(sb);
 691
 692         ext2_debug ("Checking group descriptors");
 693
 694         for (i = 0; i < sbi->s_groups_count; i++) {
 695                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 696                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 697                 ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
 698
 699                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 700                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 701                 {
 702                         ext2_error (sb, "ext2_check_descriptors",
 703                                     "Block bitmap for group %d"
 704                                     " not in group (block %lu)!",
 705                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 706                         return 0;
 707                 }
 708                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 709                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 710                 {
 711                         ext2_error (sb, "ext2_check_descriptors",
 712                                     "Inode bitmap for group %d"
 713                                     " not in group (block %lu)!",
 714                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 715                         return 0;
 716                 }
 717                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 718                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 719                     last_block)
 720                 {
 721                         ext2_error (sb, "ext2_check_descriptors",
 722                                     "Inode table for group %d"
 723                                     " not in group (block %lu)!",
 724                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 725                         return 0;
 726                 }
 727         }
 728         return 1;
 729 }
 730
 731 /*
 732  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 733  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 734  * We need to be 1 filesystem block less than the 2^32 sector limit.
 735  */
 736 static loff_t ext2_max_size(int bits)
 737 {
 738         loff_t res = EXT2_NDIR_BLOCKS;
 739         int meta_blocks;
 740         unsigned int upper_limit;
 741         unsigned int ppb = 1 << (bits-2);
 742
 743         /* This is calculated to be the largest file size for a
 744          * dense, file such that the total number of
 745          * sectors in the file, including data and all indirect blocks,
 746          * does not exceed 2^32 -1
 747          * __u32 i_blocks representing the total number of
 748          * 512 bytes blocks of the file
 749          */
 750         upper_limit = (1LL << 32) - 1;
 751
 752         /* total blocks in file system block size */
 753         upper_limit >>= (bits - 9);
 754
 755         /* Compute how many blocks we can address by block tree */
 756         res += 1LL << (bits-2);
 757         res += 1LL << (2*(bits-2));
 758         res += 1LL << (3*(bits-2));
 759         /* Does block tree limit file size? */
 760         if (res < upper_limit)
 761                 goto check_lfs;
 762
 763         res = upper_limit;
 764         /* How many metadata blocks are needed for addressing upper_limit? */
 765         upper_limit -= EXT2_NDIR_BLOCKS;
 766         /* indirect blocks */
 767         meta_blocks = 1;
 768         upper_limit -= ppb;
 769         /* double indirect blocks */
 770         if (upper_limit < ppb * ppb) {
 771                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 772                 res -= meta_blocks;
 773                 goto check_lfs;
 774         }
 775         meta_blocks += 1 + ppb;
 776         upper_limit -= ppb * ppb;
 777         /* tripple indirect blocks for the rest */
 778         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 779                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 780         res -= meta_blocks;
 781 check_lfs:
 782         res <<= bits;
 783         if (res > MAX_LFS_FILESIZE)
 784                 res = MAX_LFS_FILESIZE;
 785
 786         return res;
 787 }
 788
 789 static unsigned long descriptor_loc(struct super_block *sb,
 790                                     unsigned long logic_sb_block,
 791                                     int nr)
 792 {
 793         struct ext2_sb_info *sbi = EXT2_SB(sb);
 794         unsigned long bg, first_meta_bg;
 795
 796         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 797
 798         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 799             nr < first_meta_bg)
 800                 return (logic_sb_block + nr + 1);
 801         bg = sbi->s_desc_per_block * nr;
 802
 803         return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
 804 }
 805
 806 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 807 {
 808         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 809         struct buffer_head * bh;
 810         struct ext2_sb_info * sbi;
 811         struct ext2_super_block * es;
 812         struct inode *root;
 813         unsigned long block;
 814         unsigned long sb_block = get_sb_block(&data);
 815         unsigned long logic_sb_block;
 816         unsigned long offset = 0;
 817         unsigned long def_mount_opts;
 818         long ret = -ENOMEM;
 819         int blocksize = BLOCK_SIZE;
 820         int db_count;
 821         int i, j;
 822         __le32 features;
 823         int err;
 824         struct ext2_mount_options opts;
 825
 826         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 827         if (!sbi)
 828                 goto failed;
 829
 830         sbi->s_blockgroup_lock =
 831                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 832         if (!sbi->s_blockgroup_lock) {
 833                 kfree(sbi);
 834                 goto failed;
 835         }
 836         sb->s_fs_info = sbi;
 837         sbi->s_sb_block = sb_block;
 838         sbi->s_daxdev = dax_dev;
 839
 840         spin_lock_init(&sbi->s_lock);
 841         ret = -EINVAL;
 842
 843         /*
 844          * See what the current blocksize for the device is, and
 845          * use that as the blocksize.  Otherwise (or if the blocksize
 846          * is smaller than the default) use the default.
 847          * This is important for devices that have a hardware
 848          * sectorsize that is larger than the default.
 849          */
 850         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 851         if (!blocksize) {
 852                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 853                 goto failed_sbi;
 854         }
 855
 856         /*
 857          * If the superblock doesn't start on a hardware sector boundary,
 858          * calculate the offset.
 859          */
 860         if (blocksize != BLOCK_SIZE) {
 861                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 862                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 863         } else {
 864                 logic_sb_block = sb_block;
 865         }
 866
 867         if (!(bh = sb_bread(sb, logic_sb_block))) {
 868                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 869                 goto failed_sbi;
 870         }
 871         /*
 872          * Note: s_es must be initialized as soon as possible because
 873          *       some ext2 macro-instructions depend on its value
 874          */
 875         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 876         sbi->s_es = es;
 877         sb->s_magic = le16_to_cpu(es->s_magic);
 878
 879         if (sb->s_magic != EXT2_SUPER_MAGIC)
 880                 goto cantfind_ext2;
 881
 882         opts.s_mount_opt = 0;
 883         /* Set defaults before we parse the mount options */
 884         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 885         if (def_mount_opts & EXT2_DEFM_DEBUG)
 886                 set_opt(opts.s_mount_opt, DEBUG);
 887         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 888                 set_opt(opts.s_mount_opt, GRPID);
 889         if (def_mount_opts & EXT2_DEFM_UID16)
 890                 set_opt(opts.s_mount_opt, NO_UID32);
 891 #ifdef CONFIG_EXT2_FS_XATTR
 892         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 893                 set_opt(opts.s_mount_opt, XATTR_USER);
 894 #endif
 895 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 896         if (def_mount_opts & EXT2_DEFM_ACL)
 897                 set_opt(opts.s_mount_opt, POSIX_ACL);
 898 #endif
 899
 900         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 901                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 902         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 903                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 904         else
 905                 set_opt(opts.s_mount_opt, ERRORS_RO);
 906
 907         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 908         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 909
 910         set_opt(opts.s_mount_opt, RESERVATION);
 911
 912         if (!parse_options((char *) data, sb, &opts))
 913                 goto failed_mount;
 914
 915         sbi->s_mount_opt = opts.s_mount_opt;
 916         sbi->s_resuid = opts.s_resuid;
 917         sbi->s_resgid = opts.s_resgid;
 918
 919         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 920                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
 921         sb->s_iflags |= SB_I_CGROUPWB;
 922
 923         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 924             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 925              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 926              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 927                 ext2_msg(sb, KERN_WARNING,
 928                         "warning: feature flags set on rev 0 fs, "
 929                         "running e2fsck is recommended");
 930         /*
 931          * Check feature flags regardless of the revision level, since we
 932          * previously didn't change the revision level when setting the flags,
 933          * so there is a chance incompat flags are set on a rev 0 filesystem.
 934          */
 935         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 936         if (features) {
 937                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 938                        "unsupported optional features (%x)",
 939                         le32_to_cpu(features));
 940                 goto failed_mount;
 941         }
 942         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 943                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 944                        "unsupported optional features (%x)",
 945                        le32_to_cpu(features));
 946                 goto failed_mount;
 947         }
 948
 949         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 950
 951         if (test_opt(sb, DAX)) {
 952                 if (!dax_supported(dax_dev, sb->s_bdev, blocksize, 0,
 953                                 bdev_nr_sectors(sb->s_bdev))) {
 954                         ext2_msg(sb, KERN_ERR,
 955                                 "DAX unsupported by block device. Turning off DAX.");
 956                         clear_opt(sbi->s_mount_opt, DAX);
 957                 }
 958         }
 959
 960         /* If the blocksize doesn't match, re-read the thing.. */
 961         if (sb->s_blocksize != blocksize) {
 962                 brelse(bh);
 963
 964                 if (!sb_set_blocksize(sb, blocksize)) {
 965                         ext2_msg(sb, KERN_ERR,
 966                                 "error: bad blocksize %d", blocksize);
 967                         goto failed_sbi;
 968                 }
 969
 970                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 971                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 972                 bh = sb_bread(sb, logic_sb_block);
 973                 if(!bh) {
 974                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 975                                 "superblock on 2nd try");
 976                         goto failed_sbi;
 977                 }
 978                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 979                 sbi->s_es = es;
 980                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 981                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 982                         goto failed_mount;
 983                 }
 984         }
 985
 986         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 987         sb->s_max_links = EXT2_LINK_MAX;
 988         sb->s_time_min = S32_MIN;
 989         sb->s_time_max = S32_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         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1018         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1019                 goto cantfind_ext2;
1020         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1021                                         sbi->s_inodes_per_block;
1022         sbi->s_desc_per_block = sb->s_blocksize /
1023                                         sizeof (struct ext2_group_desc);
1024         sbi->s_sbh = bh;
1025         sbi->s_mount_state = le16_to_cpu(es->s_state);
1026         sbi->s_addr_per_block_bits =
1027                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1028         sbi->s_desc_per_block_bits =
1029                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1030
1031         if (sb->s_magic != EXT2_SUPER_MAGIC)
1032                 goto cantfind_ext2;
1033
1034         if (sb->s_blocksize != bh->b_size) {
1035                 if (!silent)
1036                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1037                 goto failed_mount;
1038         }
1039
1040         if (sb->s_blocksize != sbi->s_frag_size) {
1041                 ext2_msg(sb, KERN_ERR,
1042                         "error: fragsize %lu != blocksize %lu"
1043                         "(not supported yet)",
1044                         sbi->s_frag_size, sb->s_blocksize);
1045                 goto failed_mount;
1046         }
1047
1048         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1049                 ext2_msg(sb, KERN_ERR,
1050                         "error: #blocks per group too big: %lu",
1051                         sbi->s_blocks_per_group);
1052                 goto failed_mount;
1053         }
1054         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1055                 ext2_msg(sb, KERN_ERR,
1056                         "error: #fragments per group too big: %lu",
1057                         sbi->s_frags_per_group);
1058                 goto failed_mount;
1059         }
1060         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1061                 ext2_msg(sb, KERN_ERR,
1062                         "error: #inodes per group too big: %lu",
1063                         sbi->s_inodes_per_group);
1064                 goto failed_mount;
1065         }
1066
1067         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1068                 goto cantfind_ext2;
1069         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1070                                 le32_to_cpu(es->s_first_data_block) - 1)
1071                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1072         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1073                    EXT2_DESC_PER_BLOCK(sb);
1074         sbi->s_group_desc = kmalloc_array(db_count,
1075                                            sizeof(struct buffer_head *),
1076                                            GFP_KERNEL);
1077         if (sbi->s_group_desc == NULL) {
1078                 ret = -ENOMEM;
1079                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1080                 goto failed_mount;
1081         }
1082         bgl_lock_init(sbi->s_blockgroup_lock);
1083         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1084         if (!sbi->s_debts) {
1085                 ret = -ENOMEM;
1086                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1087                 goto failed_mount_group_desc;
1088         }
1089         for (i = 0; i < db_count; i++) {
1090                 block = descriptor_loc(sb, logic_sb_block, i);
1091                 sbi->s_group_desc[i] = sb_bread(sb, block);
1092                 if (!sbi->s_group_desc[i]) {
1093                         for (j = 0; j < i; j++)
1094                                 brelse (sbi->s_group_desc[j]);
1095                         ext2_msg(sb, KERN_ERR,
1096                                 "error: unable to read group descriptors");
1097                         goto failed_mount_group_desc;
1098                 }
1099         }
1100         if (!ext2_check_descriptors (sb)) {
1101                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1102                 goto failed_mount2;
1103         }
1104         sbi->s_gdb_count = db_count;
1105         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1106         spin_lock_init(&sbi->s_next_gen_lock);
1107
1108         /* per filesystem reservation list head & lock */
1109         spin_lock_init(&sbi->s_rsv_window_lock);
1110         sbi->s_rsv_window_root = RB_ROOT;
1111         /*
1112          * Add a single, static dummy reservation to the start of the
1113          * reservation window list --- it gives us a placeholder for
1114          * append-at-start-of-list which makes the allocation logic
1115          * _much_ simpler.
1116          */
1117         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1118         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1119         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1120         sbi->s_rsv_window_head.rsv_goal_size = 0;
1121         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1122
1123         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1124                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1125         if (!err) {
1126                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1127                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1128         }
1129         if (!err) {
1130                 err = percpu_counter_init(&sbi->s_dirs_counter,
1131                                 ext2_count_dirs(sb), GFP_KERNEL);
1132         }
1133         if (err) {
1134                 ret = err;
1135                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1136                 goto failed_mount3;
1137         }
1138
1139 #ifdef CONFIG_EXT2_FS_XATTR
1140         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1141         if (!sbi->s_ea_block_cache) {
1142                 ret = -ENOMEM;
1143                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1144                 goto failed_mount3;
1145         }
1146 #endif
1147         /*
1148          * set up enough so that it can read an inode
1149          */
1150         sb->s_op = &ext2_sops;
1151         sb->s_export_op = &ext2_export_ops;
1152         sb->s_xattr = ext2_xattr_handlers;
1153
1154 #ifdef CONFIG_QUOTA
1155         sb->dq_op = &dquot_operations;
1156         sb->s_qcop = &ext2_quotactl_ops;
1157         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1158 #endif
1159
1160         root = ext2_iget(sb, EXT2_ROOT_INO);
1161         if (IS_ERR(root)) {
1162                 ret = PTR_ERR(root);
1163                 goto failed_mount3;
1164         }
1165         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1166                 iput(root);
1167                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1168                 goto failed_mount3;
1169         }
1170
1171         sb->s_root = d_make_root(root);
1172         if (!sb->s_root) {
1173                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1174                 ret = -ENOMEM;
1175                 goto failed_mount3;
1176         }
1177         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1178                 ext2_msg(sb, KERN_WARNING,
1179                         "warning: mounting ext3 filesystem as ext2");
1180         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1181                 sb->s_flags |= SB_RDONLY;
1182         ext2_write_super(sb);
1183         return 0;
1184
1185 cantfind_ext2:
1186         if (!silent)
1187                 ext2_msg(sb, KERN_ERR,
1188                         "error: can't find an ext2 filesystem on dev %s.",
1189                         sb->s_id);
1190         goto failed_mount;
1191 failed_mount3:
1192         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1193         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1194         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1195         percpu_counter_destroy(&sbi->s_dirs_counter);
1196 failed_mount2:
1197         for (i = 0; i < db_count; i++)
1198                 brelse(sbi->s_group_desc[i]);
1199 failed_mount_group_desc:
1200         kfree(sbi->s_group_desc);
1201         kfree(sbi->s_debts);
1202 failed_mount:
1203         brelse(bh);
1204 failed_sbi:
1205         sb->s_fs_info = NULL;
1206         kfree(sbi->s_blockgroup_lock);
1207         kfree(sbi);
1208 failed:
1209         fs_put_dax(dax_dev);
1210         return ret;
1211 }
1212
1213 static void ext2_clear_super_error(struct super_block *sb)
1214 {
1215         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1216
1217         if (buffer_write_io_error(sbh)) {
1218                 /*
1219                  * Oh, dear.  A previous attempt to write the
1220                  * superblock failed.  This could happen because the
1221                  * USB device was yanked out.  Or it could happen to
1222                  * be a transient write error and maybe the block will
1223                  * be remapped.  Nothing we can do but to retry the
1224                  * write and hope for the best.
1225                  */
1226                 ext2_msg(sb, KERN_ERR,
1227                        "previous I/O error to superblock detected");
1228                 clear_buffer_write_io_error(sbh);
1229                 set_buffer_uptodate(sbh);
1230         }
1231 }
1232
1233 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1234                      int wait)
1235 {
1236         ext2_clear_super_error(sb);
1237         spin_lock(&EXT2_SB(sb)->s_lock);
1238         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1239         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1240         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1241         /* unlock before we do IO */
1242         spin_unlock(&EXT2_SB(sb)->s_lock);
1243         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1244         if (wait)
1245                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1246 }
1247
1248 /*
1249  * In the second extended file system, it is not necessary to
1250  * write the super block since we use a mapping of the
1251  * disk super block in a buffer.
1252  *
1253  * However, this function is still used to set the fs valid
1254  * flags to 0.  We need to set this flag to 0 since the fs
1255  * may have been checked while mounted and e2fsck may have
1256  * set s_state to EXT2_VALID_FS after some corrections.
1257  */
1258 static int ext2_sync_fs(struct super_block *sb, int wait)
1259 {
1260         struct ext2_sb_info *sbi = EXT2_SB(sb);
1261         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1262
1263         /*
1264          * Write quota structures to quota file, sync_blockdev() will write
1265          * them to disk later
1266          */
1267         dquot_writeback_dquots(sb, -1);
1268
1269         spin_lock(&sbi->s_lock);
1270         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1271                 ext2_debug("setting valid to 0\n");
1272                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1273         }
1274         spin_unlock(&sbi->s_lock);
1275         ext2_sync_super(sb, es, wait);
1276         return 0;
1277 }
1278
1279 static int ext2_freeze(struct super_block *sb)
1280 {
1281         struct ext2_sb_info *sbi = EXT2_SB(sb);
1282
1283         /*
1284          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1285          * because we have unattached inodes and thus filesystem is not fully
1286          * consistent.
1287          */
1288         if (atomic_long_read(&sb->s_remove_count)) {
1289                 ext2_sync_fs(sb, 1);
1290                 return 0;
1291         }
1292         /* Set EXT2_FS_VALID flag */
1293         spin_lock(&sbi->s_lock);
1294         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1295         spin_unlock(&sbi->s_lock);
1296         ext2_sync_super(sb, sbi->s_es, 1);
1297
1298         return 0;
1299 }
1300
1301 static int ext2_unfreeze(struct super_block *sb)
1302 {
1303         /* Just write sb to clear EXT2_VALID_FS flag */
1304         ext2_write_super(sb);
1305
1306         return 0;
1307 }
1308
1309 static void ext2_write_super(struct super_block *sb)
1310 {
1311         if (!sb_rdonly(sb))
1312                 ext2_sync_fs(sb, 1);
1313 }
1314
1315 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1316 {
1317         struct ext2_sb_info * sbi = EXT2_SB(sb);
1318         struct ext2_super_block * es;
1319         struct ext2_mount_options new_opts;
1320         int err;
1321
1322         sync_filesystem(sb);
1323
1324         spin_lock(&sbi->s_lock);
1325         new_opts.s_mount_opt = sbi->s_mount_opt;
1326         new_opts.s_resuid = sbi->s_resuid;
1327         new_opts.s_resgid = sbi->s_resgid;
1328         spin_unlock(&sbi->s_lock);
1329
1330         if (!parse_options(data, sb, &new_opts))
1331                 return -EINVAL;
1332
1333         spin_lock(&sbi->s_lock);
1334         es = sbi->s_es;
1335         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1336                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1337                          "dax flag with busy inodes while remounting");
1338                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1339         }
1340         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1341                 goto out_set;
1342         if (*flags & SB_RDONLY) {
1343                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1344                     !(sbi->s_mount_state & EXT2_VALID_FS))
1345                         goto out_set;
1346
1347                 /*
1348                  * OK, we are remounting a valid rw partition rdonly, so set
1349                  * the rdonly flag and then mark the partition as valid again.
1350                  */
1351                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1352                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1353                 spin_unlock(&sbi->s_lock);
1354
1355                 err = dquot_suspend(sb, -1);
1356                 if (err < 0)
1357                         return err;
1358
1359                 ext2_sync_super(sb, es, 1);
1360         } else {
1361                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1362                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1363                 if (ret) {
1364                         spin_unlock(&sbi->s_lock);
1365                         ext2_msg(sb, KERN_WARNING,
1366                                 "warning: couldn't remount RDWR because of "
1367                                 "unsupported optional features (%x).",
1368                                 le32_to_cpu(ret));
1369                         return -EROFS;
1370                 }
1371                 /*
1372                  * Mounting a RDONLY partition read-write, so reread and
1373                  * store the current valid flag.  (It may have been changed
1374                  * by e2fsck since we originally mounted the partition.)
1375                  */
1376                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1377                 if (!ext2_setup_super (sb, es, 0))
1378                         sb->s_flags &= ~SB_RDONLY;
1379                 spin_unlock(&sbi->s_lock);
1380
1381                 ext2_write_super(sb);
1382
1383                 dquot_resume(sb, -1);
1384         }
1385
1386         spin_lock(&sbi->s_lock);
1387 out_set:
1388         sbi->s_mount_opt = new_opts.s_mount_opt;
1389         sbi->s_resuid = new_opts.s_resuid;
1390         sbi->s_resgid = new_opts.s_resgid;
1391         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1392                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1393         spin_unlock(&sbi->s_lock);
1394
1395         return 0;
1396 }
1397
1398 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1399 {
1400         struct super_block *sb = dentry->d_sb;
1401         struct ext2_sb_info *sbi = EXT2_SB(sb);
1402         struct ext2_super_block *es = sbi->s_es;
1403
1404         spin_lock(&sbi->s_lock);
1405
1406         if (test_opt (sb, MINIX_DF))
1407                 sbi->s_overhead_last = 0;
1408         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1409                 unsigned long i, overhead = 0;
1410                 smp_rmb();
1411
1412                 /*
1413                  * Compute the overhead (FS structures). This is constant
1414                  * for a given filesystem unless the number of block groups
1415                  * changes so we cache the previous value until it does.
1416                  */
1417
1418                 /*
1419                  * All of the blocks before first_data_block are
1420                  * overhead
1421                  */
1422                 overhead = le32_to_cpu(es->s_first_data_block);
1423
1424                 /*
1425                  * Add the overhead attributed to the superblock and
1426                  * block group descriptors.  If the sparse superblocks
1427                  * feature is turned on, then not all groups have this.
1428                  */
1429                 for (i = 0; i < sbi->s_groups_count; i++)
1430                         overhead += ext2_bg_has_super(sb, i) +
1431                                 ext2_bg_num_gdb(sb, i);
1432
1433                 /*
1434                  * Every block group has an inode bitmap, a block
1435                  * bitmap, and an inode table.
1436                  */
1437                 overhead += (sbi->s_groups_count *
1438                              (2 + sbi->s_itb_per_group));
1439                 sbi->s_overhead_last = overhead;
1440                 smp_wmb();
1441                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1442         }
1443
1444         buf->f_type = EXT2_SUPER_MAGIC;
1445         buf->f_bsize = sb->s_blocksize;
1446         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1447         buf->f_bfree = ext2_count_free_blocks(sb);
1448         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1449         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1450         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1451                 buf->f_bavail = 0;
1452         buf->f_files = le32_to_cpu(es->s_inodes_count);
1453         buf->f_ffree = ext2_count_free_inodes(sb);
1454         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1455         buf->f_namelen = EXT2_NAME_LEN;
1456         buf->f_fsid = uuid_to_fsid(es->s_uuid);
1457         spin_unlock(&sbi->s_lock);
1458         return 0;
1459 }
1460
1461 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1462         int flags, const char *dev_name, void *data)
1463 {
1464         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1465 }
1466
1467 #ifdef CONFIG_QUOTA
1468
1469 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1470  * acquiring the locks... As quota files are never truncated and quota code
1471  * itself serializes the operations (and no one else should touch the files)
1472  * we don't have to be afraid of races */
1473 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1474                                size_t len, loff_t off)
1475 {
1476         struct inode *inode = sb_dqopt(sb)->files[type];
1477         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1478         int err = 0;
1479         int offset = off & (sb->s_blocksize - 1);
1480         int tocopy;
1481         size_t toread;
1482         struct buffer_head tmp_bh;
1483         struct buffer_head *bh;
1484         loff_t i_size = i_size_read(inode);
1485
1486         if (off > i_size)
1487                 return 0;
1488         if (off+len > i_size)
1489                 len = i_size-off;
1490         toread = len;
1491         while (toread > 0) {
1492                 tocopy = sb->s_blocksize - offset < toread ?
1493                                 sb->s_blocksize - offset : toread;
1494
1495                 tmp_bh.b_state = 0;
1496                 tmp_bh.b_size = sb->s_blocksize;
1497                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1498                 if (err < 0)
1499                         return err;
1500                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1501                         memset(data, 0, tocopy);
1502                 else {
1503                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1504                         if (!bh)
1505                                 return -EIO;
1506                         memcpy(data, bh->b_data+offset, tocopy);
1507                         brelse(bh);
1508                 }
1509                 offset = 0;
1510                 toread -= tocopy;
1511                 data += tocopy;
1512                 blk++;
1513         }
1514         return len;
1515 }
1516
1517 /* Write to quotafile */
1518 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1519                                 const char *data, size_t len, loff_t off)
1520 {
1521         struct inode *inode = sb_dqopt(sb)->files[type];
1522         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1523         int err = 0;
1524         int offset = off & (sb->s_blocksize - 1);
1525         int tocopy;
1526         size_t towrite = len;
1527         struct buffer_head tmp_bh;
1528         struct buffer_head *bh;
1529
1530         while (towrite > 0) {
1531                 tocopy = sb->s_blocksize - offset < towrite ?
1532                                 sb->s_blocksize - offset : towrite;
1533
1534                 tmp_bh.b_state = 0;
1535                 tmp_bh.b_size = sb->s_blocksize;
1536                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1537                 if (err < 0)
1538                         goto out;
1539                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1540                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1541                 else
1542                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1543                 if (unlikely(!bh)) {
1544                         err = -EIO;
1545                         goto out;
1546                 }
1547                 lock_buffer(bh);
1548                 memcpy(bh->b_data+offset, data, tocopy);
1549                 flush_dcache_page(bh->b_page);
1550                 set_buffer_uptodate(bh);
1551                 mark_buffer_dirty(bh);
1552                 unlock_buffer(bh);
1553                 brelse(bh);
1554                 offset = 0;
1555                 towrite -= tocopy;
1556                 data += tocopy;
1557                 blk++;
1558         }
1559 out:
1560         if (len == towrite)
1561                 return err;
1562         if (inode->i_size < off+len-towrite)
1563                 i_size_write(inode, off+len-towrite);
1564         inode_inc_iversion(inode);
1565         inode->i_mtime = inode->i_ctime = current_time(inode);
1566         mark_inode_dirty(inode);
1567         return len - towrite;
1568 }
1569
1570 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1571                          const struct path *path)
1572 {
1573         int err;
1574         struct inode *inode;
1575
1576         err = dquot_quota_on(sb, type, format_id, path);
1577         if (err)
1578                 return err;
1579
1580         inode = d_inode(path->dentry);
1581         inode_lock(inode);
1582         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1583         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1584                         S_NOATIME | S_IMMUTABLE);
1585         inode_unlock(inode);
1586         mark_inode_dirty(inode);
1587
1588         return 0;
1589 }
1590
1591 static int ext2_quota_off(struct super_block *sb, int type)
1592 {
1593         struct inode *inode = sb_dqopt(sb)->files[type];
1594         int err;
1595
1596         if (!inode || !igrab(inode))
1597                 goto out;
1598
1599         err = dquot_quota_off(sb, type);
1600         if (err)
1601                 goto out_put;
1602
1603         inode_lock(inode);
1604         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1605         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1606         inode_unlock(inode);
1607         mark_inode_dirty(inode);
1608 out_put:
1609         iput(inode);
1610         return err;
1611 out:
1612         return dquot_quota_off(sb, type);
1613 }
1614
1615 #endif
1616
1617 static struct file_system_type ext2_fs_type = {
1618         .owner          = THIS_MODULE,
1619         .name           = "ext2",
1620         .mount          = ext2_mount,
1621         .kill_sb        = kill_block_super,
1622         .fs_flags       = FS_REQUIRES_DEV,
1623 };
1624 MODULE_ALIAS_FS("ext2");
1625
1626 static int __init init_ext2_fs(void)
1627 {
1628         int err;
1629
1630         err = init_inodecache();
1631         if (err)
1632                 return err;
1633         err = register_filesystem(&ext2_fs_type);
1634         if (err)
1635                 goto out;
1636         return 0;
1637 out:
1638         destroy_inodecache();
1639         return err;
1640 }
1641
1642 static void __exit exit_ext2_fs(void)
1643 {
1644         unregister_filesystem(&ext2_fs_type);
1645         destroy_inodecache();
1646 }
1647
1648 MODULE_AUTHOR("Remy Card and others");
1649 MODULE_DESCRIPTION("Second Extended Filesystem");
1650 MODULE_LICENSE("GPL");
1651 module_init(init_ext2_fs)
1652 module_exit(exit_ext2_fs)