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, fs=%lu, gc=%lu, "
 685                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 686                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 687                         sbi->s_frag_size,
 688                         sbi->s_groups_count,
 689                         EXT2_BLOCKS_PER_GROUP(sb),
 690                         EXT2_INODES_PER_GROUP(sb),
 691                         sbi->s_mount_opt);
 692         return res;
 693 }
 694
 695 static int ext2_check_descriptors(struct super_block *sb)
 696 {
 697         int i;
 698         struct ext2_sb_info *sbi = EXT2_SB(sb);
 699
 700         ext2_debug ("Checking group descriptors");
 701
 702         for (i = 0; i < sbi->s_groups_count; i++) {
 703                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 704                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 705                 ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
 706
 707                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 708                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 709                 {
 710                         ext2_error (sb, "ext2_check_descriptors",
 711                                     "Block bitmap for group %d"
 712                                     " not in group (block %lu)!",
 713                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 714                         return 0;
 715                 }
 716                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 717                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 718                 {
 719                         ext2_error (sb, "ext2_check_descriptors",
 720                                     "Inode bitmap for group %d"
 721                                     " not in group (block %lu)!",
 722                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 723                         return 0;
 724                 }
 725                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 726                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 727                     last_block)
 728                 {
 729                         ext2_error (sb, "ext2_check_descriptors",
 730                                     "Inode table for group %d"
 731                                     " not in group (block %lu)!",
 732                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 733                         return 0;
 734                 }
 735         }
 736         return 1;
 737 }
 738
 739 /*
 740  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 741  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 742  * We need to be 1 filesystem block less than the 2^32 sector limit.
 743  */
 744 static loff_t ext2_max_size(int bits)
 745 {
 746         loff_t res = EXT2_NDIR_BLOCKS;
 747         int meta_blocks;
 748         unsigned int upper_limit;
 749         unsigned int ppb = 1 << (bits-2);
 750
 751         /* This is calculated to be the largest file size for a
 752          * dense, file such that the total number of
 753          * sectors in the file, including data and all indirect blocks,
 754          * does not exceed 2^32 -1
 755          * __u32 i_blocks representing the total number of
 756          * 512 bytes blocks of the file
 757          */
 758         upper_limit = (1LL << 32) - 1;
 759
 760         /* total blocks in file system block size */
 761         upper_limit >>= (bits - 9);
 762
 763         /* Compute how many blocks we can address by block tree */
 764         res += 1LL << (bits-2);
 765         res += 1LL << (2*(bits-2));
 766         res += 1LL << (3*(bits-2));
 767         /* Does block tree limit file size? */
 768         if (res < upper_limit)
 769                 goto check_lfs;
 770
 771         res = upper_limit;
 772         /* How many metadata blocks are needed for addressing upper_limit? */
 773         upper_limit -= EXT2_NDIR_BLOCKS;
 774         /* indirect blocks */
 775         meta_blocks = 1;
 776         upper_limit -= ppb;
 777         /* double indirect blocks */
 778         if (upper_limit < ppb * ppb) {
 779                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 780                 res -= meta_blocks;
 781                 goto check_lfs;
 782         }
 783         meta_blocks += 1 + ppb;
 784         upper_limit -= ppb * ppb;
 785         /* tripple indirect blocks for the rest */
 786         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 787                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 788         res -= meta_blocks;
 789 check_lfs:
 790         res <<= bits;
 791         if (res > MAX_LFS_FILESIZE)
 792                 res = MAX_LFS_FILESIZE;
 793
 794         return res;
 795 }
 796
 797 static unsigned long descriptor_loc(struct super_block *sb,
 798                                     unsigned long logic_sb_block,
 799                                     int nr)
 800 {
 801         struct ext2_sb_info *sbi = EXT2_SB(sb);
 802         unsigned long bg, first_meta_bg;
 803
 804         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 805
 806         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 807             nr < first_meta_bg)
 808                 return (logic_sb_block + nr + 1);
 809         bg = sbi->s_desc_per_block * nr;
 810
 811         return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
 812 }
 813
 814 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 815 {
 816         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 817         struct buffer_head * bh;
 818         struct ext2_sb_info * sbi;
 819         struct ext2_super_block * es;
 820         struct inode *root;
 821         unsigned long block;
 822         unsigned long sb_block = get_sb_block(&data);
 823         unsigned long logic_sb_block;
 824         unsigned long offset = 0;
 825         unsigned long def_mount_opts;
 826         long ret = -ENOMEM;
 827         int blocksize = BLOCK_SIZE;
 828         int db_count;
 829         int i, j;
 830         __le32 features;
 831         int err;
 832         struct ext2_mount_options opts;
 833
 834         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 835         if (!sbi)
 836                 goto failed;
 837
 838         sbi->s_blockgroup_lock =
 839                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 840         if (!sbi->s_blockgroup_lock) {
 841                 kfree(sbi);
 842                 goto failed;
 843         }
 844         sb->s_fs_info = sbi;
 845         sbi->s_sb_block = sb_block;
 846         sbi->s_daxdev = dax_dev;
 847
 848         spin_lock_init(&sbi->s_lock);
 849         ret = -EINVAL;
 850
 851         /*
 852          * See what the current blocksize for the device is, and
 853          * use that as the blocksize.  Otherwise (or if the blocksize
 854          * is smaller than the default) use the default.
 855          * This is important for devices that have a hardware
 856          * sectorsize that is larger than the default.
 857          */
 858         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 859         if (!blocksize) {
 860                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 861                 goto failed_sbi;
 862         }
 863
 864         /*
 865          * If the superblock doesn't start on a hardware sector boundary,
 866          * calculate the offset.
 867          */
 868         if (blocksize != BLOCK_SIZE) {
 869                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 870                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 871         } else {
 872                 logic_sb_block = sb_block;
 873         }
 874
 875         if (!(bh = sb_bread(sb, logic_sb_block))) {
 876                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 877                 goto failed_sbi;
 878         }
 879         /*
 880          * Note: s_es must be initialized as soon as possible because
 881          *       some ext2 macro-instructions depend on its value
 882          */
 883         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 884         sbi->s_es = es;
 885         sb->s_magic = le16_to_cpu(es->s_magic);
 886
 887         if (sb->s_magic != EXT2_SUPER_MAGIC)
 888                 goto cantfind_ext2;
 889
 890         opts.s_mount_opt = 0;
 891         /* Set defaults before we parse the mount options */
 892         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 893         if (def_mount_opts & EXT2_DEFM_DEBUG)
 894                 set_opt(opts.s_mount_opt, DEBUG);
 895         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 896                 set_opt(opts.s_mount_opt, GRPID);
 897         if (def_mount_opts & EXT2_DEFM_UID16)
 898                 set_opt(opts.s_mount_opt, NO_UID32);
 899 #ifdef CONFIG_EXT2_FS_XATTR
 900         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 901                 set_opt(opts.s_mount_opt, XATTR_USER);
 902 #endif
 903 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 904         if (def_mount_opts & EXT2_DEFM_ACL)
 905                 set_opt(opts.s_mount_opt, POSIX_ACL);
 906 #endif
 907
 908         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 909                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 910         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 911                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 912         else
 913                 set_opt(opts.s_mount_opt, ERRORS_RO);
 914
 915         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 916         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 917
 918         set_opt(opts.s_mount_opt, RESERVATION);
 919
 920         if (!parse_options((char *) data, sb, &opts))
 921                 goto failed_mount;
 922
 923         sbi->s_mount_opt = opts.s_mount_opt;
 924         sbi->s_resuid = opts.s_resuid;
 925         sbi->s_resgid = opts.s_resgid;
 926
 927         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 928                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
 929         sb->s_iflags |= SB_I_CGROUPWB;
 930
 931         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 932             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 933              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 934              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 935                 ext2_msg(sb, KERN_WARNING,
 936                         "warning: feature flags set on rev 0 fs, "
 937                         "running e2fsck is recommended");
 938         /*
 939          * Check feature flags regardless of the revision level, since we
 940          * previously didn't change the revision level when setting the flags,
 941          * so there is a chance incompat flags are set on a rev 0 filesystem.
 942          */
 943         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 944         if (features) {
 945                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 946                        "unsupported optional features (%x)",
 947                         le32_to_cpu(features));
 948                 goto failed_mount;
 949         }
 950         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 951                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 952                        "unsupported optional features (%x)",
 953                        le32_to_cpu(features));
 954                 goto failed_mount;
 955         }
 956
 957         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 958
 959         if (test_opt(sb, DAX)) {
 960                 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
 961                         ext2_msg(sb, KERN_ERR,
 962                                 "DAX unsupported by block device. Turning off DAX.");
 963                         clear_opt(sbi->s_mount_opt, DAX);
 964                 }
 965         }
 966
 967         /* If the blocksize doesn't match, re-read the thing.. */
 968         if (sb->s_blocksize != blocksize) {
 969                 brelse(bh);
 970
 971                 if (!sb_set_blocksize(sb, blocksize)) {
 972                         ext2_msg(sb, KERN_ERR,
 973                                 "error: bad blocksize %d", blocksize);
 974                         goto failed_sbi;
 975                 }
 976
 977                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 978                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 979                 bh = sb_bread(sb, logic_sb_block);
 980                 if(!bh) {
 981                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 982                                 "superblock on 2nd try");
 983                         goto failed_sbi;
 984                 }
 985                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 986                 sbi->s_es = es;
 987                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 988                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 989                         goto failed_mount;
 990                 }
 991         }
 992
 993         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 994         sb->s_max_links = EXT2_LINK_MAX;
 995         sb->s_time_min = S32_MIN;
 996         sb->s_time_max = S32_MAX;
 997
 998         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 999                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1000                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1001         } else {
1002                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1003                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1004                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1005                     !is_power_of_2(sbi->s_inode_size) ||
1006                     (sbi->s_inode_size > blocksize)) {
1007                         ext2_msg(sb, KERN_ERR,
1008                                 "error: unsupported inode size: %d",
1009                                 sbi->s_inode_size);
1010                         goto failed_mount;
1011                 }
1012         }
1013
1014         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1015                                    le32_to_cpu(es->s_log_frag_size);
1016         if (sbi->s_frag_size == 0)
1017                 goto cantfind_ext2;
1018         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1019
1020         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1021         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1022         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1023
1024         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1025         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1026                 goto cantfind_ext2;
1027         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1028                                         sbi->s_inodes_per_block;
1029         sbi->s_desc_per_block = sb->s_blocksize /
1030                                         sizeof (struct ext2_group_desc);
1031         sbi->s_sbh = bh;
1032         sbi->s_mount_state = le16_to_cpu(es->s_state);
1033         sbi->s_addr_per_block_bits =
1034                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1035         sbi->s_desc_per_block_bits =
1036                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1037
1038         if (sb->s_magic != EXT2_SUPER_MAGIC)
1039                 goto cantfind_ext2;
1040
1041         if (sb->s_blocksize != bh->b_size) {
1042                 if (!silent)
1043                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1044                 goto failed_mount;
1045         }
1046
1047         if (sb->s_blocksize != sbi->s_frag_size) {
1048                 ext2_msg(sb, KERN_ERR,
1049                         "error: fragsize %lu != blocksize %lu"
1050                         "(not supported yet)",
1051                         sbi->s_frag_size, sb->s_blocksize);
1052                 goto failed_mount;
1053         }
1054
1055         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1056                 ext2_msg(sb, KERN_ERR,
1057                         "error: #blocks per group too big: %lu",
1058                         sbi->s_blocks_per_group);
1059                 goto failed_mount;
1060         }
1061         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1062                 ext2_msg(sb, KERN_ERR,
1063                         "error: #fragments per group too big: %lu",
1064                         sbi->s_frags_per_group);
1065                 goto failed_mount;
1066         }
1067         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1068                 ext2_msg(sb, KERN_ERR,
1069                         "error: #inodes per group too big: %lu",
1070                         sbi->s_inodes_per_group);
1071                 goto failed_mount;
1072         }
1073
1074         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1075                 goto cantfind_ext2;
1076         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1077                                 le32_to_cpu(es->s_first_data_block) - 1)
1078                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1079         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1080                    EXT2_DESC_PER_BLOCK(sb);
1081         sbi->s_group_desc = kmalloc_array (db_count,
1082                                            sizeof(struct buffer_head *),
1083                                            GFP_KERNEL);
1084         if (sbi->s_group_desc == NULL) {
1085                 ret = -ENOMEM;
1086                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1087                 goto failed_mount;
1088         }
1089         bgl_lock_init(sbi->s_blockgroup_lock);
1090         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1091         if (!sbi->s_debts) {
1092                 ret = -ENOMEM;
1093                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1094                 goto failed_mount_group_desc;
1095         }
1096         for (i = 0; i < db_count; i++) {
1097                 block = descriptor_loc(sb, logic_sb_block, i);
1098                 sbi->s_group_desc[i] = sb_bread(sb, block);
1099                 if (!sbi->s_group_desc[i]) {
1100                         for (j = 0; j < i; j++)
1101                                 brelse (sbi->s_group_desc[j]);
1102                         ext2_msg(sb, KERN_ERR,
1103                                 "error: unable to read group descriptors");
1104                         goto failed_mount_group_desc;
1105                 }
1106         }
1107         if (!ext2_check_descriptors (sb)) {
1108                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1109                 goto failed_mount2;
1110         }
1111         sbi->s_gdb_count = db_count;
1112         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1113         spin_lock_init(&sbi->s_next_gen_lock);
1114
1115         /* per fileystem reservation list head & lock */
1116         spin_lock_init(&sbi->s_rsv_window_lock);
1117         sbi->s_rsv_window_root = RB_ROOT;
1118         /*
1119          * Add a single, static dummy reservation to the start of the
1120          * reservation window list --- it gives us a placeholder for
1121          * append-at-start-of-list which makes the allocation logic
1122          * _much_ simpler.
1123          */
1124         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1125         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1126         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1127         sbi->s_rsv_window_head.rsv_goal_size = 0;
1128         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1129
1130         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1131                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1132         if (!err) {
1133                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1134                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1135         }
1136         if (!err) {
1137                 err = percpu_counter_init(&sbi->s_dirs_counter,
1138                                 ext2_count_dirs(sb), GFP_KERNEL);
1139         }
1140         if (err) {
1141                 ret = err;
1142                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1143                 goto failed_mount3;
1144         }
1145
1146 #ifdef CONFIG_EXT2_FS_XATTR
1147         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1148         if (!sbi->s_ea_block_cache) {
1149                 ret = -ENOMEM;
1150                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1151                 goto failed_mount3;
1152         }
1153 #endif
1154         /*
1155          * set up enough so that it can read an inode
1156          */
1157         sb->s_op = &ext2_sops;
1158         sb->s_export_op = &ext2_export_ops;
1159         sb->s_xattr = ext2_xattr_handlers;
1160
1161 #ifdef CONFIG_QUOTA
1162         sb->dq_op = &dquot_operations;
1163         sb->s_qcop = &ext2_quotactl_ops;
1164         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1165 #endif
1166
1167         root = ext2_iget(sb, EXT2_ROOT_INO);
1168         if (IS_ERR(root)) {
1169                 ret = PTR_ERR(root);
1170                 goto failed_mount3;
1171         }
1172         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1173                 iput(root);
1174                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1175                 goto failed_mount3;
1176         }
1177
1178         sb->s_root = d_make_root(root);
1179         if (!sb->s_root) {
1180                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1181                 ret = -ENOMEM;
1182                 goto failed_mount3;
1183         }
1184         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1185                 ext2_msg(sb, KERN_WARNING,
1186                         "warning: mounting ext3 filesystem as ext2");
1187         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1188                 sb->s_flags |= SB_RDONLY;
1189         ext2_write_super(sb);
1190         return 0;
1191
1192 cantfind_ext2:
1193         if (!silent)
1194                 ext2_msg(sb, KERN_ERR,
1195                         "error: can't find an ext2 filesystem on dev %s.",
1196                         sb->s_id);
1197         goto failed_mount;
1198 failed_mount3:
1199         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1200         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1201         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1202         percpu_counter_destroy(&sbi->s_dirs_counter);
1203 failed_mount2:
1204         for (i = 0; i < db_count; i++)
1205                 brelse(sbi->s_group_desc[i]);
1206 failed_mount_group_desc:
1207         kfree(sbi->s_group_desc);
1208         kfree(sbi->s_debts);
1209 failed_mount:
1210         brelse(bh);
1211 failed_sbi:
1212         sb->s_fs_info = NULL;
1213         kfree(sbi->s_blockgroup_lock);
1214         kfree(sbi);
1215 failed:
1216         fs_put_dax(dax_dev);
1217         return ret;
1218 }
1219
1220 static void ext2_clear_super_error(struct super_block *sb)
1221 {
1222         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1223
1224         if (buffer_write_io_error(sbh)) {
1225                 /*
1226                  * Oh, dear.  A previous attempt to write the
1227                  * superblock failed.  This could happen because the
1228                  * USB device was yanked out.  Or it could happen to
1229                  * be a transient write error and maybe the block will
1230                  * be remapped.  Nothing we can do but to retry the
1231                  * write and hope for the best.
1232                  */
1233                 ext2_msg(sb, KERN_ERR,
1234                        "previous I/O error to superblock detected");
1235                 clear_buffer_write_io_error(sbh);
1236                 set_buffer_uptodate(sbh);
1237         }
1238 }
1239
1240 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1241                      int wait)
1242 {
1243         ext2_clear_super_error(sb);
1244         spin_lock(&EXT2_SB(sb)->s_lock);
1245         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1246         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1247         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1248         /* unlock before we do IO */
1249         spin_unlock(&EXT2_SB(sb)->s_lock);
1250         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1251         if (wait)
1252                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1253 }
1254
1255 /*
1256  * In the second extended file system, it is not necessary to
1257  * write the super block since we use a mapping of the
1258  * disk super block in a buffer.
1259  *
1260  * However, this function is still used to set the fs valid
1261  * flags to 0.  We need to set this flag to 0 since the fs
1262  * may have been checked while mounted and e2fsck may have
1263  * set s_state to EXT2_VALID_FS after some corrections.
1264  */
1265 static int ext2_sync_fs(struct super_block *sb, int wait)
1266 {
1267         struct ext2_sb_info *sbi = EXT2_SB(sb);
1268         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1269
1270         /*
1271          * Write quota structures to quota file, sync_blockdev() will write
1272          * them to disk later
1273          */
1274         dquot_writeback_dquots(sb, -1);
1275
1276         spin_lock(&sbi->s_lock);
1277         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1278                 ext2_debug("setting valid to 0\n");
1279                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1280         }
1281         spin_unlock(&sbi->s_lock);
1282         ext2_sync_super(sb, es, wait);
1283         return 0;
1284 }
1285
1286 static int ext2_freeze(struct super_block *sb)
1287 {
1288         struct ext2_sb_info *sbi = EXT2_SB(sb);
1289
1290         /*
1291          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1292          * because we have unattached inodes and thus filesystem is not fully
1293          * consistent.
1294          */
1295         if (atomic_long_read(&sb->s_remove_count)) {
1296                 ext2_sync_fs(sb, 1);
1297                 return 0;
1298         }
1299         /* Set EXT2_FS_VALID flag */
1300         spin_lock(&sbi->s_lock);
1301         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1302         spin_unlock(&sbi->s_lock);
1303         ext2_sync_super(sb, sbi->s_es, 1);
1304
1305         return 0;
1306 }
1307
1308 static int ext2_unfreeze(struct super_block *sb)
1309 {
1310         /* Just write sb to clear EXT2_VALID_FS flag */
1311         ext2_write_super(sb);
1312
1313         return 0;
1314 }
1315
1316 static void ext2_write_super(struct super_block *sb)
1317 {
1318         if (!sb_rdonly(sb))
1319                 ext2_sync_fs(sb, 1);
1320 }
1321
1322 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1323 {
1324         struct ext2_sb_info * sbi = EXT2_SB(sb);
1325         struct ext2_super_block * es;
1326         struct ext2_mount_options new_opts;
1327         int err;
1328
1329         sync_filesystem(sb);
1330
1331         spin_lock(&sbi->s_lock);
1332         new_opts.s_mount_opt = sbi->s_mount_opt;
1333         new_opts.s_resuid = sbi->s_resuid;
1334         new_opts.s_resgid = sbi->s_resgid;
1335         spin_unlock(&sbi->s_lock);
1336
1337         if (!parse_options(data, sb, &new_opts))
1338                 return -EINVAL;
1339
1340         spin_lock(&sbi->s_lock);
1341         es = sbi->s_es;
1342         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1343                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1344                          "dax flag with busy inodes while remounting");
1345                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1346         }
1347         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1348                 goto out_set;
1349         if (*flags & SB_RDONLY) {
1350                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1351                     !(sbi->s_mount_state & EXT2_VALID_FS))
1352                         goto out_set;
1353
1354                 /*
1355                  * OK, we are remounting a valid rw partition rdonly, so set
1356                  * the rdonly flag and then mark the partition as valid again.
1357                  */
1358                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1359                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1360                 spin_unlock(&sbi->s_lock);
1361
1362                 err = dquot_suspend(sb, -1);
1363                 if (err < 0)
1364                         return err;
1365
1366                 ext2_sync_super(sb, es, 1);
1367         } else {
1368                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1369                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1370                 if (ret) {
1371                         spin_unlock(&sbi->s_lock);
1372                         ext2_msg(sb, KERN_WARNING,
1373                                 "warning: couldn't remount RDWR because of "
1374                                 "unsupported optional features (%x).",
1375                                 le32_to_cpu(ret));
1376                         return -EROFS;
1377                 }
1378                 /*
1379                  * Mounting a RDONLY partition read-write, so reread and
1380                  * store the current valid flag.  (It may have been changed
1381                  * by e2fsck since we originally mounted the partition.)
1382                  */
1383                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1384                 if (!ext2_setup_super (sb, es, 0))
1385                         sb->s_flags &= ~SB_RDONLY;
1386                 spin_unlock(&sbi->s_lock);
1387
1388                 ext2_write_super(sb);
1389
1390                 dquot_resume(sb, -1);
1391         }
1392
1393         spin_lock(&sbi->s_lock);
1394 out_set:
1395         sbi->s_mount_opt = new_opts.s_mount_opt;
1396         sbi->s_resuid = new_opts.s_resuid;
1397         sbi->s_resgid = new_opts.s_resgid;
1398         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1399                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1400         spin_unlock(&sbi->s_lock);
1401
1402         return 0;
1403 }
1404
1405 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1406 {
1407         struct super_block *sb = dentry->d_sb;
1408         struct ext2_sb_info *sbi = EXT2_SB(sb);
1409         struct ext2_super_block *es = sbi->s_es;
1410         u64 fsid;
1411
1412         spin_lock(&sbi->s_lock);
1413
1414         if (test_opt (sb, MINIX_DF))
1415                 sbi->s_overhead_last = 0;
1416         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1417                 unsigned long i, overhead = 0;
1418                 smp_rmb();
1419
1420                 /*
1421                  * Compute the overhead (FS structures). This is constant
1422                  * for a given filesystem unless the number of block groups
1423                  * changes so we cache the previous value until it does.
1424                  */
1425
1426                 /*
1427                  * All of the blocks before first_data_block are
1428                  * overhead
1429                  */
1430                 overhead = le32_to_cpu(es->s_first_data_block);
1431
1432                 /*
1433                  * Add the overhead attributed to the superblock and
1434                  * block group descriptors.  If the sparse superblocks
1435                  * feature is turned on, then not all groups have this.
1436                  */
1437                 for (i = 0; i < sbi->s_groups_count; i++)
1438                         overhead += ext2_bg_has_super(sb, i) +
1439                                 ext2_bg_num_gdb(sb, i);
1440
1441                 /*
1442                  * Every block group has an inode bitmap, a block
1443                  * bitmap, and an inode table.
1444                  */
1445                 overhead += (sbi->s_groups_count *
1446                              (2 + sbi->s_itb_per_group));
1447                 sbi->s_overhead_last = overhead;
1448                 smp_wmb();
1449                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1450         }
1451
1452         buf->f_type = EXT2_SUPER_MAGIC;
1453         buf->f_bsize = sb->s_blocksize;
1454         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1455         buf->f_bfree = ext2_count_free_blocks(sb);
1456         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1457         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1458         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1459                 buf->f_bavail = 0;
1460         buf->f_files = le32_to_cpu(es->s_inodes_count);
1461         buf->f_ffree = ext2_count_free_inodes(sb);
1462         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1463         buf->f_namelen = EXT2_NAME_LEN;
1464         fsid = le64_to_cpup((void *)es->s_uuid) ^
1465                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1466         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1467         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1468         spin_unlock(&sbi->s_lock);
1469         return 0;
1470 }
1471
1472 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1473         int flags, const char *dev_name, void *data)
1474 {
1475         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1476 }
1477
1478 #ifdef CONFIG_QUOTA
1479
1480 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1481  * acquiring the locks... As quota files are never truncated and quota code
1482  * itself serializes the operations (and no one else should touch the files)
1483  * we don't have to be afraid of races */
1484 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1485                                size_t len, loff_t off)
1486 {
1487         struct inode *inode = sb_dqopt(sb)->files[type];
1488         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1489         int err = 0;
1490         int offset = off & (sb->s_blocksize - 1);
1491         int tocopy;
1492         size_t toread;
1493         struct buffer_head tmp_bh;
1494         struct buffer_head *bh;
1495         loff_t i_size = i_size_read(inode);
1496
1497         if (off > i_size)
1498                 return 0;
1499         if (off+len > i_size)
1500                 len = i_size-off;
1501         toread = len;
1502         while (toread > 0) {
1503                 tocopy = sb->s_blocksize - offset < toread ?
1504                                 sb->s_blocksize - offset : toread;
1505
1506                 tmp_bh.b_state = 0;
1507                 tmp_bh.b_size = sb->s_blocksize;
1508                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1509                 if (err < 0)
1510                         return err;
1511                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1512                         memset(data, 0, tocopy);
1513                 else {
1514                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1515                         if (!bh)
1516                                 return -EIO;
1517                         memcpy(data, bh->b_data+offset, tocopy);
1518                         brelse(bh);
1519                 }
1520                 offset = 0;
1521                 toread -= tocopy;
1522                 data += tocopy;
1523                 blk++;
1524         }
1525         return len;
1526 }
1527
1528 /* Write to quotafile */
1529 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1530                                 const char *data, size_t len, loff_t off)
1531 {
1532         struct inode *inode = sb_dqopt(sb)->files[type];
1533         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1534         int err = 0;
1535         int offset = off & (sb->s_blocksize - 1);
1536         int tocopy;
1537         size_t towrite = len;
1538         struct buffer_head tmp_bh;
1539         struct buffer_head *bh;
1540
1541         while (towrite > 0) {
1542                 tocopy = sb->s_blocksize - offset < towrite ?
1543                                 sb->s_blocksize - offset : towrite;
1544
1545                 tmp_bh.b_state = 0;
1546                 tmp_bh.b_size = sb->s_blocksize;
1547                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1548                 if (err < 0)
1549                         goto out;
1550                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1551                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1552                 else
1553                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1554                 if (unlikely(!bh)) {
1555                         err = -EIO;
1556                         goto out;
1557                 }
1558                 lock_buffer(bh);
1559                 memcpy(bh->b_data+offset, data, tocopy);
1560                 flush_dcache_page(bh->b_page);
1561                 set_buffer_uptodate(bh);
1562                 mark_buffer_dirty(bh);
1563                 unlock_buffer(bh);
1564                 brelse(bh);
1565                 offset = 0;
1566                 towrite -= tocopy;
1567                 data += tocopy;
1568                 blk++;
1569         }
1570 out:
1571         if (len == towrite)
1572                 return err;
1573         if (inode->i_size < off+len-towrite)
1574                 i_size_write(inode, off+len-towrite);
1575         inode_inc_iversion(inode);
1576         inode->i_mtime = inode->i_ctime = current_time(inode);
1577         mark_inode_dirty(inode);
1578         return len - towrite;
1579 }
1580
1581 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1582                          const struct path *path)
1583 {
1584         int err;
1585         struct inode *inode;
1586
1587         err = dquot_quota_on(sb, type, format_id, path);
1588         if (err)
1589                 return err;
1590
1591         inode = d_inode(path->dentry);
1592         inode_lock(inode);
1593         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1594         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1595                         S_NOATIME | S_IMMUTABLE);
1596         inode_unlock(inode);
1597         mark_inode_dirty(inode);
1598
1599         return 0;
1600 }
1601
1602 static int ext2_quota_off(struct super_block *sb, int type)
1603 {
1604         struct inode *inode = sb_dqopt(sb)->files[type];
1605         int err;
1606
1607         if (!inode || !igrab(inode))
1608                 goto out;
1609
1610         err = dquot_quota_off(sb, type);
1611         if (err)
1612                 goto out_put;
1613
1614         inode_lock(inode);
1615         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1616         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1617         inode_unlock(inode);
1618         mark_inode_dirty(inode);
1619 out_put:
1620         iput(inode);
1621         return err;
1622 out:
1623         return dquot_quota_off(sb, type);
1624 }
1625
1626 #endif
1627
1628 static struct file_system_type ext2_fs_type = {
1629         .owner          = THIS_MODULE,
1630         .name           = "ext2",
1631         .mount          = ext2_mount,
1632         .kill_sb        = kill_block_super,
1633         .fs_flags       = FS_REQUIRES_DEV,
1634 };
1635 MODULE_ALIAS_FS("ext2");
1636
1637 static int __init init_ext2_fs(void)
1638 {
1639         int err;
1640
1641         err = init_inodecache();
1642         if (err)
1643                 return err;
1644         err = register_filesystem(&ext2_fs_type);
1645         if (err)
1646                 goto out;
1647         return 0;
1648 out:
1649         destroy_inodecache();
1650         return err;
1651 }
1652
1653 static void __exit exit_ext2_fs(void)
1654 {
1655         unregister_filesystem(&ext2_fs_type);
1656         destroy_inodecache();
1657 }
1658
1659 MODULE_AUTHOR("Remy Card and others");
1660 MODULE_DESCRIPTION("Second Extended Filesystem");
1661 MODULE_LICENSE("GPL");
1662 module_init(init_ext2_fs)
1663 module_exit(exit_ext2_fs)