fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/vfs.h>
  30 #include <linux/seq_file.h>
  31 #include <linux/mount.h>
  32 #include <linux/log2.h>
  33 #include <linux/quotaops.h>
  34 #include <linux/uaccess.h>
  35 #include <linux/dax.h>
  36 #include <linux/iversion.h>
  37 #include "ext2.h"
  38 #include "xattr.h"
  39 #include "acl.h"
  40
  41 static void ext2_write_super(struct super_block *sb);
  42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  44 static int ext2_sync_fs(struct super_block *sb, int wait);
  45 static int ext2_freeze(struct super_block *sb);
  46 static int ext2_unfreeze(struct super_block *sb);
  47
  48 void ext2_error(struct super_block *sb, const char *function,
  49                 const char *fmt, ...)
  50 {
  51         struct va_format vaf;
  52         va_list args;
  53         struct ext2_sb_info *sbi = EXT2_SB(sb);
  54         struct ext2_super_block *es = sbi->s_es;
  55
  56         if (!sb_rdonly(sb)) {
  57                 spin_lock(&sbi->s_lock);
  58                 sbi->s_mount_state |= EXT2_ERROR_FS;
  59                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  60                 spin_unlock(&sbi->s_lock);
  61                 ext2_sync_super(sb, es, 1);
  62         }
  63
  64         va_start(args, fmt);
  65
  66         vaf.fmt = fmt;
  67         vaf.va = &args;
  68
  69         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  70                sb->s_id, function, &vaf);
  71
  72         va_end(args);
  73
  74         if (test_opt(sb, ERRORS_PANIC))
  75                 panic("EXT2-fs: panic from previous error\n");
  76         if (test_opt(sb, ERRORS_RO)) {
  77                 ext2_msg(sb, KERN_CRIT,
  78                              "error: remounting filesystem read-only");
  79                 sb->s_flags |= SB_RDONLY;
  80         }
  81 }
  82
  83 void ext2_msg(struct super_block *sb, const char *prefix,
  84                 const char *fmt, ...)
  85 {
  86         struct va_format vaf;
  87         va_list args;
  88
  89         va_start(args, fmt);
  90
  91         vaf.fmt = fmt;
  92         vaf.va = &args;
  93
  94         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  95
  96         va_end(args);
  97 }
  98
  99 /*
 100  * This must be called with sbi->s_lock held.
 101  */
 102 void ext2_update_dynamic_rev(struct super_block *sb)
 103 {
 104         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 105
 106         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 107                 return;
 108
 109         ext2_msg(sb, KERN_WARNING,
 110                      "warning: updating to rev %d because of "
 111                      "new feature flag, running e2fsck is recommended",
 112                      EXT2_DYNAMIC_REV);
 113
 114         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 115         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 116         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 117         /* leave es->s_feature_*compat flags alone */
 118         /* es->s_uuid will be set by e2fsck if empty */
 119
 120         /*
 121          * The rest of the superblock fields should be zero, and if not it
 122          * means they are likely already in use, so leave them alone.  We
 123          * can leave it up to e2fsck to clean up any inconsistencies there.
 124          */
 125 }
 126
 127 #ifdef CONFIG_QUOTA
 128 static int ext2_quota_off(struct super_block *sb, int type);
 129
 130 static void ext2_quota_off_umount(struct super_block *sb)
 131 {
 132         int type;
 133
 134         for (type = 0; type < MAXQUOTAS; type++)
 135                 ext2_quota_off(sb, type);
 136 }
 137 #else
 138 static inline void ext2_quota_off_umount(struct super_block *sb)
 139 {
 140 }
 141 #endif
 142
 143 static void ext2_put_super (struct super_block * sb)
 144 {
 145         int db_count;
 146         int i;
 147         struct ext2_sb_info *sbi = EXT2_SB(sb);
 148
 149         ext2_quota_off_umount(sb);
 150
 151         if (sbi->s_ea_block_cache) {
 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                 if (sbi->s_group_desc[i])
 166                         brelse (sbi->s_group_desc[i]);
 167         kfree(sbi->s_group_desc);
 168         kfree(sbi->s_debts);
 169         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 170         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 171         percpu_counter_destroy(&sbi->s_dirs_counter);
 172         brelse (sbi->s_sbh);
 173         sb->s_fs_info = NULL;
 174         kfree(sbi->s_blockgroup_lock);
 175         fs_put_dax(sbi->s_daxdev);
 176         kfree(sbi);
 177 }
 178
 179 static struct kmem_cache * ext2_inode_cachep;
 180
 181 static struct inode *ext2_alloc_inode(struct super_block *sb)
 182 {
 183         struct ext2_inode_info *ei;
 184         ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 185         if (!ei)
 186                 return NULL;
 187         ei->i_block_alloc_info = NULL;
 188         inode_set_iversion(&ei->vfs_inode, 1);
 189 #ifdef CONFIG_QUOTA
 190         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 191 #endif
 192
 193         return &ei->vfs_inode;
 194 }
 195
 196 static void ext2_i_callback(struct rcu_head *head)
 197 {
 198         struct inode *inode = container_of(head, struct inode, i_rcu);
 199         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 200 }
 201
 202 static void ext2_destroy_inode(struct inode *inode)
 203 {
 204         call_rcu(&inode->i_rcu, ext2_i_callback);
 205 }
 206
 207 static void init_once(void *foo)
 208 {
 209         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 210
 211         rwlock_init(&ei->i_meta_lock);
 212 #ifdef CONFIG_EXT2_FS_XATTR
 213         init_rwsem(&ei->xattr_sem);
 214 #endif
 215         mutex_init(&ei->truncate_mutex);
 216 #ifdef CONFIG_FS_DAX
 217         init_rwsem(&ei->dax_sem);
 218 #endif
 219         inode_init_once(&ei->vfs_inode);
 220 }
 221
 222 static int __init init_inodecache(void)
 223 {
 224         ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
 225                                 sizeof(struct ext2_inode_info), 0,
 226                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
 227                                         SLAB_ACCOUNT),
 228                                 offsetof(struct ext2_inode_info, i_data),
 229                                 sizeof_field(struct ext2_inode_info, i_data),
 230                                 init_once);
 231         if (ext2_inode_cachep == NULL)
 232                 return -ENOMEM;
 233         return 0;
 234 }
 235
 236 static void destroy_inodecache(void)
 237 {
 238         /*
 239          * Make sure all delayed rcu free inodes are flushed before we
 240          * destroy cache.
 241          */
 242         rcu_barrier();
 243         kmem_cache_destroy(ext2_inode_cachep);
 244 }
 245
 246 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 247 {
 248         struct super_block *sb = root->d_sb;
 249         struct ext2_sb_info *sbi = EXT2_SB(sb);
 250         struct ext2_super_block *es = sbi->s_es;
 251         unsigned long def_mount_opts;
 252
 253         spin_lock(&sbi->s_lock);
 254         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 255
 256         if (sbi->s_sb_block != 1)
 257                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 258         if (test_opt(sb, MINIX_DF))
 259                 seq_puts(seq, ",minixdf");
 260         if (test_opt(sb, GRPID))
 261                 seq_puts(seq, ",grpid");
 262         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 263                 seq_puts(seq, ",nogrpid");
 264         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 265             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 266                 seq_printf(seq, ",resuid=%u",
 267                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 268         }
 269         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 270             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 271                 seq_printf(seq, ",resgid=%u",
 272                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 273         }
 274         if (test_opt(sb, ERRORS_RO)) {
 275                 int def_errors = le16_to_cpu(es->s_errors);
 276
 277                 if (def_errors == EXT2_ERRORS_PANIC ||
 278                     def_errors == EXT2_ERRORS_CONTINUE) {
 279                         seq_puts(seq, ",errors=remount-ro");
 280                 }
 281         }
 282         if (test_opt(sb, ERRORS_CONT))
 283                 seq_puts(seq, ",errors=continue");
 284         if (test_opt(sb, ERRORS_PANIC))
 285                 seq_puts(seq, ",errors=panic");
 286         if (test_opt(sb, NO_UID32))
 287                 seq_puts(seq, ",nouid32");
 288         if (test_opt(sb, DEBUG))
 289                 seq_puts(seq, ",debug");
 290         if (test_opt(sb, OLDALLOC))
 291                 seq_puts(seq, ",oldalloc");
 292
 293 #ifdef CONFIG_EXT2_FS_XATTR
 294         if (test_opt(sb, XATTR_USER))
 295                 seq_puts(seq, ",user_xattr");
 296         if (!test_opt(sb, XATTR_USER) &&
 297             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 298                 seq_puts(seq, ",nouser_xattr");
 299         }
 300 #endif
 301
 302 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 303         if (test_opt(sb, POSIX_ACL))
 304                 seq_puts(seq, ",acl");
 305         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 306                 seq_puts(seq, ",noacl");
 307 #endif
 308
 309         if (test_opt(sb, NOBH))
 310                 seq_puts(seq, ",nobh");
 311
 312 #if defined(CONFIG_QUOTA)
 313         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 314                 seq_puts(seq, ",usrquota");
 315
 316         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 317                 seq_puts(seq, ",grpquota");
 318 #endif
 319
 320 #ifdef CONFIG_FS_DAX
 321         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 322                 seq_puts(seq, ",xip");
 323         if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
 324                 seq_puts(seq, ",dax");
 325 #endif
 326
 327         if (!test_opt(sb, RESERVATION))
 328                 seq_puts(seq, ",noreservation");
 329
 330         spin_unlock(&sbi->s_lock);
 331         return 0;
 332 }
 333
 334 #ifdef CONFIG_QUOTA
 335 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 336 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 337 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
 338                          const struct path *path);
 339 static struct dquot **ext2_get_dquots(struct inode *inode)
 340 {
 341         return EXT2_I(inode)->i_dquot;
 342 }
 343
 344 static const struct quotactl_ops ext2_quotactl_ops = {
 345         .quota_on       = ext2_quota_on,
 346         .quota_off      = ext2_quota_off,
 347         .quota_sync     = dquot_quota_sync,
 348         .get_state      = dquot_get_state,
 349         .set_info       = dquot_set_dqinfo,
 350         .get_dqblk      = dquot_get_dqblk,
 351         .set_dqblk      = dquot_set_dqblk,
 352         .get_nextdqblk  = dquot_get_next_dqblk,
 353 };
 354 #endif
 355
 356 static const struct super_operations ext2_sops = {
 357         .alloc_inode    = ext2_alloc_inode,
 358         .destroy_inode  = ext2_destroy_inode,
 359         .write_inode    = ext2_write_inode,
 360         .evict_inode    = ext2_evict_inode,
 361         .put_super      = ext2_put_super,
 362         .sync_fs        = ext2_sync_fs,
 363         .freeze_fs      = ext2_freeze,
 364         .unfreeze_fs    = ext2_unfreeze,
 365         .statfs         = ext2_statfs,
 366         .remount_fs     = ext2_remount,
 367         .show_options   = ext2_show_options,
 368 #ifdef CONFIG_QUOTA
 369         .quota_read     = ext2_quota_read,
 370         .quota_write    = ext2_quota_write,
 371         .get_dquots     = ext2_get_dquots,
 372 #endif
 373 };
 374
 375 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 376                 u64 ino, u32 generation)
 377 {
 378         struct inode *inode;
 379
 380         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 381                 return ERR_PTR(-ESTALE);
 382         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 383                 return ERR_PTR(-ESTALE);
 384
 385         /*
 386          * ext2_iget isn't quite right if the inode is currently unallocated!
 387          * However ext2_iget currently does appropriate checks to handle stale
 388          * inodes so everything is OK.
 389          */
 390         inode = ext2_iget(sb, ino);
 391         if (IS_ERR(inode))
 392                 return ERR_CAST(inode);
 393         if (generation && inode->i_generation != generation) {
 394                 /* we didn't find the right inode.. */
 395                 iput(inode);
 396                 return ERR_PTR(-ESTALE);
 397         }
 398         return inode;
 399 }
 400
 401 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 402                 int fh_len, int fh_type)
 403 {
 404         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 405                                     ext2_nfs_get_inode);
 406 }
 407
 408 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 409                 int fh_len, int fh_type)
 410 {
 411         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 412                                     ext2_nfs_get_inode);
 413 }
 414
 415 static const struct export_operations ext2_export_ops = {
 416         .fh_to_dentry = ext2_fh_to_dentry,
 417         .fh_to_parent = ext2_fh_to_parent,
 418         .get_parent = ext2_get_parent,
 419 };
 420
 421 static unsigned long get_sb_block(void **data)
 422 {
 423         unsigned long   sb_block;
 424         char            *options = (char *) *data;
 425
 426         if (!options || strncmp(options, "sb=", 3) != 0)
 427                 return 1;       /* Default location */
 428         options += 3;
 429         sb_block = simple_strtoul(options, &options, 0);
 430         if (*options && *options != ',') {
 431                 printk("EXT2-fs: Invalid sb specification: %s\n",
 432                        (char *) *data);
 433                 return 1;
 434         }
 435         if (*options == ',')
 436                 options++;
 437         *data = (void *) options;
 438         return sb_block;
 439 }
 440
 441 enum {
 442         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 443         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 444         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 445         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 446         Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
 447         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 448 };
 449
 450 static const match_table_t tokens = {
 451         {Opt_bsd_df, "bsddf"},
 452         {Opt_minix_df, "minixdf"},
 453         {Opt_grpid, "grpid"},
 454         {Opt_grpid, "bsdgroups"},
 455         {Opt_nogrpid, "nogrpid"},
 456         {Opt_nogrpid, "sysvgroups"},
 457         {Opt_resgid, "resgid=%u"},
 458         {Opt_resuid, "resuid=%u"},
 459         {Opt_sb, "sb=%u"},
 460         {Opt_err_cont, "errors=continue"},
 461         {Opt_err_panic, "errors=panic"},
 462         {Opt_err_ro, "errors=remount-ro"},
 463         {Opt_nouid32, "nouid32"},
 464         {Opt_nocheck, "check=none"},
 465         {Opt_nocheck, "nocheck"},
 466         {Opt_debug, "debug"},
 467         {Opt_oldalloc, "oldalloc"},
 468         {Opt_orlov, "orlov"},
 469         {Opt_nobh, "nobh"},
 470         {Opt_user_xattr, "user_xattr"},
 471         {Opt_nouser_xattr, "nouser_xattr"},
 472         {Opt_acl, "acl"},
 473         {Opt_noacl, "noacl"},
 474         {Opt_xip, "xip"},
 475         {Opt_dax, "dax"},
 476         {Opt_grpquota, "grpquota"},
 477         {Opt_ignore, "noquota"},
 478         {Opt_quota, "quota"},
 479         {Opt_usrquota, "usrquota"},
 480         {Opt_reservation, "reservation"},
 481         {Opt_noreservation, "noreservation"},
 482         {Opt_err, NULL}
 483 };
 484
 485 static int parse_options(char *options, struct super_block *sb,
 486                          struct ext2_mount_options *opts)
 487 {
 488         char *p;
 489         substring_t args[MAX_OPT_ARGS];
 490         int option;
 491         kuid_t uid;
 492         kgid_t gid;
 493
 494         if (!options)
 495                 return 1;
 496
 497         while ((p = strsep (&options, ",")) != NULL) {
 498                 int token;
 499                 if (!*p)
 500                         continue;
 501
 502                 token = match_token(p, tokens, args);
 503                 switch (token) {
 504                 case Opt_bsd_df:
 505                         clear_opt (opts->s_mount_opt, MINIX_DF);
 506                         break;
 507                 case Opt_minix_df:
 508                         set_opt (opts->s_mount_opt, MINIX_DF);
 509                         break;
 510                 case Opt_grpid:
 511                         set_opt (opts->s_mount_opt, GRPID);
 512                         break;
 513                 case Opt_nogrpid:
 514                         clear_opt (opts->s_mount_opt, GRPID);
 515                         break;
 516                 case Opt_resuid:
 517                         if (match_int(&args[0], &option))
 518                                 return 0;
 519                         uid = make_kuid(current_user_ns(), option);
 520                         if (!uid_valid(uid)) {
 521                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 522                                 return 0;
 523
 524                         }
 525                         opts->s_resuid = uid;
 526                         break;
 527                 case Opt_resgid:
 528                         if (match_int(&args[0], &option))
 529                                 return 0;
 530                         gid = make_kgid(current_user_ns(), option);
 531                         if (!gid_valid(gid)) {
 532                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 533                                 return 0;
 534                         }
 535                         opts->s_resgid = gid;
 536                         break;
 537                 case Opt_sb:
 538                         /* handled by get_sb_block() instead of here */
 539                         /* *sb_block = match_int(&args[0]); */
 540                         break;
 541                 case Opt_err_panic:
 542                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 543                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 544                         set_opt (opts->s_mount_opt, ERRORS_PANIC);
 545                         break;
 546                 case Opt_err_ro:
 547                         clear_opt (opts->s_mount_opt, ERRORS_CONT);
 548                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 549                         set_opt (opts->s_mount_opt, ERRORS_RO);
 550                         break;
 551                 case Opt_err_cont:
 552                         clear_opt (opts->s_mount_opt, ERRORS_RO);
 553                         clear_opt (opts->s_mount_opt, ERRORS_PANIC);
 554                         set_opt (opts->s_mount_opt, ERRORS_CONT);
 555                         break;
 556                 case Opt_nouid32:
 557                         set_opt (opts->s_mount_opt, NO_UID32);
 558                         break;
 559                 case Opt_nocheck:
 560                         clear_opt (opts->s_mount_opt, CHECK);
 561                         break;
 562                 case Opt_debug:
 563                         set_opt (opts->s_mount_opt, DEBUG);
 564                         break;
 565                 case Opt_oldalloc:
 566                         set_opt (opts->s_mount_opt, OLDALLOC);
 567                         break;
 568                 case Opt_orlov:
 569                         clear_opt (opts->s_mount_opt, OLDALLOC);
 570                         break;
 571                 case Opt_nobh:
 572                         set_opt (opts->s_mount_opt, NOBH);
 573                         break;
 574 #ifdef CONFIG_EXT2_FS_XATTR
 575                 case Opt_user_xattr:
 576                         set_opt (opts->s_mount_opt, XATTR_USER);
 577                         break;
 578                 case Opt_nouser_xattr:
 579                         clear_opt (opts->s_mount_opt, XATTR_USER);
 580                         break;
 581 #else
 582                 case Opt_user_xattr:
 583                 case Opt_nouser_xattr:
 584                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 585                                 "not supported");
 586                         break;
 587 #endif
 588 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 589                 case Opt_acl:
 590                         set_opt(opts->s_mount_opt, POSIX_ACL);
 591                         break;
 592                 case Opt_noacl:
 593                         clear_opt(opts->s_mount_opt, POSIX_ACL);
 594                         break;
 595 #else
 596                 case Opt_acl:
 597                 case Opt_noacl:
 598                         ext2_msg(sb, KERN_INFO,
 599                                 "(no)acl options not supported");
 600                         break;
 601 #endif
 602                 case Opt_xip:
 603                         ext2_msg(sb, KERN_INFO, "use dax instead of xip");
 604                         set_opt(opts->s_mount_opt, XIP);
 605                         /* Fall through */
 606                 case Opt_dax:
 607 #ifdef CONFIG_FS_DAX
 608                         ext2_msg(sb, KERN_WARNING,
 609                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 610                         set_opt(opts->s_mount_opt, DAX);
 611 #else
 612                         ext2_msg(sb, KERN_INFO, "dax option not supported");
 613 #endif
 614                         break;
 615
 616 #if defined(CONFIG_QUOTA)
 617                 case Opt_quota:
 618                 case Opt_usrquota:
 619                         set_opt(opts->s_mount_opt, USRQUOTA);
 620                         break;
 621
 622                 case Opt_grpquota:
 623                         set_opt(opts->s_mount_opt, GRPQUOTA);
 624                         break;
 625 #else
 626                 case Opt_quota:
 627                 case Opt_usrquota:
 628                 case Opt_grpquota:
 629                         ext2_msg(sb, KERN_INFO,
 630                                 "quota operations not supported");
 631                         break;
 632 #endif
 633
 634                 case Opt_reservation:
 635                         set_opt(opts->s_mount_opt, RESERVATION);
 636                         ext2_msg(sb, KERN_INFO, "reservations ON");
 637                         break;
 638                 case Opt_noreservation:
 639                         clear_opt(opts->s_mount_opt, RESERVATION);
 640                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 641                         break;
 642                 case Opt_ignore:
 643                         break;
 644                 default:
 645                         return 0;
 646                 }
 647         }
 648         return 1;
 649 }
 650
 651 static int ext2_setup_super (struct super_block * sb,
 652                               struct ext2_super_block * es,
 653                               int read_only)
 654 {
 655         int res = 0;
 656         struct ext2_sb_info *sbi = EXT2_SB(sb);
 657
 658         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 659                 ext2_msg(sb, KERN_ERR,
 660                         "error: revision level too high, "
 661                         "forcing read-only mode");
 662                 res = SB_RDONLY;
 663         }
 664         if (read_only)
 665                 return res;
 666         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 667                 ext2_msg(sb, KERN_WARNING,
 668                         "warning: mounting unchecked fs, "
 669                         "running e2fsck is recommended");
 670         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 671                 ext2_msg(sb, KERN_WARNING,
 672                         "warning: mounting fs with errors, "
 673                         "running e2fsck is recommended");
 674         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 675                  le16_to_cpu(es->s_mnt_count) >=
 676                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 677                 ext2_msg(sb, KERN_WARNING,
 678                         "warning: maximal mount count reached, "
 679                         "running e2fsck is recommended");
 680         else if (le32_to_cpu(es->s_checkinterval) &&
 681                 (le32_to_cpu(es->s_lastcheck) +
 682                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 683                 ext2_msg(sb, KERN_WARNING,
 684                         "warning: checktime reached, "
 685                         "running e2fsck is recommended");
 686         if (!le16_to_cpu(es->s_max_mnt_count))
 687                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 688         le16_add_cpu(&es->s_mnt_count, 1);
 689         if (test_opt (sb, DEBUG))
 690                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 691                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 692                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 693                         sbi->s_frag_size,
 694                         sbi->s_groups_count,
 695                         EXT2_BLOCKS_PER_GROUP(sb),
 696                         EXT2_INODES_PER_GROUP(sb),
 697                         sbi->s_mount_opt);
 698         return res;
 699 }
 700
 701 static int ext2_check_descriptors(struct super_block *sb)
 702 {
 703         int i;
 704         struct ext2_sb_info *sbi = EXT2_SB(sb);
 705
 706         ext2_debug ("Checking group descriptors");
 707
 708         for (i = 0; i < sbi->s_groups_count; i++) {
 709                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 710                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 711                 ext2_fsblk_t last_block;
 712
 713                 if (i == sbi->s_groups_count - 1)
 714                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 715                 else
 716                         last_block = first_block +
 717                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 718
 719                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 720                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 721                 {
 722                         ext2_error (sb, "ext2_check_descriptors",
 723                                     "Block bitmap for group %d"
 724                                     " not in group (block %lu)!",
 725                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 726                         return 0;
 727                 }
 728                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 729                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 730                 {
 731                         ext2_error (sb, "ext2_check_descriptors",
 732                                     "Inode bitmap for group %d"
 733                                     " not in group (block %lu)!",
 734                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 735                         return 0;
 736                 }
 737                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 738                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 739                     last_block)
 740                 {
 741                         ext2_error (sb, "ext2_check_descriptors",
 742                                     "Inode table for group %d"
 743                                     " not in group (block %lu)!",
 744                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 745                         return 0;
 746                 }
 747         }
 748         return 1;
 749 }
 750
 751 /*
 752  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 753  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 754  * We need to be 1 filesystem block less than the 2^32 sector limit.
 755  */
 756 static loff_t ext2_max_size(int bits)
 757 {
 758         loff_t res = EXT2_NDIR_BLOCKS;
 759         int meta_blocks;
 760         loff_t upper_limit;
 761
 762         /* This is calculated to be the largest file size for a
 763          * dense, file such that the total number of
 764          * sectors in the file, including data and all indirect blocks,
 765          * does not exceed 2^32 -1
 766          * __u32 i_blocks representing the total number of
 767          * 512 bytes blocks of the file
 768          */
 769         upper_limit = (1LL << 32) - 1;
 770
 771         /* total blocks in file system block size */
 772         upper_limit >>= (bits - 9);
 773
 774
 775         /* indirect blocks */
 776         meta_blocks = 1;
 777         /* double indirect blocks */
 778         meta_blocks += 1 + (1LL << (bits-2));
 779         /* tripple indirect blocks */
 780         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 781
 782         upper_limit -= meta_blocks;
 783         upper_limit <<= bits;
 784
 785         res += 1LL << (bits-2);
 786         res += 1LL << (2*(bits-2));
 787         res += 1LL << (3*(bits-2));
 788         res <<= bits;
 789         if (res > upper_limit)
 790                 res = upper_limit;
 791
 792         if (res > MAX_LFS_FILESIZE)
 793                 res = MAX_LFS_FILESIZE;
 794
 795         return res;
 796 }
 797
 798 static unsigned long descriptor_loc(struct super_block *sb,
 799                                     unsigned long logic_sb_block,
 800                                     int nr)
 801 {
 802         struct ext2_sb_info *sbi = EXT2_SB(sb);
 803         unsigned long bg, first_meta_bg;
 804         int has_super = 0;
 805
 806         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 807
 808         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 809             nr < first_meta_bg)
 810                 return (logic_sb_block + nr + 1);
 811         bg = sbi->s_desc_per_block * nr;
 812         if (ext2_bg_has_super(sb, bg))
 813                 has_super = 1;
 814
 815         return ext2_group_first_block_no(sb, bg) + has_super;
 816 }
 817
 818 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 819 {
 820         struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 821         struct buffer_head * bh;
 822         struct ext2_sb_info * sbi;
 823         struct ext2_super_block * es;
 824         struct inode *root;
 825         unsigned long block;
 826         unsigned long sb_block = get_sb_block(&data);
 827         unsigned long logic_sb_block;
 828         unsigned long offset = 0;
 829         unsigned long def_mount_opts;
 830         long ret = -EINVAL;
 831         int blocksize = BLOCK_SIZE;
 832         int db_count;
 833         int i, j;
 834         __le32 features;
 835         int err;
 836         struct ext2_mount_options opts;
 837
 838         err = -ENOMEM;
 839         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 840         if (!sbi)
 841                 goto failed;
 842
 843         sbi->s_blockgroup_lock =
 844                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 845         if (!sbi->s_blockgroup_lock) {
 846                 kfree(sbi);
 847                 goto failed;
 848         }
 849         sb->s_fs_info = sbi;
 850         sbi->s_sb_block = sb_block;
 851         sbi->s_daxdev = dax_dev;
 852
 853         spin_lock_init(&sbi->s_lock);
 854
 855         /*
 856          * See what the current blocksize for the device is, and
 857          * use that as the blocksize.  Otherwise (or if the blocksize
 858          * is smaller than the default) use the default.
 859          * This is important for devices that have a hardware
 860          * sectorsize that is larger than the default.
 861          */
 862         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 863         if (!blocksize) {
 864                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 865                 goto failed_sbi;
 866         }
 867
 868         /*
 869          * If the superblock doesn't start on a hardware sector boundary,
 870          * calculate the offset.
 871          */
 872         if (blocksize != BLOCK_SIZE) {
 873                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 874                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 875         } else {
 876                 logic_sb_block = sb_block;
 877         }
 878
 879         if (!(bh = sb_bread(sb, logic_sb_block))) {
 880                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 881                 goto failed_sbi;
 882         }
 883         /*
 884          * Note: s_es must be initialized as soon as possible because
 885          *       some ext2 macro-instructions depend on its value
 886          */
 887         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 888         sbi->s_es = es;
 889         sb->s_magic = le16_to_cpu(es->s_magic);
 890
 891         if (sb->s_magic != EXT2_SUPER_MAGIC)
 892                 goto cantfind_ext2;
 893
 894         /* Set defaults before we parse the mount options */
 895         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 896         if (def_mount_opts & EXT2_DEFM_DEBUG)
 897                 set_opt(opts.s_mount_opt, DEBUG);
 898         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 899                 set_opt(opts.s_mount_opt, GRPID);
 900         if (def_mount_opts & EXT2_DEFM_UID16)
 901                 set_opt(opts.s_mount_opt, NO_UID32);
 902 #ifdef CONFIG_EXT2_FS_XATTR
 903         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 904                 set_opt(opts.s_mount_opt, XATTR_USER);
 905 #endif
 906 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 907         if (def_mount_opts & EXT2_DEFM_ACL)
 908                 set_opt(opts.s_mount_opt, POSIX_ACL);
 909 #endif
 910
 911         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 912                 set_opt(opts.s_mount_opt, ERRORS_PANIC);
 913         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 914                 set_opt(opts.s_mount_opt, ERRORS_CONT);
 915         else
 916                 set_opt(opts.s_mount_opt, ERRORS_RO);
 917
 918         opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 919         opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 920
 921         set_opt(opts.s_mount_opt, RESERVATION);
 922
 923         if (!parse_options((char *) data, sb, &opts))
 924                 goto failed_mount;
 925
 926         sbi->s_mount_opt = opts.s_mount_opt;
 927         sbi->s_resuid = opts.s_resuid;
 928         sbi->s_resgid = opts.s_resgid;
 929
 930         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 931                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 932                  SB_POSIXACL : 0);
 933         sb->s_iflags |= SB_I_CGROUPWB;
 934
 935         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 936             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 937              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 938              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 939                 ext2_msg(sb, KERN_WARNING,
 940                         "warning: feature flags set on rev 0 fs, "
 941                         "running e2fsck is recommended");
 942         /*
 943          * Check feature flags regardless of the revision level, since we
 944          * previously didn't change the revision level when setting the flags,
 945          * so there is a chance incompat flags are set on a rev 0 filesystem.
 946          */
 947         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 948         if (features) {
 949                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 950                        "unsupported optional features (%x)",
 951                         le32_to_cpu(features));
 952                 goto failed_mount;
 953         }
 954         if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 955                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 956                        "unsupported optional features (%x)",
 957                        le32_to_cpu(features));
 958                 goto failed_mount;
 959         }
 960
 961         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 962
 963         if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
 964                 err = bdev_dax_supported(sb, blocksize);
 965                 if (err) {
 966                         ext2_msg(sb, KERN_ERR,
 967                                 "DAX unsupported by block device. Turning off DAX.");
 968                         sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
 969                 }
 970         }
 971
 972         /* If the blocksize doesn't match, re-read the thing.. */
 973         if (sb->s_blocksize != blocksize) {
 974                 brelse(bh);
 975
 976                 if (!sb_set_blocksize(sb, blocksize)) {
 977                         ext2_msg(sb, KERN_ERR,
 978                                 "error: bad blocksize %d", blocksize);
 979                         goto failed_sbi;
 980                 }
 981
 982                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 983                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 984                 bh = sb_bread(sb, logic_sb_block);
 985                 if(!bh) {
 986                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 987                                 "superblock on 2nd try");
 988                         goto failed_sbi;
 989                 }
 990                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 991                 sbi->s_es = es;
 992                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 993                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 994                         goto failed_mount;
 995                 }
 996         }
 997
 998         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 999         sb->s_max_links = EXT2_LINK_MAX;
1000
1001         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1002                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1003                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1004         } else {
1005                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1006                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1007                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1008                     !is_power_of_2(sbi->s_inode_size) ||
1009                     (sbi->s_inode_size > blocksize)) {
1010                         ext2_msg(sb, KERN_ERR,
1011                                 "error: unsupported inode size: %d",
1012                                 sbi->s_inode_size);
1013                         goto failed_mount;
1014                 }
1015         }
1016
1017         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1018                                    le32_to_cpu(es->s_log_frag_size);
1019         if (sbi->s_frag_size == 0)
1020                 goto cantfind_ext2;
1021         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1022
1023         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1024         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1025         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1026
1027         if (EXT2_INODE_SIZE(sb) == 0)
1028                 goto cantfind_ext2;
1029         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1030         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1031                 goto cantfind_ext2;
1032         sbi->s_itb_per_group = sbi->s_inodes_per_group /
1033                                         sbi->s_inodes_per_block;
1034         sbi->s_desc_per_block = sb->s_blocksize /
1035                                         sizeof (struct ext2_group_desc);
1036         sbi->s_sbh = bh;
1037         sbi->s_mount_state = le16_to_cpu(es->s_state);
1038         sbi->s_addr_per_block_bits =
1039                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1040         sbi->s_desc_per_block_bits =
1041                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1042
1043         if (sb->s_magic != EXT2_SUPER_MAGIC)
1044                 goto cantfind_ext2;
1045
1046         if (sb->s_blocksize != bh->b_size) {
1047                 if (!silent)
1048                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1049                 goto failed_mount;
1050         }
1051
1052         if (sb->s_blocksize != sbi->s_frag_size) {
1053                 ext2_msg(sb, KERN_ERR,
1054                         "error: fragsize %lu != blocksize %lu"
1055                         "(not supported yet)",
1056                         sbi->s_frag_size, sb->s_blocksize);
1057                 goto failed_mount;
1058         }
1059
1060         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1061                 ext2_msg(sb, KERN_ERR,
1062                         "error: #blocks per group too big: %lu",
1063                         sbi->s_blocks_per_group);
1064                 goto failed_mount;
1065         }
1066         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1067                 ext2_msg(sb, KERN_ERR,
1068                         "error: #fragments per group too big: %lu",
1069                         sbi->s_frags_per_group);
1070                 goto failed_mount;
1071         }
1072         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1073                 ext2_msg(sb, KERN_ERR,
1074                         "error: #inodes per group too big: %lu",
1075                         sbi->s_inodes_per_group);
1076                 goto failed_mount;
1077         }
1078
1079         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1080                 goto cantfind_ext2;
1081         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1082                                 le32_to_cpu(es->s_first_data_block) - 1)
1083                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1084         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1085                    EXT2_DESC_PER_BLOCK(sb);
1086         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1087         if (sbi->s_group_desc == NULL) {
1088                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1089                 goto failed_mount;
1090         }
1091         bgl_lock_init(sbi->s_blockgroup_lock);
1092         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1093         if (!sbi->s_debts) {
1094                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1095                 goto failed_mount_group_desc;
1096         }
1097         for (i = 0; i < db_count; i++) {
1098                 block = descriptor_loc(sb, logic_sb_block, i);
1099                 sbi->s_group_desc[i] = sb_bread(sb, block);
1100                 if (!sbi->s_group_desc[i]) {
1101                         for (j = 0; j < i; j++)
1102                                 brelse (sbi->s_group_desc[j]);
1103                         ext2_msg(sb, KERN_ERR,
1104                                 "error: unable to read group descriptors");
1105                         goto failed_mount_group_desc;
1106                 }
1107         }
1108         if (!ext2_check_descriptors (sb)) {
1109                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1110                 goto failed_mount2;
1111         }
1112         sbi->s_gdb_count = db_count;
1113         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1114         spin_lock_init(&sbi->s_next_gen_lock);
1115
1116         /* per fileystem reservation list head & lock */
1117         spin_lock_init(&sbi->s_rsv_window_lock);
1118         sbi->s_rsv_window_root = RB_ROOT;
1119         /*
1120          * Add a single, static dummy reservation to the start of the
1121          * reservation window list --- it gives us a placeholder for
1122          * append-at-start-of-list which makes the allocation logic
1123          * _much_ simpler.
1124          */
1125         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1126         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1127         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1128         sbi->s_rsv_window_head.rsv_goal_size = 0;
1129         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1130
1131         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1132                                 ext2_count_free_blocks(sb), GFP_KERNEL);
1133         if (!err) {
1134                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1135                                 ext2_count_free_inodes(sb), GFP_KERNEL);
1136         }
1137         if (!err) {
1138                 err = percpu_counter_init(&sbi->s_dirs_counter,
1139                                 ext2_count_dirs(sb), GFP_KERNEL);
1140         }
1141         if (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                 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1150                 goto failed_mount3;
1151         }
1152 #endif
1153         /*
1154          * set up enough so that it can read an inode
1155          */
1156         sb->s_op = &ext2_sops;
1157         sb->s_export_op = &ext2_export_ops;
1158         sb->s_xattr = ext2_xattr_handlers;
1159
1160 #ifdef CONFIG_QUOTA
1161         sb->dq_op = &dquot_operations;
1162         sb->s_qcop = &ext2_quotactl_ops;
1163         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1164 #endif
1165
1166         root = ext2_iget(sb, EXT2_ROOT_INO);
1167         if (IS_ERR(root)) {
1168                 ret = PTR_ERR(root);
1169                 goto failed_mount3;
1170         }
1171         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1172                 iput(root);
1173                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1174                 goto failed_mount3;
1175         }
1176
1177         sb->s_root = d_make_root(root);
1178         if (!sb->s_root) {
1179                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1180                 ret = -ENOMEM;
1181                 goto failed_mount3;
1182         }
1183         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1184                 ext2_msg(sb, KERN_WARNING,
1185                         "warning: mounting ext3 filesystem as ext2");
1186         if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1187                 sb->s_flags |= SB_RDONLY;
1188         ext2_write_super(sb);
1189         return 0;
1190
1191 cantfind_ext2:
1192         if (!silent)
1193                 ext2_msg(sb, KERN_ERR,
1194                         "error: can't find an ext2 filesystem on dev %s.",
1195                         sb->s_id);
1196         goto failed_mount;
1197 failed_mount3:
1198         if (sbi->s_ea_block_cache)
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\n");
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(get_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         /*
1338          * Allow the "check" option to be passed as a remount option.
1339          */
1340         if (!parse_options(data, sb, &new_opts))
1341                 return -EINVAL;
1342
1343         spin_lock(&sbi->s_lock);
1344         es = sbi->s_es;
1345         if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1346                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1347                          "dax flag with busy inodes while remounting");
1348                 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1349         }
1350         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1351                 goto out_set;
1352         if (*flags & SB_RDONLY) {
1353                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1354                     !(sbi->s_mount_state & EXT2_VALID_FS))
1355                         goto out_set;
1356
1357                 /*
1358                  * OK, we are remounting a valid rw partition rdonly, so set
1359                  * the rdonly flag and then mark the partition as valid again.
1360                  */
1361                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1362                 es->s_mtime = cpu_to_le32(get_seconds());
1363                 spin_unlock(&sbi->s_lock);
1364
1365                 err = dquot_suspend(sb, -1);
1366                 if (err < 0)
1367                         return err;
1368
1369                 ext2_sync_super(sb, es, 1);
1370         } else {
1371                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1372                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1373                 if (ret) {
1374                         spin_unlock(&sbi->s_lock);
1375                         ext2_msg(sb, KERN_WARNING,
1376                                 "warning: couldn't remount RDWR because of "
1377                                 "unsupported optional features (%x).",
1378                                 le32_to_cpu(ret));
1379                         return -EROFS;
1380                 }
1381                 /*
1382                  * Mounting a RDONLY partition read-write, so reread and
1383                  * store the current valid flag.  (It may have been changed
1384                  * by e2fsck since we originally mounted the partition.)
1385                  */
1386                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1387                 if (!ext2_setup_super (sb, es, 0))
1388                         sb->s_flags &= ~SB_RDONLY;
1389                 spin_unlock(&sbi->s_lock);
1390
1391                 ext2_write_super(sb);
1392
1393                 dquot_resume(sb, -1);
1394         }
1395
1396         spin_lock(&sbi->s_lock);
1397 out_set:
1398         sbi->s_mount_opt = new_opts.s_mount_opt;
1399         sbi->s_resuid = new_opts.s_resuid;
1400         sbi->s_resgid = new_opts.s_resgid;
1401         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1402                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1403         spin_unlock(&sbi->s_lock);
1404
1405         return 0;
1406 }
1407
1408 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1409 {
1410         struct super_block *sb = dentry->d_sb;
1411         struct ext2_sb_info *sbi = EXT2_SB(sb);
1412         struct ext2_super_block *es = sbi->s_es;
1413         u64 fsid;
1414
1415         spin_lock(&sbi->s_lock);
1416
1417         if (test_opt (sb, MINIX_DF))
1418                 sbi->s_overhead_last = 0;
1419         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1420                 unsigned long i, overhead = 0;
1421                 smp_rmb();
1422
1423                 /*
1424                  * Compute the overhead (FS structures). This is constant
1425                  * for a given filesystem unless the number of block groups
1426                  * changes so we cache the previous value until it does.
1427                  */
1428
1429                 /*
1430                  * All of the blocks before first_data_block are
1431                  * overhead
1432                  */
1433                 overhead = le32_to_cpu(es->s_first_data_block);
1434
1435                 /*
1436                  * Add the overhead attributed to the superblock and
1437                  * block group descriptors.  If the sparse superblocks
1438                  * feature is turned on, then not all groups have this.
1439                  */
1440                 for (i = 0; i < sbi->s_groups_count; i++)
1441                         overhead += ext2_bg_has_super(sb, i) +
1442                                 ext2_bg_num_gdb(sb, i);
1443
1444                 /*
1445                  * Every block group has an inode bitmap, a block
1446                  * bitmap, and an inode table.
1447                  */
1448                 overhead += (sbi->s_groups_count *
1449                              (2 + sbi->s_itb_per_group));
1450                 sbi->s_overhead_last = overhead;
1451                 smp_wmb();
1452                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1453         }
1454
1455         buf->f_type = EXT2_SUPER_MAGIC;
1456         buf->f_bsize = sb->s_blocksize;
1457         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1458         buf->f_bfree = ext2_count_free_blocks(sb);
1459         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1460         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1461         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1462                 buf->f_bavail = 0;
1463         buf->f_files = le32_to_cpu(es->s_inodes_count);
1464         buf->f_ffree = ext2_count_free_inodes(sb);
1465         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1466         buf->f_namelen = EXT2_NAME_LEN;
1467         fsid = le64_to_cpup((void *)es->s_uuid) ^
1468                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1469         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1470         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1471         spin_unlock(&sbi->s_lock);
1472         return 0;
1473 }
1474
1475 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1476         int flags, const char *dev_name, void *data)
1477 {
1478         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1479 }
1480
1481 #ifdef CONFIG_QUOTA
1482
1483 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1484  * acquiring the locks... As quota files are never truncated and quota code
1485  * itself serializes the operations (and no one else should touch the files)
1486  * we don't have to be afraid of races */
1487 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1488                                size_t len, loff_t off)
1489 {
1490         struct inode *inode = sb_dqopt(sb)->files[type];
1491         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1492         int err = 0;
1493         int offset = off & (sb->s_blocksize - 1);
1494         int tocopy;
1495         size_t toread;
1496         struct buffer_head tmp_bh;
1497         struct buffer_head *bh;
1498         loff_t i_size = i_size_read(inode);
1499
1500         if (off > i_size)
1501                 return 0;
1502         if (off+len > i_size)
1503                 len = i_size-off;
1504         toread = len;
1505         while (toread > 0) {
1506                 tocopy = sb->s_blocksize - offset < toread ?
1507                                 sb->s_blocksize - offset : toread;
1508
1509                 tmp_bh.b_state = 0;
1510                 tmp_bh.b_size = sb->s_blocksize;
1511                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1512                 if (err < 0)
1513                         return err;
1514                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1515                         memset(data, 0, tocopy);
1516                 else {
1517                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1518                         if (!bh)
1519                                 return -EIO;
1520                         memcpy(data, bh->b_data+offset, tocopy);
1521                         brelse(bh);
1522                 }
1523                 offset = 0;
1524                 toread -= tocopy;
1525                 data += tocopy;
1526                 blk++;
1527         }
1528         return len;
1529 }
1530
1531 /* Write to quotafile */
1532 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1533                                 const char *data, size_t len, loff_t off)
1534 {
1535         struct inode *inode = sb_dqopt(sb)->files[type];
1536         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1537         int err = 0;
1538         int offset = off & (sb->s_blocksize - 1);
1539         int tocopy;
1540         size_t towrite = len;
1541         struct buffer_head tmp_bh;
1542         struct buffer_head *bh;
1543
1544         while (towrite > 0) {
1545                 tocopy = sb->s_blocksize - offset < towrite ?
1546                                 sb->s_blocksize - offset : towrite;
1547
1548                 tmp_bh.b_state = 0;
1549                 tmp_bh.b_size = sb->s_blocksize;
1550                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1551                 if (err < 0)
1552                         goto out;
1553                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1554                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1555                 else
1556                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1557                 if (unlikely(!bh)) {
1558                         err = -EIO;
1559                         goto out;
1560                 }
1561                 lock_buffer(bh);
1562                 memcpy(bh->b_data+offset, data, tocopy);
1563                 flush_dcache_page(bh->b_page);
1564                 set_buffer_uptodate(bh);
1565                 mark_buffer_dirty(bh);
1566                 unlock_buffer(bh);
1567                 brelse(bh);
1568                 offset = 0;
1569                 towrite -= tocopy;
1570                 data += tocopy;
1571                 blk++;
1572         }
1573 out:
1574         if (len == towrite)
1575                 return err;
1576         if (inode->i_size < off+len-towrite)
1577                 i_size_write(inode, off+len-towrite);
1578         inode_inc_iversion(inode);
1579         inode->i_mtime = inode->i_ctime = current_time(inode);
1580         mark_inode_dirty(inode);
1581         return len - towrite;
1582 }
1583
1584 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1585                          const struct path *path)
1586 {
1587         int err;
1588         struct inode *inode;
1589
1590         err = dquot_quota_on(sb, type, format_id, path);
1591         if (err)
1592                 return err;
1593
1594         inode = d_inode(path->dentry);
1595         inode_lock(inode);
1596         EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1597         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1598                         S_NOATIME | S_IMMUTABLE);
1599         inode_unlock(inode);
1600         mark_inode_dirty(inode);
1601
1602         return 0;
1603 }
1604
1605 static int ext2_quota_off(struct super_block *sb, int type)
1606 {
1607         struct inode *inode = sb_dqopt(sb)->files[type];
1608         int err;
1609
1610         if (!inode || !igrab(inode))
1611                 goto out;
1612
1613         err = dquot_quota_off(sb, type);
1614         if (err)
1615                 goto out_put;
1616
1617         inode_lock(inode);
1618         EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1619         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1620         inode_unlock(inode);
1621         mark_inode_dirty(inode);
1622 out_put:
1623         iput(inode);
1624         return err;
1625 out:
1626         return dquot_quota_off(sb, type);
1627 }
1628
1629 #endif
1630
1631 static struct file_system_type ext2_fs_type = {
1632         .owner          = THIS_MODULE,
1633         .name           = "ext2",
1634         .mount          = ext2_mount,
1635         .kill_sb        = kill_block_super,
1636         .fs_flags       = FS_REQUIRES_DEV,
1637 };
1638 MODULE_ALIAS_FS("ext2");
1639
1640 static int __init init_ext2_fs(void)
1641 {
1642         int err;
1643
1644         err = init_inodecache();
1645         if (err)
1646                 return err;
1647         err = register_filesystem(&ext2_fs_type);
1648         if (err)
1649                 goto out;
1650         return 0;
1651 out:
1652         destroy_inodecache();
1653         return err;
1654 }
1655
1656 static void __exit exit_ext2_fs(void)
1657 {
1658         unregister_filesystem(&ext2_fs_type);
1659         destroy_inodecache();
1660 }
1661
1662 MODULE_AUTHOR("Remy Card and others");
1663 MODULE_DESCRIPTION("Second Extended Filesystem");
1664 MODULE_LICENSE("GPL");
1665 module_init(init_ext2_fs)
1666 module_exit(exit_ext2_fs)