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;
 706
 707                 if (i == sbi->s_groups_count - 1)
 708                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 709                 else
 710                         last_block = first_block +
 711                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 712
 713                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 714                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 715                 {
 716                         ext2_error (sb, "ext2_check_descriptors",
 717                                     "Block bitmap for group %d"
 718                                     " not in group (block %lu)!",
 719                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 720                         return 0;
 721                 }
 722                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 723                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 724                 {
 725                         ext2_error (sb, "ext2_check_descriptors",
 726                                     "Inode bitmap for group %d"
 727                                     " not in group (block %lu)!",
 728                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 729                         return 0;
 730                 }
 731                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 732                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 733                     last_block)
 734                 {
 735                         ext2_error (sb, "ext2_check_descriptors",
 736                                     "Inode table for group %d"
 737                                     " not in group (block %lu)!",
 738                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 739                         return 0;
 740                 }
 741         }
 742         return 1;
 743 }
 744
 745 /*
 746  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 747  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 748  * We need to be 1 filesystem block less than the 2^32 sector limit.
 749  */
 750 static loff_t ext2_max_size(int bits)
 751 {
 752         loff_t res = EXT2_NDIR_BLOCKS;
 753         int meta_blocks;
 754         unsigned int upper_limit;
 755         unsigned int ppb = 1 << (bits-2);
 756
 757         /* This is calculated to be the largest file size for a
 758          * dense, file such that the total number of
 759          * sectors in the file, including data and all indirect blocks,
 760          * does not exceed 2^32 -1
 761          * __u32 i_blocks representing the total number of
 762          * 512 bytes blocks of the file
 763          */
 764         upper_limit = (1LL << 32) - 1;
 765
 766         /* total blocks in file system block size */
 767         upper_limit >>= (bits - 9);
 768
 769         /* Compute how many blocks we can address by block tree */
 770         res += 1LL << (bits-2);
 771         res += 1LL << (2*(bits-2));
 772         res += 1LL << (3*(bits-2));
 773         /* Does block tree limit file size? */
 774         if (res < upper_limit)
 775                 goto check_lfs;
 776
 777         res = upper_limit;
 778         /* How many metadata blocks are needed for addressing upper_limit? */
 779         upper_limit -= EXT2_NDIR_BLOCKS;
 780         /* indirect blocks */
 781         meta_blocks = 1;
 782         upper_limit -= ppb;
 783         /* double indirect blocks */
 784         if (upper_limit < ppb * ppb) {
 785                 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
 786                 res -= meta_blocks;
 787                 goto check_lfs;
 788         }
 789         meta_blocks += 1 + ppb;
 790         upper_limit -= ppb * ppb;
 791         /* tripple indirect blocks for the rest */
 792         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
 793                 DIV_ROUND_UP(upper_limit, ppb*ppb);
 794         res -= meta_blocks;
 795 check_lfs:
 796         res <<= bits;
 797         if (res > MAX_LFS_FILESIZE)
 798                 res = MAX_LFS_FILESIZE;
 799
 800         return res;
 801 }
 802
 803 static unsigned long descriptor_loc(struct super_block *sb,
 804                                     unsigned long logic_sb_block,
 805                                     int nr)
 806 {
 807         struct ext2_sb_info *sbi = EXT2_SB(sb);
 808         unsigned long bg, first_meta_bg;
 809         int has_super = 0;
 810
 811         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 812
 813         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 814             nr < first_meta_bg)
 815                 return (logic_sb_block + nr + 1);
 816         bg = sbi->s_desc_per_block * nr;
 817         if (ext2_bg_has_super(sb, bg))
 818                 has_super = 1;
 819
 820         return ext2_group_first_block_no(sb, bg) + has_super;
 821 }
 822
 823 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 824 {
 825         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 826         struct buffer_head * bh;
 827         struct ext2_sb_info * sbi;
 828         struct ext2_super_block * es;
 829         struct inode *root;
 830         unsigned long block;
 831         unsigned long sb_block = get_sb_block(&data);
 832         unsigned long logic_sb_block;
 833         unsigned long offset = 0;
 834         unsigned long def_mount_opts;
 835         long ret = -ENOMEM;
 836         int blocksize = BLOCK_SIZE;
 837         int db_count;
 838         int i, j;
 839         __le32 features;
 840         int err;
 841         struct ext2_mount_options opts;
 842
 843         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 844         if (!sbi)
 845                 goto failed;
 846
 847         sbi->s_blockgroup_lock =
 848                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 849         if (!sbi->s_blockgroup_lock) {
 850                 kfree(sbi);
 851                 goto failed;
 852         }
 853         sb->s_fs_info = sbi;
 854         sbi->s_sb_block = sb_block;
 855         sbi->s_daxdev = dax_dev;
 856
 857         spin_lock_init(&sbi->s_lock);
 858         ret = -EINVAL;
 859
 860         /*
 861          * See what the current blocksize for the device is, and
 862          * use that as the blocksize.  Otherwise (or if the blocksize
 863          * is smaller than the default) use the default.
 864          * This is important for devices that have a hardware
 865          * sectorsize that is larger than the default.
 866          */
 867         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 868         if (!blocksize) {
 869                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 870                 goto failed_sbi;
 871         }
 872
 873         /*
 874          * If the superblock doesn't start on a hardware sector boundary,
 875          * calculate the offset.
 876          */
 877         if (blocksize != BLOCK_SIZE) {
 878                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 879                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 880         } else {
 881                 logic_sb_block = sb_block;
 882         }
 883
 884         if (!(bh = sb_bread(sb, logic_sb_block))) {
 885                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 886                 goto failed_sbi;
 887         }
 888         /*
 889          * Note: s_es must be initialized as soon as possible because
 890          *       some ext2 macro-instructions depend on its value
 891          */
 892         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 893         sbi->s_es = es;
 894         sb->s_magic = le16_to_cpu(es->s_magic);
 895
 896         if (sb->s_magic != EXT2_SUPER_MAGIC)
 897                 goto cantfind_ext2;
 898
 899         opts.s_mount_opt = 0;
 900         /* Set defaults before we parse the mount options */
 901         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 902         if (def_mount_opts & EXT2_DEFM_DEBUG)
 903                 set_opt(opts.s_mount_opt, DEBUG);
 904         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 905                 set_opt(opts.s_mount_opt, GRPID);
 906         if (def_mount_opts & EXT2_DEFM_UID16)
 907                 set_opt(opts.s_mount_opt, NO_UID32);
 908 #ifdef CONFIG_EXT2_FS_XATTR
 909         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 910                 set_opt(opts.s_mount_opt, XATTR_USER);
 911 #endif
 912 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 913         if (def_mount_opts & EXT2_DEFM_ACL)
 914                 set_opt(opts.s_mount_opt, POSIX_ACL);
 915 #endif
 916
 917         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 918                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 919         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 920                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 921         else
 922                 set_opt(opts.s_mount_opt, ERRORS_RO);
 923
 924         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 925         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 926
 927         set_opt(opts.s_mount_opt, RESERVATION);
 928
 929         if (!parse_options((char *) data, sb, &opts))
 930                 goto failed_mount;
 931
 932         sbi->s_mount_opt = opts.s_mount_opt;
 933         sbi->s_resuid = opts.s_resuid;
 934         sbi->s_resgid = opts.s_resgid;
 935
 936         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 937                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
 938         sb->s_iflags |= SB_I_CGROUPWB;
 939
 940         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 941             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 942              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 943              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 944                 ext2_msg(sb, KERN_WARNING,
 945                         "warning: feature flags set on rev 0 fs, "
 946                         "running e2fsck is recommended");
 947         /*
 948          * Check feature flags regardless of the revision level, since we
 949          * previously didn't change the revision level when setting the flags,
 950          * so there is a chance incompat flags are set on a rev 0 filesystem.
 951          */
 952         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 953         if (features) {
 954                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 955                        "unsupported optional features (%x)",
 956                         le32_to_cpu(features));
 957                 goto failed_mount;
 958         }
 959         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 960                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 961                        "unsupported optional features (%x)",
 962                        le32_to_cpu(features));
 963                 goto failed_mount;
 964         }
 965
 966         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 967
 968         if (test_opt(sb, DAX)) {
 969                 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
 970                         ext2_msg(sb, KERN_ERR,
 971                                 "DAX unsupported by block device. Turning off DAX.");
 972                         clear_opt(sbi->s_mount_opt, DAX);
 973                 }
 974         }
 975
 976         /* If the blocksize doesn't match, re-read the thing.. */
 977         if (sb->s_blocksize != blocksize) {
 978                 brelse(bh);
 979
 980                 if (!sb_set_blocksize(sb, blocksize)) {
 981                         ext2_msg(sb, KERN_ERR,
 982                                 "error: bad blocksize %d", blocksize);
 983                         goto failed_sbi;
 984                 }
 985
 986                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 987                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 988                 bh = sb_bread(sb, logic_sb_block);
 989                 if(!bh) {
 990                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 991                                 "superblock on 2nd try");
 992                         goto failed_sbi;
 993                 }
 994                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 995                 sbi->s_es = es;
 996                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 997                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 998                         goto failed_mount;
 999                 }
1000         }
1001
1002         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1003         sb->s_max_links = EXT2_LINK_MAX;
1004         sb->s_time_min = S32_MIN;
1005         sb->s_time_max = S32_MAX;
1006
1007         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1008                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1009                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1010         } else {
1011                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1012                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1013                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1014                     !is_power_of_2(sbi->s_inode_size) ||
1015                     (sbi->s_inode_size > blocksize)) {
1016                         ext2_msg(sb, KERN_ERR,
1017                                 "error: unsupported inode size: %d",
1018                                 sbi->s_inode_size);
1019                         goto failed_mount;
1020                 }
1021         }
1022
1023         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1024                                    le32_to_cpu(es->s_log_frag_size);
1025         if (sbi->s_frag_size == 0)
1026                 goto cantfind_ext2;
1027         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1028
1029         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1030         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1031         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1032
1033         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1034         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1035                 goto cantfind_ext2;
1036         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1037                                         sbi->s_inodes_per_block;
1038         sbi->s_desc_per_block = sb->s_blocksize /
1039                                         sizeof (struct ext2_group_desc);
1040         sbi->s_sbh = bh;
1041         sbi->s_mount_state = le16_to_cpu(es->s_state);
1042         sbi->s_addr_per_block_bits =
1043                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1044         sbi->s_desc_per_block_bits =
1045                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1046
1047         if (sb->s_magic != EXT2_SUPER_MAGIC)
1048                 goto cantfind_ext2;
1049
1050         if (sb->s_blocksize != bh->b_size) {
1051                 if (!silent)
1052                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1053                 goto failed_mount;
1054         }
1055
1056         if (sb->s_blocksize != sbi->s_frag_size) {
1057                 ext2_msg(sb, KERN_ERR,
1058                         "error: fragsize %lu != blocksize %lu"
1059                         "(not supported yet)",
1060                         sbi->s_frag_size, sb->s_blocksize);
1061                 goto failed_mount;
1062         }
1063
1064         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1065                 ext2_msg(sb, KERN_ERR,
1066                         "error: #blocks per group too big: %lu",
1067                         sbi->s_blocks_per_group);
1068                 goto failed_mount;
1069         }
1070         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1071                 ext2_msg(sb, KERN_ERR,
1072                         "error: #fragments per group too big: %lu",
1073                         sbi->s_frags_per_group);
1074                 goto failed_mount;
1075         }
1076         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1077                 ext2_msg(sb, KERN_ERR,
1078                         "error: #inodes per group too big: %lu",
1079                         sbi->s_inodes_per_group);
1080                 goto failed_mount;
1081         }
1082
1083         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1084                 goto cantfind_ext2;
1085         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1086                                 le32_to_cpu(es->s_first_data_block) - 1)
1087                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1088         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1089                    EXT2_DESC_PER_BLOCK(sb);
1090         sbi->s_group_desc = kmalloc_array (db_count,
1091                                            sizeof(struct buffer_head *),
1092                                            GFP_KERNEL);
1093         if (sbi->s_group_desc == NULL) {
1094                 ret = -ENOMEM;
1095                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1096                 goto failed_mount;
1097         }
1098         bgl_lock_init(sbi->s_blockgroup_lock);
1099         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1100         if (!sbi->s_debts) {
1101                 ret = -ENOMEM;
1102                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1103                 goto failed_mount_group_desc;
1104         }
1105         for (i = 0; i < db_count; i++) {
1106                 block = descriptor_loc(sb, logic_sb_block, i);
1107                 sbi->s_group_desc[i] = sb_bread(sb, block);
1108                 if (!sbi->s_group_desc[i]) {
1109                         for (j = 0; j < i; j++)
1110                                 brelse (sbi->s_group_desc[j]);
1111                         ext2_msg(sb, KERN_ERR,
1112                                 "error: unable to read group descriptors");
1113                         goto failed_mount_group_desc;
1114                 }
1115         }
1116         if (!ext2_check_descriptors (sb)) {
1117                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1118                 goto failed_mount2;
1119         }
1120         sbi->s_gdb_count = db_count;
1121         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1122         spin_lock_init(&sbi->s_next_gen_lock);
1123
1124         /* per fileystem reservation list head & lock */
1125         spin_lock_init(&sbi->s_rsv_window_lock);
1126         sbi->s_rsv_window_root = RB_ROOT;
1127         /*
1128          * Add a single, static dummy reservation to the start of the
1129          * reservation window list --- it gives us a placeholder for
1130          * append-at-start-of-list which makes the allocation logic
1131          * _much_ simpler.
1132          */
1133         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1134         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1135         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1136         sbi->s_rsv_window_head.rsv_goal_size = 0;
1137         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1138
1139         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1140                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1141         if (!err) {
1142                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1143                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1144         }
1145         if (!err) {
1146                 err = percpu_counter_init(&sbi->s_dirs_counter,
1147                                 ext2_count_dirs(sb), GFP_KERNEL);
1148         }
1149         if (err) {
1150                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1151                 goto failed_mount3;
1152         }
1153
1154 #ifdef CONFIG_EXT2_FS_XATTR
1155         sbi->s_ea_block_cache = ext2_xattr_create_cache();
1156         if (!sbi->s_ea_block_cache) {
1157                 ret = -ENOMEM;
1158                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1159                 goto failed_mount3;
1160         }
1161 #endif
1162         /*
1163          * set up enough so that it can read an inode
1164          */
1165         sb->s_op = &ext2_sops;
1166         sb->s_export_op = &ext2_export_ops;
1167         sb->s_xattr = ext2_xattr_handlers;
1168
1169 #ifdef CONFIG_QUOTA
1170         sb->dq_op = &dquot_operations;
1171         sb->s_qcop = &ext2_quotactl_ops;
1172         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1173 #endif
1174
1175         root = ext2_iget(sb, EXT2_ROOT_INO);
1176         if (IS_ERR(root)) {
1177                 ret = PTR_ERR(root);
1178                 goto failed_mount3;
1179         }
1180         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1181                 iput(root);
1182                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1183                 goto failed_mount3;
1184         }
1185
1186         sb->s_root = d_make_root(root);
1187         if (!sb->s_root) {
1188                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1189                 ret = -ENOMEM;
1190                 goto failed_mount3;
1191         }
1192         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1193                 ext2_msg(sb, KERN_WARNING,
1194                         "warning: mounting ext3 filesystem as ext2");
1195         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1196                 sb->s_flags |= SB_RDONLY;
1197         ext2_write_super(sb);
1198         return 0;
1199
1200 cantfind_ext2:
1201         if (!silent)
1202                 ext2_msg(sb, KERN_ERR,
1203                         "error: can't find an ext2 filesystem on dev %s.",
1204                         sb->s_id);
1205         goto failed_mount;
1206 failed_mount3:
1207         ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1208         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1209         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1210         percpu_counter_destroy(&sbi->s_dirs_counter);
1211 failed_mount2:
1212         for (i = 0; i < db_count; i++)
1213                 brelse(sbi->s_group_desc[i]);
1214 failed_mount_group_desc:
1215         kfree(sbi->s_group_desc);
1216         kfree(sbi->s_debts);
1217 failed_mount:
1218         brelse(bh);
1219 failed_sbi:
1220         sb->s_fs_info = NULL;
1221         kfree(sbi->s_blockgroup_lock);
1222         kfree(sbi);
1223 failed:
1224         fs_put_dax(dax_dev);
1225         return ret;
1226 }
1227
1228 static void ext2_clear_super_error(struct super_block *sb)
1229 {
1230         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1231
1232         if (buffer_write_io_error(sbh)) {
1233                 /*
1234                  * Oh, dear.  A previous attempt to write the
1235                  * superblock failed.  This could happen because the
1236                  * USB device was yanked out.  Or it could happen to
1237                  * be a transient write error and maybe the block will
1238                  * be remapped.  Nothing we can do but to retry the
1239                  * write and hope for the best.
1240                  */
1241                 ext2_msg(sb, KERN_ERR,
1242                        "previous I/O error to superblock detected");
1243                 clear_buffer_write_io_error(sbh);
1244                 set_buffer_uptodate(sbh);
1245         }
1246 }
1247
1248 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1249                      int wait)
1250 {
1251         ext2_clear_super_error(sb);
1252         spin_lock(&EXT2_SB(sb)->s_lock);
1253         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1254         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1255         es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1256         /* unlock before we do IO */
1257         spin_unlock(&EXT2_SB(sb)->s_lock);
1258         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1259         if (wait)
1260                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1261 }
1262
1263 /*
1264  * In the second extended file system, it is not necessary to
1265  * write the super block since we use a mapping of the
1266  * disk super block in a buffer.
1267  *
1268  * However, this function is still used to set the fs valid
1269  * flags to 0.  We need to set this flag to 0 since the fs
1270  * may have been checked while mounted and e2fsck may have
1271  * set s_state to EXT2_VALID_FS after some corrections.
1272  */
1273 static int ext2_sync_fs(struct super_block *sb, int wait)
1274 {
1275         struct ext2_sb_info *sbi = EXT2_SB(sb);
1276         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1277
1278         /*
1279          * Write quota structures to quota file, sync_blockdev() will write
1280          * them to disk later
1281          */
1282         dquot_writeback_dquots(sb, -1);
1283
1284         spin_lock(&sbi->s_lock);
1285         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1286                 ext2_debug("setting valid to 0\n");
1287                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1288         }
1289         spin_unlock(&sbi->s_lock);
1290         ext2_sync_super(sb, es, wait);
1291         return 0;
1292 }
1293
1294 static int ext2_freeze(struct super_block *sb)
1295 {
1296         struct ext2_sb_info *sbi = EXT2_SB(sb);
1297
1298         /*
1299          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1300          * because we have unattached inodes and thus filesystem is not fully
1301          * consistent.
1302          */
1303         if (atomic_long_read(&sb->s_remove_count)) {
1304                 ext2_sync_fs(sb, 1);
1305                 return 0;
1306         }
1307         /* Set EXT2_FS_VALID flag */
1308         spin_lock(&sbi->s_lock);
1309         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1310         spin_unlock(&sbi->s_lock);
1311         ext2_sync_super(sb, sbi->s_es, 1);
1312
1313         return 0;
1314 }
1315
1316 static int ext2_unfreeze(struct super_block *sb)
1317 {
1318         /* Just write sb to clear EXT2_VALID_FS flag */
1319         ext2_write_super(sb);
1320
1321         return 0;
1322 }
1323
1324 static void ext2_write_super(struct super_block *sb)
1325 {
1326         if (!sb_rdonly(sb))
1327                 ext2_sync_fs(sb, 1);
1328 }
1329
1330 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1331 {
1332         struct ext2_sb_info * sbi = EXT2_SB(sb);
1333         struct ext2_super_block * es;
1334         struct ext2_mount_options new_opts;
1335         int err;
1336
1337         sync_filesystem(sb);
1338
1339         spin_lock(&sbi->s_lock);
1340         new_opts.s_mount_opt = sbi->s_mount_opt;
1341         new_opts.s_resuid = sbi->s_resuid;
1342         new_opts.s_resgid = sbi->s_resgid;
1343         spin_unlock(&sbi->s_lock);
1344
1345         if (!parse_options(data, sb, &new_opts))
1346                 return -EINVAL;
1347
1348         spin_lock(&sbi->s_lock);
1349         es = sbi->s_es;
1350         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1351                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1352                          "dax flag with busy inodes while remounting");
1353                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1354         }
1355         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1356                 goto out_set;
1357         if (*flags & SB_RDONLY) {
1358                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1359                     !(sbi->s_mount_state & EXT2_VALID_FS))
1360                         goto out_set;
1361
1362                 /*
1363                  * OK, we are remounting a valid rw partition rdonly, so set
1364                  * the rdonly flag and then mark the partition as valid again.
1365                  */
1366                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1367                 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1368                 spin_unlock(&sbi->s_lock);
1369
1370                 err = dquot_suspend(sb, -1);
1371                 if (err < 0)
1372                         return err;
1373
1374                 ext2_sync_super(sb, es, 1);
1375         } else {
1376                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1377                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1378                 if (ret) {
1379                         spin_unlock(&sbi->s_lock);
1380                         ext2_msg(sb, KERN_WARNING,
1381                                 "warning: couldn't remount RDWR because of "
1382                                 "unsupported optional features (%x).",
1383                                 le32_to_cpu(ret));
1384                         return -EROFS;
1385                 }
1386                 /*
1387                  * Mounting a RDONLY partition read-write, so reread and
1388                  * store the current valid flag.  (It may have been changed
1389                  * by e2fsck since we originally mounted the partition.)
1390                  */
1391                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1392                 if (!ext2_setup_super (sb, es, 0))
1393                         sb->s_flags &= ~SB_RDONLY;
1394                 spin_unlock(&sbi->s_lock);
1395
1396                 ext2_write_super(sb);
1397
1398                 dquot_resume(sb, -1);
1399         }
1400
1401         spin_lock(&sbi->s_lock);
1402 out_set:
1403         sbi->s_mount_opt = new_opts.s_mount_opt;
1404         sbi->s_resuid = new_opts.s_resuid;
1405         sbi->s_resgid = new_opts.s_resgid;
1406         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1407                 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1408         spin_unlock(&sbi->s_lock);
1409
1410         return 0;
1411 }
1412
1413 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1414 {
1415         struct super_block *sb = dentry->d_sb;
1416         struct ext2_sb_info *sbi = EXT2_SB(sb);
1417         struct ext2_super_block *es = sbi->s_es;
1418         u64 fsid;
1419
1420         spin_lock(&sbi->s_lock);
1421
1422         if (test_opt (sb, MINIX_DF))
1423                 sbi->s_overhead_last = 0;
1424         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1425                 unsigned long i, overhead = 0;
1426                 smp_rmb();
1427
1428                 /*
1429                  * Compute the overhead (FS structures). This is constant
1430                  * for a given filesystem unless the number of block groups
1431                  * changes so we cache the previous value until it does.
1432                  */
1433
1434                 /*
1435                  * All of the blocks before first_data_block are
1436                  * overhead
1437                  */
1438                 overhead = le32_to_cpu(es->s_first_data_block);
1439
1440                 /*
1441                  * Add the overhead attributed to the superblock and
1442                  * block group descriptors.  If the sparse superblocks
1443                  * feature is turned on, then not all groups have this.
1444                  */
1445                 for (i = 0; i < sbi->s_groups_count; i++)
1446                         overhead += ext2_bg_has_super(sb, i) +
1447                                 ext2_bg_num_gdb(sb, i);
1448
1449                 /*
1450                  * Every block group has an inode bitmap, a block
1451                  * bitmap, and an inode table.
1452                  */
1453                 overhead += (sbi->s_groups_count *
1454                              (2 + sbi->s_itb_per_group));
1455                 sbi->s_overhead_last = overhead;
1456                 smp_wmb();
1457                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1458         }
1459
1460         buf->f_type = EXT2_SUPER_MAGIC;
1461         buf->f_bsize = sb->s_blocksize;
1462         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1463         buf->f_bfree = ext2_count_free_blocks(sb);
1464         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1465         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1466         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1467                 buf->f_bavail = 0;
1468         buf->f_files = le32_to_cpu(es->s_inodes_count);
1469         buf->f_ffree = ext2_count_free_inodes(sb);
1470         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1471         buf->f_namelen = EXT2_NAME_LEN;
1472         fsid = le64_to_cpup((void *)es->s_uuid) ^
1473                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1474         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1475         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1476         spin_unlock(&sbi->s_lock);
1477         return 0;
1478 }
1479
1480 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1481         int flags, const char *dev_name, void *data)
1482 {
1483         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1484 }
1485
1486 #ifdef CONFIG_QUOTA
1487
1488 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1489  * acquiring the locks... As quota files are never truncated and quota code
1490  * itself serializes the operations (and no one else should touch the files)
1491  * we don't have to be afraid of races */
1492 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1493                                size_t len, loff_t off)
1494 {
1495         struct inode *inode = sb_dqopt(sb)->files[type];
1496         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1497         int err = 0;
1498         int offset = off & (sb->s_blocksize - 1);
1499         int tocopy;
1500         size_t toread;
1501         struct buffer_head tmp_bh;
1502         struct buffer_head *bh;
1503         loff_t i_size = i_size_read(inode);
1504
1505         if (off > i_size)
1506                 return 0;
1507         if (off+len > i_size)
1508                 len = i_size-off;
1509         toread = len;
1510         while (toread > 0) {
1511                 tocopy = sb->s_blocksize - offset < toread ?
1512                                 sb->s_blocksize - offset : toread;
1513
1514                 tmp_bh.b_state = 0;
1515                 tmp_bh.b_size = sb->s_blocksize;
1516                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1517                 if (err < 0)
1518                         return err;
1519                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1520                         memset(data, 0, tocopy);
1521                 else {
1522                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1523                         if (!bh)
1524                                 return -EIO;
1525                         memcpy(data, bh->b_data+offset, tocopy);
1526                         brelse(bh);
1527                 }
1528                 offset = 0;
1529                 toread -= tocopy;
1530                 data += tocopy;
1531                 blk++;
1532         }
1533         return len;
1534 }
1535
1536 /* Write to quotafile */
1537 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1538                                 const char *data, size_t len, loff_t off)
1539 {
1540         struct inode *inode = sb_dqopt(sb)->files[type];
1541         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1542         int err = 0;
1543         int offset = off & (sb->s_blocksize - 1);
1544         int tocopy;
1545         size_t towrite = len;
1546         struct buffer_head tmp_bh;
1547         struct buffer_head *bh;
1548
1549         while (towrite > 0) {
1550                 tocopy = sb->s_blocksize - offset < towrite ?
1551                                 sb->s_blocksize - offset : towrite;
1552
1553                 tmp_bh.b_state = 0;
1554                 tmp_bh.b_size = sb->s_blocksize;
1555                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1556                 if (err < 0)
1557                         goto out;
1558                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1559                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1560                 else
1561                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1562                 if (unlikely(!bh)) {
1563                         err = -EIO;
1564                         goto out;
1565                 }
1566                 lock_buffer(bh);
1567                 memcpy(bh->b_data+offset, data, tocopy);
1568                 flush_dcache_page(bh->b_page);
1569                 set_buffer_uptodate(bh);
1570                 mark_buffer_dirty(bh);
1571                 unlock_buffer(bh);
1572                 brelse(bh);
1573                 offset = 0;
1574                 towrite -= tocopy;
1575                 data += tocopy;
1576                 blk++;
1577         }
1578 out:
1579         if (len == towrite)
1580                 return err;
1581         if (inode->i_size < off+len-towrite)
1582                 i_size_write(inode, off+len-towrite);
1583         inode_inc_iversion(inode);
1584         inode->i_mtime = inode->i_ctime = current_time(inode);
1585         mark_inode_dirty(inode);
1586         return len - towrite;
1587 }
1588
1589 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1590                          const struct path *path)
1591 {
1592         int err;
1593         struct inode *inode;
1594
1595         err = dquot_quota_on(sb, type, format_id, path);
1596         if (err)
1597                 return err;
1598
1599         inode = d_inode(path->dentry);
1600         inode_lock(inode);
1601         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1602         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1603                         S_NOATIME | S_IMMUTABLE);
1604         inode_unlock(inode);
1605         mark_inode_dirty(inode);
1606
1607         return 0;
1608 }
1609
1610 static int ext2_quota_off(struct super_block *sb, int type)
1611 {
1612         struct inode *inode = sb_dqopt(sb)->files[type];
1613         int err;
1614
1615         if (!inode || !igrab(inode))
1616                 goto out;
1617
1618         err = dquot_quota_off(sb, type);
1619         if (err)
1620                 goto out_put;
1621
1622         inode_lock(inode);
1623         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1624         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1625         inode_unlock(inode);
1626         mark_inode_dirty(inode);
1627 out_put:
1628         iput(inode);
1629         return err;
1630 out:
1631         return dquot_quota_off(sb, type);
1632 }
1633
1634 #endif
1635
1636 static struct file_system_type ext2_fs_type = {
1637         .owner          = THIS_MODULE,
1638         .name           = "ext2",
1639         .mount          = ext2_mount,
1640         .kill_sb        = kill_block_super,
1641         .fs_flags       = FS_REQUIRES_DEV,
1642 };
1643 MODULE_ALIAS_FS("ext2");
1644
1645 static int __init init_ext2_fs(void)
1646 {
1647         int err;
1648
1649         err = init_inodecache();
1650         if (err)
1651                 return err;
1652         err = register_filesystem(&ext2_fs_type);
1653         if (err)
1654                 goto out;
1655         return 0;
1656 out:
1657         destroy_inodecache();
1658         return err;
1659 }
1660
1661 static void __exit exit_ext2_fs(void)
1662 {
1663         unregister_filesystem(&ext2_fs_type);
1664         destroy_inodecache();
1665 }
1666
1667 MODULE_AUTHOR("Remy Card and others");
1668 MODULE_DESCRIPTION("Second Extended Filesystem");
1669 MODULE_LICENSE("GPL");
1670 module_init(init_ext2_fs)
1671 module_exit(exit_ext2_fs)