fs/reiserfs/super.c

   1 /*
   2  * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
   3  *
   4  * Trivial changes by Alan Cox to add the LFS fixes
   5  *
   6  * Trivial Changes:
   7  * Rights granted to Hans Reiser to redistribute under other terms providing
   8  * he accepts all liability including but not limited to patent, fitness
   9  * for purpose, and direct or indirect claims arising from failure to perform.
  10  *
  11  * NO WARRANTY
  12  */
  13
  14 #include <linux/module.h>
  15 #include <linux/slab.h>
  16 #include <linux/vmalloc.h>
  17 #include <linux/time.h>
  18 #include <linux/uaccess.h>
  19 #include "reiserfs.h"
  20 #include "acl.h"
  21 #include "xattr.h"
  22 #include <linux/init.h>
  23 #include <linux/blkdev.h>
  24 #include <linux/backing-dev.h>
  25 #include <linux/buffer_head.h>
  26 #include <linux/exportfs.h>
  27 #include <linux/quotaops.h>
  28 #include <linux/vfs.h>
  29 #include <linux/mount.h>
  30 #include <linux/namei.h>
  31 #include <linux/crc32.h>
  32 #include <linux/seq_file.h>
  33
  34 struct file_system_type reiserfs_fs_type;
  35
  36 static const char reiserfs_3_5_magic_string[] = REISERFS_SUPER_MAGIC_STRING;
  37 static const char reiserfs_3_6_magic_string[] = REISER2FS_SUPER_MAGIC_STRING;
  38 static const char reiserfs_jr_magic_string[] = REISER2FS_JR_SUPER_MAGIC_STRING;
  39
  40 int is_reiserfs_3_5(struct reiserfs_super_block *rs)
  41 {
  42         return !strncmp(rs->s_v1.s_magic, reiserfs_3_5_magic_string,
  43                         strlen(reiserfs_3_5_magic_string));
  44 }
  45
  46 int is_reiserfs_3_6(struct reiserfs_super_block *rs)
  47 {
  48         return !strncmp(rs->s_v1.s_magic, reiserfs_3_6_magic_string,
  49                         strlen(reiserfs_3_6_magic_string));
  50 }
  51
  52 int is_reiserfs_jr(struct reiserfs_super_block *rs)
  53 {
  54         return !strncmp(rs->s_v1.s_magic, reiserfs_jr_magic_string,
  55                         strlen(reiserfs_jr_magic_string));
  56 }
  57
  58 static int is_any_reiserfs_magic_string(struct reiserfs_super_block *rs)
  59 {
  60         return (is_reiserfs_3_5(rs) || is_reiserfs_3_6(rs) ||
  61                 is_reiserfs_jr(rs));
  62 }
  63
  64 static int reiserfs_remount(struct super_block *s, int *flags, char *data);
  65 static int reiserfs_statfs(struct dentry *dentry, struct kstatfs *buf);
  66
  67 static int reiserfs_sync_fs(struct super_block *s, int wait)
  68 {
  69         struct reiserfs_transaction_handle th;
  70
  71         /*
  72          * Writeback quota in non-journalled quota case - journalled quota has
  73          * no dirty dquots
  74          */
  75         dquot_writeback_dquots(s, -1);
  76         reiserfs_write_lock(s);
  77         if (!journal_begin(&th, s, 1))
  78                 if (!journal_end_sync(&th))
  79                         reiserfs_flush_old_commits(s);
  80         reiserfs_write_unlock(s);
  81         return 0;
  82 }
  83
  84 static void flush_old_commits(struct work_struct *work)
  85 {
  86         struct reiserfs_sb_info *sbi;
  87         struct super_block *s;
  88
  89         sbi = container_of(work, struct reiserfs_sb_info, old_work.work);
  90         s = sbi->s_journal->j_work_sb;
  91
  92         /*
  93          * We need s_umount for protecting quota writeback. We have to use
  94          * trylock as reiserfs_cancel_old_flush() may be waiting for this work
  95          * to complete with s_umount held.
  96          */
  97         if (!down_read_trylock(&s->s_umount)) {
  98                 /* Requeue work if we are not cancelling it */
  99                 spin_lock(&sbi->old_work_lock);
 100                 if (sbi->work_queued == 1)
 101                         queue_delayed_work(system_long_wq, &sbi->old_work, HZ);
 102                 spin_unlock(&sbi->old_work_lock);
 103                 return;
 104         }
 105         spin_lock(&sbi->old_work_lock);
 106         /* Avoid clobbering the cancel state... */
 107         if (sbi->work_queued == 1)
 108                 sbi->work_queued = 0;
 109         spin_unlock(&sbi->old_work_lock);
 110
 111         reiserfs_sync_fs(s, 1);
 112         up_read(&s->s_umount);
 113 }
 114
 115 void reiserfs_schedule_old_flush(struct super_block *s)
 116 {
 117         struct reiserfs_sb_info *sbi = REISERFS_SB(s);
 118         unsigned long delay;
 119
 120         /*
 121          * Avoid scheduling flush when sb is being shut down. It can race
 122          * with journal shutdown and free still queued delayed work.
 123          */
 124         if (s->s_flags & MS_RDONLY || !(s->s_flags & MS_ACTIVE))
 125                 return;
 126
 127         spin_lock(&sbi->old_work_lock);
 128         if (!sbi->work_queued) {
 129                 delay = msecs_to_jiffies(dirty_writeback_interval * 10);
 130                 queue_delayed_work(system_long_wq, &sbi->old_work, delay);
 131                 sbi->work_queued = 1;
 132         }
 133         spin_unlock(&sbi->old_work_lock);
 134 }
 135
 136 void reiserfs_cancel_old_flush(struct super_block *s)
 137 {
 138         struct reiserfs_sb_info *sbi = REISERFS_SB(s);
 139
 140         spin_lock(&sbi->old_work_lock);
 141         /* Make sure no new flushes will be queued */
 142         sbi->work_queued = 2;
 143         spin_unlock(&sbi->old_work_lock);
 144         cancel_delayed_work_sync(&REISERFS_SB(s)->old_work);
 145 }
 146
 147 static int reiserfs_freeze(struct super_block *s)
 148 {
 149         struct reiserfs_transaction_handle th;
 150
 151         reiserfs_cancel_old_flush(s);
 152
 153         reiserfs_write_lock(s);
 154         if (!(s->s_flags & MS_RDONLY)) {
 155                 int err = journal_begin(&th, s, 1);
 156                 if (err) {
 157                         reiserfs_block_writes(&th);
 158                 } else {
 159                         reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s),
 160                                                      1);
 161                         journal_mark_dirty(&th, SB_BUFFER_WITH_SB(s));
 162                         reiserfs_block_writes(&th);
 163                         journal_end_sync(&th);
 164                 }
 165         }
 166         reiserfs_write_unlock(s);
 167         return 0;
 168 }
 169
 170 static int reiserfs_unfreeze(struct super_block *s)
 171 {
 172         struct reiserfs_sb_info *sbi = REISERFS_SB(s);
 173
 174         reiserfs_allow_writes(s);
 175         spin_lock(&sbi->old_work_lock);
 176         /* Allow old_work to run again */
 177         sbi->work_queued = 0;
 178         spin_unlock(&sbi->old_work_lock);
 179         return 0;
 180 }
 181
 182 extern const struct in_core_key MAX_IN_CORE_KEY;
 183
 184 /*
 185  * this is used to delete "save link" when there are no items of a
 186  * file it points to. It can either happen if unlink is completed but
 187  * "save unlink" removal, or if file has both unlink and truncate
 188  * pending and as unlink completes first (because key of "save link"
 189  * protecting unlink is bigger that a key lf "save link" which
 190  * protects truncate), so there left no items to make truncate
 191  * completion on
 192  */
 193 static int remove_save_link_only(struct super_block *s,
 194                                  struct reiserfs_key *key, int oid_free)
 195 {
 196         struct reiserfs_transaction_handle th;
 197         int err;
 198
 199         /* we are going to do one balancing */
 200         err = journal_begin(&th, s, JOURNAL_PER_BALANCE_CNT);
 201         if (err)
 202                 return err;
 203
 204         reiserfs_delete_solid_item(&th, NULL, key);
 205         if (oid_free)
 206                 /* removals are protected by direct items */
 207                 reiserfs_release_objectid(&th, le32_to_cpu(key->k_objectid));
 208
 209         return journal_end(&th);
 210 }
 211
 212 #ifdef CONFIG_QUOTA
 213 static int reiserfs_quota_on_mount(struct super_block *, int);
 214 #endif
 215
 216 /*
 217  * Look for uncompleted unlinks and truncates and complete them
 218  *
 219  * Called with superblock write locked.  If quotas are enabled, we have to
 220  * release/retake lest we call dquot_quota_on_mount(), proceed to
 221  * schedule_on_each_cpu() in invalidate_bdev() and deadlock waiting for the per
 222  * cpu worklets to complete flush_async_commits() that in turn wait for the
 223  * superblock write lock.
 224  */
 225 static int finish_unfinished(struct super_block *s)
 226 {
 227         INITIALIZE_PATH(path);
 228         struct cpu_key max_cpu_key, obj_key;
 229         struct reiserfs_key save_link_key, last_inode_key;
 230         int retval = 0;
 231         struct item_head *ih;
 232         struct buffer_head *bh;
 233         int item_pos;
 234         char *item;
 235         int done;
 236         struct inode *inode;
 237         int truncate;
 238 #ifdef CONFIG_QUOTA
 239         int i;
 240         int ms_active_set;
 241         int quota_enabled[REISERFS_MAXQUOTAS];
 242 #endif
 243
 244         /* compose key to look for "save" links */
 245         max_cpu_key.version = KEY_FORMAT_3_5;
 246         max_cpu_key.on_disk_key.k_dir_id = ~0U;
 247         max_cpu_key.on_disk_key.k_objectid = ~0U;
 248         set_cpu_key_k_offset(&max_cpu_key, ~0U);
 249         max_cpu_key.key_length = 3;
 250
 251         memset(&last_inode_key, 0, sizeof(last_inode_key));
 252
 253 #ifdef CONFIG_QUOTA
 254         /* Needed for iput() to work correctly and not trash data */
 255         if (s->s_flags & MS_ACTIVE) {
 256                 ms_active_set = 0;
 257         } else {
 258                 ms_active_set = 1;
 259                 s->s_flags |= MS_ACTIVE;
 260         }
 261         /* Turn on quotas so that they are updated correctly */
 262         for (i = 0; i < REISERFS_MAXQUOTAS; i++) {
 263                 quota_enabled[i] = 1;
 264                 if (REISERFS_SB(s)->s_qf_names[i]) {
 265                         int ret;
 266
 267                         if (sb_has_quota_active(s, i)) {
 268                                 quota_enabled[i] = 0;
 269                                 continue;
 270                         }
 271                         reiserfs_write_unlock(s);
 272                         ret = reiserfs_quota_on_mount(s, i);
 273                         reiserfs_write_lock(s);
 274                         if (ret < 0)
 275                                 reiserfs_warning(s, "reiserfs-2500",
 276                                                  "cannot turn on journaled "
 277                                                  "quota: error %d", ret);
 278                 }
 279         }
 280 #endif
 281
 282         done = 0;
 283         REISERFS_SB(s)->s_is_unlinked_ok = 1;
 284         while (!retval) {
 285                 int depth;
 286                 retval = search_item(s, &max_cpu_key, &path);
 287                 if (retval != ITEM_NOT_FOUND) {
 288                         reiserfs_error(s, "vs-2140",
 289                                        "search_by_key returned %d", retval);
 290                         break;
 291                 }
 292
 293                 bh = get_last_bh(&path);
 294                 item_pos = get_item_pos(&path);
 295                 if (item_pos != B_NR_ITEMS(bh)) {
 296                         reiserfs_warning(s, "vs-2060",
 297                                          "wrong position found");
 298                         break;
 299                 }
 300                 item_pos--;
 301                 ih = item_head(bh, item_pos);
 302
 303                 if (le32_to_cpu(ih->ih_key.k_dir_id) != MAX_KEY_OBJECTID)
 304                         /* there are no "save" links anymore */
 305                         break;
 306
 307                 save_link_key = ih->ih_key;
 308                 if (is_indirect_le_ih(ih))
 309                         truncate = 1;
 310                 else
 311                         truncate = 0;
 312
 313                 /* reiserfs_iget needs k_dirid and k_objectid only */
 314                 item = ih_item_body(bh, ih);
 315                 obj_key.on_disk_key.k_dir_id = le32_to_cpu(*(__le32 *) item);
 316                 obj_key.on_disk_key.k_objectid =
 317                     le32_to_cpu(ih->ih_key.k_objectid);
 318                 obj_key.on_disk_key.k_offset = 0;
 319                 obj_key.on_disk_key.k_type = 0;
 320
 321                 pathrelse(&path);
 322
 323                 inode = reiserfs_iget(s, &obj_key);
 324                 if (IS_ERR_OR_NULL(inode)) {
 325                         /*
 326                          * the unlink almost completed, it just did not
 327                          * manage to remove "save" link and release objectid
 328                          */
 329                         reiserfs_warning(s, "vs-2180", "iget failed for %K",
 330                                          &obj_key);
 331                         retval = remove_save_link_only(s, &save_link_key, 1);
 332                         continue;
 333                 }
 334
 335                 if (!truncate && inode->i_nlink) {
 336                         /* file is not unlinked */
 337                         reiserfs_warning(s, "vs-2185",
 338                                          "file %K is not unlinked",
 339                                          &obj_key);
 340                         retval = remove_save_link_only(s, &save_link_key, 0);
 341                         continue;
 342                 }
 343                 depth = reiserfs_write_unlock_nested(inode->i_sb);
 344                 dquot_initialize(inode);
 345                 reiserfs_write_lock_nested(inode->i_sb, depth);
 346
 347                 if (truncate && S_ISDIR(inode->i_mode)) {
 348                         /*
 349                          * We got a truncate request for a dir which
 350                          * is impossible.  The only imaginable way is to
 351                          * execute unfinished truncate request then boot
 352                          * into old kernel, remove the file and create dir
 353                          * with the same key.
 354                          */
 355                         reiserfs_warning(s, "green-2101",
 356                                          "impossible truncate on a "
 357                                          "directory %k. Please report",
 358                                          INODE_PKEY(inode));
 359                         retval = remove_save_link_only(s, &save_link_key, 0);
 360                         truncate = 0;
 361                         iput(inode);
 362                         continue;
 363                 }
 364
 365                 if (truncate) {
 366                         REISERFS_I(inode)->i_flags |=
 367                             i_link_saved_truncate_mask;
 368                         /*
 369                          * not completed truncate found. New size was
 370                          * committed together with "save" link
 371                          */
 372                         reiserfs_info(s, "Truncating %k to %lld ..",
 373                                       INODE_PKEY(inode), inode->i_size);
 374
 375                         /* don't update modification time */
 376                         reiserfs_truncate_file(inode, 0);
 377
 378                         retval = remove_save_link(inode, truncate);
 379                 } else {
 380                         REISERFS_I(inode)->i_flags |= i_link_saved_unlink_mask;
 381                         /* not completed unlink (rmdir) found */
 382                         reiserfs_info(s, "Removing %k..", INODE_PKEY(inode));
 383                         if (memcmp(&last_inode_key, INODE_PKEY(inode),
 384                                         sizeof(last_inode_key))){
 385                                 last_inode_key = *INODE_PKEY(inode);
 386                                 /* removal gets completed in iput */
 387                                 retval = 0;
 388                         } else {
 389                                 reiserfs_warning(s, "super-2189", "Dead loop "
 390                                                  "in finish_unfinished "
 391                                                  "detected, just remove "
 392                                                  "save link\n");
 393                                 retval = remove_save_link_only(s,
 394                                                         &save_link_key, 0);
 395                         }
 396                 }
 397
 398                 iput(inode);
 399                 printk("done\n");
 400                 done++;
 401         }
 402         REISERFS_SB(s)->s_is_unlinked_ok = 0;
 403
 404 #ifdef CONFIG_QUOTA
 405         /* Turn quotas off */
 406         reiserfs_write_unlock(s);
 407         for (i = 0; i < REISERFS_MAXQUOTAS; i++) {
 408                 if (sb_dqopt(s)->files[i] && quota_enabled[i])
 409                         dquot_quota_off(s, i);
 410         }
 411         reiserfs_write_lock(s);
 412         if (ms_active_set)
 413                 /* Restore the flag back */
 414                 s->s_flags &= ~MS_ACTIVE;
 415 #endif
 416         pathrelse(&path);
 417         if (done)
 418                 reiserfs_info(s, "There were %d uncompleted unlinks/truncates. "
 419                               "Completed\n", done);
 420         return retval;
 421 }
 422
 423 /*
 424  * to protect file being unlinked from getting lost we "safe" link files
 425  * being unlinked. This link will be deleted in the same transaction with last
 426  * item of file. mounting the filesystem we scan all these links and remove
 427  * files which almost got lost
 428  */
 429 void add_save_link(struct reiserfs_transaction_handle *th,
 430                    struct inode *inode, int truncate)
 431 {
 432         INITIALIZE_PATH(path);
 433         int retval;
 434         struct cpu_key key;
 435         struct item_head ih;
 436         __le32 link;
 437
 438         BUG_ON(!th->t_trans_id);
 439
 440         /* file can only get one "save link" of each kind */
 441         RFALSE(truncate &&
 442                (REISERFS_I(inode)->i_flags & i_link_saved_truncate_mask),
 443                "saved link already exists for truncated inode %lx",
 444                (long)inode->i_ino);
 445         RFALSE(!truncate &&
 446                (REISERFS_I(inode)->i_flags & i_link_saved_unlink_mask),
 447                "saved link already exists for unlinked inode %lx",
 448                (long)inode->i_ino);
 449
 450         /* setup key of "save" link */
 451         key.version = KEY_FORMAT_3_5;
 452         key.on_disk_key.k_dir_id = MAX_KEY_OBJECTID;
 453         key.on_disk_key.k_objectid = inode->i_ino;
 454         if (!truncate) {
 455                 /* unlink, rmdir, rename */
 456                 set_cpu_key_k_offset(&key, 1 + inode->i_sb->s_blocksize);
 457                 set_cpu_key_k_type(&key, TYPE_DIRECT);
 458
 459                 /* item head of "safe" link */
 460                 make_le_item_head(&ih, &key, key.version,
 461                                   1 + inode->i_sb->s_blocksize, TYPE_DIRECT,
 462                                   4 /*length */ , 0xffff /*free space */ );
 463         } else {
 464                 /* truncate */
 465                 if (S_ISDIR(inode->i_mode))
 466                         reiserfs_warning(inode->i_sb, "green-2102",
 467                                          "Adding a truncate savelink for "
 468                                          "a directory %k! Please report",
 469                                          INODE_PKEY(inode));
 470                 set_cpu_key_k_offset(&key, 1);
 471                 set_cpu_key_k_type(&key, TYPE_INDIRECT);
 472
 473                 /* item head of "safe" link */
 474                 make_le_item_head(&ih, &key, key.version, 1, TYPE_INDIRECT,
 475                                   4 /*length */ , 0 /*free space */ );
 476         }
 477         key.key_length = 3;
 478
 479         /* look for its place in the tree */
 480         retval = search_item(inode->i_sb, &key, &path);
 481         if (retval != ITEM_NOT_FOUND) {
 482                 if (retval != -ENOSPC)
 483                         reiserfs_error(inode->i_sb, "vs-2100",
 484                                        "search_by_key (%K) returned %d", &key,
 485                                        retval);
 486                 pathrelse(&path);
 487                 return;
 488         }
 489
 490         /* body of "save" link */
 491         link = INODE_PKEY(inode)->k_dir_id;
 492
 493         /* put "save" link into tree, don't charge quota to anyone */
 494         retval =
 495             reiserfs_insert_item(th, &path, &key, &ih, NULL, (char *)&link);
 496         if (retval) {
 497                 if (retval != -ENOSPC)
 498                         reiserfs_error(inode->i_sb, "vs-2120",
 499                                        "insert_item returned %d", retval);
 500         } else {
 501                 if (truncate)
 502                         REISERFS_I(inode)->i_flags |=
 503                             i_link_saved_truncate_mask;
 504                 else
 505                         REISERFS_I(inode)->i_flags |= i_link_saved_unlink_mask;
 506         }
 507 }
 508
 509 /* this opens transaction unlike add_save_link */
 510 int remove_save_link(struct inode *inode, int truncate)
 511 {
 512         struct reiserfs_transaction_handle th;
 513         struct reiserfs_key key;
 514         int err;
 515
 516         /* we are going to do one balancing only */
 517         err = journal_begin(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT);
 518         if (err)
 519                 return err;
 520
 521         /* setup key of "save" link */
 522         key.k_dir_id = cpu_to_le32(MAX_KEY_OBJECTID);
 523         key.k_objectid = INODE_PKEY(inode)->k_objectid;
 524         if (!truncate) {
 525                 /* unlink, rmdir, rename */
 526                 set_le_key_k_offset(KEY_FORMAT_3_5, &key,
 527                                     1 + inode->i_sb->s_blocksize);
 528                 set_le_key_k_type(KEY_FORMAT_3_5, &key, TYPE_DIRECT);
 529         } else {
 530                 /* truncate */
 531                 set_le_key_k_offset(KEY_FORMAT_3_5, &key, 1);
 532                 set_le_key_k_type(KEY_FORMAT_3_5, &key, TYPE_INDIRECT);
 533         }
 534
 535         if ((truncate &&
 536              (REISERFS_I(inode)->i_flags & i_link_saved_truncate_mask)) ||
 537             (!truncate &&
 538              (REISERFS_I(inode)->i_flags & i_link_saved_unlink_mask)))
 539                 /* don't take quota bytes from anywhere */
 540                 reiserfs_delete_solid_item(&th, NULL, &key);
 541         if (!truncate) {
 542                 reiserfs_release_objectid(&th, inode->i_ino);
 543                 REISERFS_I(inode)->i_flags &= ~i_link_saved_unlink_mask;
 544         } else
 545                 REISERFS_I(inode)->i_flags &= ~i_link_saved_truncate_mask;
 546
 547         return journal_end(&th);
 548 }
 549
 550 static void reiserfs_kill_sb(struct super_block *s)
 551 {
 552         if (REISERFS_SB(s)) {
 553                 reiserfs_proc_info_done(s);
 554                 /*
 555                  * Force any pending inode evictions to occur now. Any
 556                  * inodes to be removed that have extended attributes
 557                  * associated with them need to clean them up before
 558                  * we can release the extended attribute root dentries.
 559                  * shrink_dcache_for_umount will BUG if we don't release
 560                  * those before it's called so ->put_super is too late.
 561                  */
 562                 shrink_dcache_sb(s);
 563
 564                 dput(REISERFS_SB(s)->xattr_root);
 565                 REISERFS_SB(s)->xattr_root = NULL;
 566                 dput(REISERFS_SB(s)->priv_root);
 567                 REISERFS_SB(s)->priv_root = NULL;
 568         }
 569
 570         kill_block_super(s);
 571 }
 572
 573 #ifdef CONFIG_QUOTA
 574 static int reiserfs_quota_off(struct super_block *sb, int type);
 575
 576 static void reiserfs_quota_off_umount(struct super_block *s)
 577 {
 578         int type;
 579
 580         for (type = 0; type < REISERFS_MAXQUOTAS; type++)
 581                 reiserfs_quota_off(s, type);
 582 }
 583 #else
 584 static inline void reiserfs_quota_off_umount(struct super_block *s)
 585 {
 586 }
 587 #endif
 588
 589 static void reiserfs_put_super(struct super_block *s)
 590 {
 591         struct reiserfs_transaction_handle th;
 592         th.t_trans_id = 0;
 593
 594         reiserfs_quota_off_umount(s);
 595
 596         reiserfs_write_lock(s);
 597
 598         /*
 599          * change file system state to current state if it was mounted
 600          * with read-write permissions
 601          */
 602         if (!(s->s_flags & MS_RDONLY)) {
 603                 if (!journal_begin(&th, s, 10)) {
 604                         reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s),
 605                                                      1);
 606                         set_sb_umount_state(SB_DISK_SUPER_BLOCK(s),
 607                                             REISERFS_SB(s)->s_mount_state);
 608                         journal_mark_dirty(&th, SB_BUFFER_WITH_SB(s));
 609                 }
 610         }
 611
 612         /*
 613          * note, journal_release checks for readonly mount, and can
 614          * decide not to do a journal_end
 615          */
 616         journal_release(&th, s);
 617
 618         reiserfs_free_bitmap_cache(s);
 619
 620         brelse(SB_BUFFER_WITH_SB(s));
 621
 622         print_statistics(s);
 623
 624         if (REISERFS_SB(s)->reserved_blocks != 0) {
 625                 reiserfs_warning(s, "green-2005", "reserved blocks left %d",
 626                                  REISERFS_SB(s)->reserved_blocks);
 627         }
 628
 629         reiserfs_write_unlock(s);
 630         mutex_destroy(&REISERFS_SB(s)->lock);
 631         destroy_workqueue(REISERFS_SB(s)->commit_wq);
 632         kfree(s->s_fs_info);
 633         s->s_fs_info = NULL;
 634 }
 635
 636 static struct kmem_cache *reiserfs_inode_cachep;
 637
 638 static struct inode *reiserfs_alloc_inode(struct super_block *sb)
 639 {
 640         struct reiserfs_inode_info *ei;
 641         ei = kmem_cache_alloc(reiserfs_inode_cachep, GFP_KERNEL);
 642         if (!ei)
 643                 return NULL;
 644         atomic_set(&ei->openers, 0);
 645         mutex_init(&ei->tailpack);
 646 #ifdef CONFIG_QUOTA
 647         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 648 #endif
 649
 650         return &ei->vfs_inode;
 651 }
 652
 653 static void reiserfs_i_callback(struct rcu_head *head)
 654 {
 655         struct inode *inode = container_of(head, struct inode, i_rcu);
 656         kmem_cache_free(reiserfs_inode_cachep, REISERFS_I(inode));
 657 }
 658
 659 static void reiserfs_destroy_inode(struct inode *inode)
 660 {
 661         call_rcu(&inode->i_rcu, reiserfs_i_callback);
 662 }
 663
 664 static void init_once(void *foo)
 665 {
 666         struct reiserfs_inode_info *ei = (struct reiserfs_inode_info *)foo;
 667
 668         INIT_LIST_HEAD(&ei->i_prealloc_list);
 669         inode_init_once(&ei->vfs_inode);
 670 }
 671
 672 static int __init init_inodecache(void)
 673 {
 674         reiserfs_inode_cachep = kmem_cache_create("reiser_inode_cache",
 675                                                   sizeof(struct
 676                                                          reiserfs_inode_info),
 677                                                   0, (SLAB_RECLAIM_ACCOUNT|
 678                                                       SLAB_MEM_SPREAD|
 679                                                       SLAB_ACCOUNT),
 680                                                   init_once);
 681         if (reiserfs_inode_cachep == NULL)
 682                 return -ENOMEM;
 683         return 0;
 684 }
 685
 686 static void destroy_inodecache(void)
 687 {
 688         /*
 689          * Make sure all delayed rcu free inodes are flushed before we
 690          * destroy cache.
 691          */
 692         rcu_barrier();
 693         kmem_cache_destroy(reiserfs_inode_cachep);
 694 }
 695
 696 /* we don't mark inodes dirty, we just log them */
 697 static void reiserfs_dirty_inode(struct inode *inode, int flags)
 698 {
 699         struct reiserfs_transaction_handle th;
 700
 701         int err = 0;
 702
 703         if (inode->i_sb->s_flags & MS_RDONLY) {
 704                 reiserfs_warning(inode->i_sb, "clm-6006",
 705                                  "writing inode %lu on readonly FS",
 706                                  inode->i_ino);
 707                 return;
 708         }
 709         reiserfs_write_lock(inode->i_sb);
 710
 711         /*
 712          * this is really only used for atime updates, so they don't have
 713          * to be included in O_SYNC or fsync
 714          */
 715         err = journal_begin(&th, inode->i_sb, 1);
 716         if (err)
 717                 goto out;
 718
 719         reiserfs_update_sd(&th, inode);
 720         journal_end(&th);
 721
 722 out:
 723         reiserfs_write_unlock(inode->i_sb);
 724 }
 725
 726 static int reiserfs_show_options(struct seq_file *seq, struct dentry *root)
 727 {
 728         struct super_block *s = root->d_sb;
 729         struct reiserfs_journal *journal = SB_JOURNAL(s);
 730         long opts = REISERFS_SB(s)->s_mount_opt;
 731
 732         if (opts & (1 << REISERFS_LARGETAIL))
 733                 seq_puts(seq, ",tails=on");
 734         else if (!(opts & (1 << REISERFS_SMALLTAIL)))
 735                 seq_puts(seq, ",notail");
 736         /* tails=small is default so we don't show it */
 737
 738         if (!(opts & (1 << REISERFS_BARRIER_FLUSH)))
 739                 seq_puts(seq, ",barrier=none");
 740         /* barrier=flush is default so we don't show it */
 741
 742         if (opts & (1 << REISERFS_ERROR_CONTINUE))
 743                 seq_puts(seq, ",errors=continue");
 744         else if (opts & (1 << REISERFS_ERROR_PANIC))
 745                 seq_puts(seq, ",errors=panic");
 746         /* errors=ro is default so we don't show it */
 747
 748         if (opts & (1 << REISERFS_DATA_LOG))
 749                 seq_puts(seq, ",data=journal");
 750         else if (opts & (1 << REISERFS_DATA_WRITEBACK))
 751                 seq_puts(seq, ",data=writeback");
 752         /* data=ordered is default so we don't show it */
 753
 754         if (opts & (1 << REISERFS_ATTRS))
 755                 seq_puts(seq, ",attrs");
 756
 757         if (opts & (1 << REISERFS_XATTRS_USER))
 758                 seq_puts(seq, ",user_xattr");
 759
 760         if (opts & (1 << REISERFS_EXPOSE_PRIVROOT))
 761                 seq_puts(seq, ",expose_privroot");
 762
 763         if (opts & (1 << REISERFS_POSIXACL))
 764                 seq_puts(seq, ",acl");
 765
 766         if (REISERFS_SB(s)->s_jdev)
 767                 seq_show_option(seq, "jdev", REISERFS_SB(s)->s_jdev);
 768
 769         if (journal->j_max_commit_age != journal->j_default_max_commit_age)
 770                 seq_printf(seq, ",commit=%d", journal->j_max_commit_age);
 771
 772 #ifdef CONFIG_QUOTA
 773         if (REISERFS_SB(s)->s_qf_names[USRQUOTA])
 774                 seq_show_option(seq, "usrjquota",
 775                                 REISERFS_SB(s)->s_qf_names[USRQUOTA]);
 776         else if (opts & (1 << REISERFS_USRQUOTA))
 777                 seq_puts(seq, ",usrquota");
 778         if (REISERFS_SB(s)->s_qf_names[GRPQUOTA])
 779                 seq_show_option(seq, "grpjquota",
 780                                 REISERFS_SB(s)->s_qf_names[GRPQUOTA]);
 781         else if (opts & (1 << REISERFS_GRPQUOTA))
 782                 seq_puts(seq, ",grpquota");
 783         if (REISERFS_SB(s)->s_jquota_fmt) {
 784                 if (REISERFS_SB(s)->s_jquota_fmt == QFMT_VFS_OLD)
 785                         seq_puts(seq, ",jqfmt=vfsold");
 786                 else if (REISERFS_SB(s)->s_jquota_fmt == QFMT_VFS_V0)
 787                         seq_puts(seq, ",jqfmt=vfsv0");
 788         }
 789 #endif
 790
 791         /* Block allocator options */
 792         if (opts & (1 << REISERFS_NO_BORDER))
 793                 seq_puts(seq, ",block-allocator=noborder");
 794         if (opts & (1 << REISERFS_NO_UNHASHED_RELOCATION))
 795                 seq_puts(seq, ",block-allocator=no_unhashed_relocation");
 796         if (opts & (1 << REISERFS_HASHED_RELOCATION))
 797                 seq_puts(seq, ",block-allocator=hashed_relocation");
 798         if (opts & (1 << REISERFS_TEST4))
 799                 seq_puts(seq, ",block-allocator=test4");
 800         show_alloc_options(seq, s);
 801         return 0;
 802 }
 803
 804 #ifdef CONFIG_QUOTA
 805 static ssize_t reiserfs_quota_write(struct super_block *, int, const char *,
 806                                     size_t, loff_t);
 807 static ssize_t reiserfs_quota_read(struct super_block *, int, char *, size_t,
 808                                    loff_t);
 809
 810 static struct dquot **reiserfs_get_dquots(struct inode *inode)
 811 {
 812         return REISERFS_I(inode)->i_dquot;
 813 }
 814 #endif
 815
 816 static const struct super_operations reiserfs_sops = {
 817         .alloc_inode = reiserfs_alloc_inode,
 818         .destroy_inode = reiserfs_destroy_inode,
 819         .write_inode = reiserfs_write_inode,
 820         .dirty_inode = reiserfs_dirty_inode,
 821         .evict_inode = reiserfs_evict_inode,
 822         .put_super = reiserfs_put_super,
 823         .sync_fs = reiserfs_sync_fs,
 824         .freeze_fs = reiserfs_freeze,
 825         .unfreeze_fs = reiserfs_unfreeze,
 826         .statfs = reiserfs_statfs,
 827         .remount_fs = reiserfs_remount,
 828         .show_options = reiserfs_show_options,
 829 #ifdef CONFIG_QUOTA
 830         .quota_read = reiserfs_quota_read,
 831         .quota_write = reiserfs_quota_write,
 832         .get_dquots = reiserfs_get_dquots,
 833 #endif
 834 };
 835
 836 #ifdef CONFIG_QUOTA
 837 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
 838
 839 static int reiserfs_write_dquot(struct dquot *);
 840 static int reiserfs_acquire_dquot(struct dquot *);
 841 static int reiserfs_release_dquot(struct dquot *);
 842 static int reiserfs_mark_dquot_dirty(struct dquot *);
 843 static int reiserfs_write_info(struct super_block *, int);
 844 static int reiserfs_quota_on(struct super_block *, int, int, const struct path *);
 845
 846 static const struct dquot_operations reiserfs_quota_operations = {
 847         .write_dquot = reiserfs_write_dquot,
 848         .acquire_dquot = reiserfs_acquire_dquot,
 849         .release_dquot = reiserfs_release_dquot,
 850         .mark_dirty = reiserfs_mark_dquot_dirty,
 851         .write_info = reiserfs_write_info,
 852         .alloc_dquot    = dquot_alloc,
 853         .destroy_dquot  = dquot_destroy,
 854         .get_next_id    = dquot_get_next_id,
 855 };
 856
 857 static const struct quotactl_ops reiserfs_qctl_operations = {
 858         .quota_on = reiserfs_quota_on,
 859         .quota_off = reiserfs_quota_off,
 860         .quota_sync = dquot_quota_sync,
 861         .get_state = dquot_get_state,
 862         .set_info = dquot_set_dqinfo,
 863         .get_dqblk = dquot_get_dqblk,
 864         .set_dqblk = dquot_set_dqblk,
 865 };
 866 #endif
 867
 868 static const struct export_operations reiserfs_export_ops = {
 869         .encode_fh = reiserfs_encode_fh,
 870         .fh_to_dentry = reiserfs_fh_to_dentry,
 871         .fh_to_parent = reiserfs_fh_to_parent,
 872         .get_parent = reiserfs_get_parent,
 873 };
 874
 875 /*
 876  * this struct is used in reiserfs_getopt () for containing the value for
 877  * those mount options that have values rather than being toggles.
 878  */
 879 typedef struct {
 880         char *value;
 881         /*
 882          * bitmask which is to set on mount_options bitmask
 883          * when this value is found, 0 is no bits are to be changed.
 884          */
 885         int setmask;
 886         /*
 887          * bitmask which is to clear on mount_options bitmask
 888          * when this value is found, 0 is no bits are to be changed.
 889          * This is applied BEFORE setmask
 890          */
 891         int clrmask;
 892 } arg_desc_t;
 893
 894 /* Set this bit in arg_required to allow empty arguments */
 895 #define REISERFS_OPT_ALLOWEMPTY 31
 896
 897 /*
 898  * this struct is used in reiserfs_getopt() for describing the
 899  * set of reiserfs mount options
 900  */
 901 typedef struct {
 902         char *option_name;
 903
 904         /* 0 if argument is not required, not 0 otherwise */
 905         int arg_required;
 906
 907         /* list of values accepted by an option */
 908         const arg_desc_t *values;
 909
 910         /*
 911          * bitmask which is to set on mount_options bitmask
 912          * when this value is found, 0 is no bits are to be changed.
 913          */
 914         int setmask;
 915
 916         /*
 917          * bitmask which is to clear on mount_options bitmask
 918          * when this value is found, 0 is no bits are to be changed.
 919          * This is applied BEFORE setmask
 920          */
 921         int clrmask;
 922 } opt_desc_t;
 923
 924 /* possible values for -o data= */
 925 static const arg_desc_t logging_mode[] = {
 926         {"ordered", 1 << REISERFS_DATA_ORDERED,
 927          (1 << REISERFS_DATA_LOG | 1 << REISERFS_DATA_WRITEBACK)},
 928         {"journal", 1 << REISERFS_DATA_LOG,
 929          (1 << REISERFS_DATA_ORDERED | 1 << REISERFS_DATA_WRITEBACK)},
 930         {"writeback", 1 << REISERFS_DATA_WRITEBACK,
 931          (1 << REISERFS_DATA_ORDERED | 1 << REISERFS_DATA_LOG)},
 932         {.value = NULL}
 933 };
 934
 935 /* possible values for -o barrier= */
 936 static const arg_desc_t barrier_mode[] = {
 937         {"none", 1 << REISERFS_BARRIER_NONE, 1 << REISERFS_BARRIER_FLUSH},
 938         {"flush", 1 << REISERFS_BARRIER_FLUSH, 1 << REISERFS_BARRIER_NONE},
 939         {.value = NULL}
 940 };
 941
 942 /*
 943  * possible values for "-o block-allocator=" and bits which are to be set in
 944  * s_mount_opt of reiserfs specific part of in-core super block
 945  */
 946 static const arg_desc_t balloc[] = {
 947         {"noborder", 1 << REISERFS_NO_BORDER, 0},
 948         {"border", 0, 1 << REISERFS_NO_BORDER},
 949         {"no_unhashed_relocation", 1 << REISERFS_NO_UNHASHED_RELOCATION, 0},
 950         {"hashed_relocation", 1 << REISERFS_HASHED_RELOCATION, 0},
 951         {"test4", 1 << REISERFS_TEST4, 0},
 952         {"notest4", 0, 1 << REISERFS_TEST4},
 953         {NULL, 0, 0}
 954 };
 955
 956 static const arg_desc_t tails[] = {
 957         {"on", 1 << REISERFS_LARGETAIL, 1 << REISERFS_SMALLTAIL},
 958         {"off", 0, (1 << REISERFS_LARGETAIL) | (1 << REISERFS_SMALLTAIL)},
 959         {"small", 1 << REISERFS_SMALLTAIL, 1 << REISERFS_LARGETAIL},
 960         {NULL, 0, 0}
 961 };
 962
 963 static const arg_desc_t error_actions[] = {
 964         {"panic", 1 << REISERFS_ERROR_PANIC,
 965          (1 << REISERFS_ERROR_RO | 1 << REISERFS_ERROR_CONTINUE)},
 966         {"ro-remount", 1 << REISERFS_ERROR_RO,
 967          (1 << REISERFS_ERROR_PANIC | 1 << REISERFS_ERROR_CONTINUE)},
 968 #ifdef REISERFS_JOURNAL_ERROR_ALLOWS_NO_LOG
 969         {"continue", 1 << REISERFS_ERROR_CONTINUE,
 970          (1 << REISERFS_ERROR_PANIC | 1 << REISERFS_ERROR_RO)},
 971 #endif
 972         {NULL, 0, 0},
 973 };
 974
 975 /*
 976  * proceed only one option from a list *cur - string containing of mount
 977  * options
 978  * opts - array of options which are accepted
 979  * opt_arg - if option is found and requires an argument and if it is specifed
 980  * in the input - pointer to the argument is stored here
 981  * bit_flags - if option requires to set a certain bit - it is set here
 982  * return -1 if unknown option is found, opt->arg_required otherwise
 983  */
 984 static int reiserfs_getopt(struct super_block *s, char **cur, opt_desc_t * opts,
 985                            char **opt_arg, unsigned long *bit_flags)
 986 {
 987         char *p;
 988         /*
 989          * foo=bar,
 990          * ^   ^  ^
 991          * |   |  +-- option_end
 992          * |   +-- arg_start
 993          * +-- option_start
 994          */
 995         const opt_desc_t *opt;
 996         const arg_desc_t *arg;
 997
 998         p = *cur;
 999
1000         /* assume argument cannot contain commas */
1001         *cur = strchr(p, ',');
1002         if (*cur) {
1003                 *(*cur) = '\0';
1004                 (*cur)++;
1005         }
1006
1007         if (!strncmp(p, "alloc=", 6)) {
1008                 /*
1009                  * Ugly special case, probably we should redo options
1010                  * parser so that it can understand several arguments for
1011                  * some options, also so that it can fill several bitfields
1012                  * with option values.
1013                  */
1014                 if (reiserfs_parse_alloc_options(s, p + 6)) {
1015                         return -1;
1016                 } else {
1017                         return 0;
1018                 }
1019         }
1020
1021         /* for every option in the list */
1022         for (opt = opts; opt->option_name; opt++) {
1023                 if (!strncmp(p, opt->option_name, strlen(opt->option_name))) {
1024                         if (bit_flags) {
1025                                 if (opt->clrmask ==
1026                                     (1 << REISERFS_UNSUPPORTED_OPT))
1027                                         reiserfs_warning(s, "super-6500",
1028                                                          "%s not supported.\n",
1029                                                          p);
1030                                 else
1031                                         *bit_flags &= ~opt->clrmask;
1032                                 if (opt->setmask ==
1033                                     (1 << REISERFS_UNSUPPORTED_OPT))
1034                                         reiserfs_warning(s, "super-6501",
1035                                                          "%s not supported.\n",
1036                                                          p);
1037                                 else
1038                                         *bit_flags |= opt->setmask;
1039                         }
1040                         break;
1041                 }
1042         }
1043         if (!opt->option_name) {
1044                 reiserfs_warning(s, "super-6502",
1045                                  "unknown mount option \"%s\"", p);
1046                 return -1;
1047         }
1048
1049         p += strlen(opt->option_name);
1050         switch (*p) {
1051         case '=':
1052                 if (!opt->arg_required) {
1053                         reiserfs_warning(s, "super-6503",
1054                                          "the option \"%s\" does not "
1055                                          "require an argument\n",
1056                                          opt->option_name);
1057                         return -1;
1058                 }
1059                 break;
1060
1061         case 0:
1062                 if (opt->arg_required) {
1063                         reiserfs_warning(s, "super-6504",
1064                                          "the option \"%s\" requires an "
1065                                          "argument\n", opt->option_name);
1066                         return -1;
1067                 }
1068                 break;
1069         default:
1070                 reiserfs_warning(s, "super-6505",
1071                                  "head of option \"%s\" is only correct\n",
1072                                  opt->option_name);
1073                 return -1;
1074         }
1075
1076         /*
1077          * move to the argument, or to next option if argument is not
1078          * required
1079          */
1080         p++;
1081
1082         if (opt->arg_required
1083             && !(opt->arg_required & (1 << REISERFS_OPT_ALLOWEMPTY))
1084             && !strlen(p)) {
1085                 /* this catches "option=," if not allowed */
1086                 reiserfs_warning(s, "super-6506",
1087                                  "empty argument for \"%s\"\n",
1088                                  opt->option_name);
1089                 return -1;
1090         }
1091
1092         if (!opt->values) {
1093                 /* *=NULLopt_arg contains pointer to argument */
1094                 *opt_arg = p;
1095                 return opt->arg_required & ~(1 << REISERFS_OPT_ALLOWEMPTY);
1096         }
1097
1098         /* values possible for this option are listed in opt->values */
1099         for (arg = opt->values; arg->value; arg++) {
1100                 if (!strcmp(p, arg->value)) {
1101                         if (bit_flags) {
1102                                 *bit_flags &= ~arg->clrmask;
1103                                 *bit_flags |= arg->setmask;
1104                         }
1105                         return opt->arg_required;
1106                 }
1107         }
1108
1109         reiserfs_warning(s, "super-6506",
1110                          "bad value \"%s\" for option \"%s\"\n", p,
1111                          opt->option_name);
1112         return -1;
1113 }
1114
1115 /* returns 0 if something is wrong in option string, 1 - otherwise */
1116 static int reiserfs_parse_options(struct super_block *s,
1117
1118                                   /* string given via mount's -o */
1119                                   char *options,
1120
1121                                   /*
1122                                    * after the parsing phase, contains the
1123                                    * collection of bitflags defining what
1124                                    * mount options were selected.
1125                                    */
1126                                   unsigned long *mount_options,
1127
1128                                   /* strtol-ed from NNN of resize=NNN */
1129                                   unsigned long *blocks,
1130                                   char **jdev_name,
1131                                   unsigned int *commit_max_age,
1132                                   char **qf_names,
1133                                   unsigned int *qfmt)
1134 {
1135         int c;
1136         char *arg = NULL;
1137         char *pos;
1138         opt_desc_t opts[] = {
1139                 /*
1140                  * Compatibility stuff, so that -o notail for old
1141                  * setups still work
1142                  */
1143                 {"tails",.arg_required = 't',.values = tails},
1144                 {"notail",.clrmask =
1145                  (1 << REISERFS_LARGETAIL) | (1 << REISERFS_SMALLTAIL)},
1146                 {"conv",.setmask = 1 << REISERFS_CONVERT},
1147                 {"attrs",.setmask = 1 << REISERFS_ATTRS},
1148                 {"noattrs",.clrmask = 1 << REISERFS_ATTRS},
1149                 {"expose_privroot", .setmask = 1 << REISERFS_EXPOSE_PRIVROOT},
1150 #ifdef CONFIG_REISERFS_FS_XATTR
1151                 {"user_xattr",.setmask = 1 << REISERFS_XATTRS_USER},
1152                 {"nouser_xattr",.clrmask = 1 << REISERFS_XATTRS_USER},
1153 #else
1154                 {"user_xattr",.setmask = 1 << REISERFS_UNSUPPORTED_OPT},
1155                 {"nouser_xattr",.clrmask = 1 << REISERFS_UNSUPPORTED_OPT},
1156 #endif
1157 #ifdef CONFIG_REISERFS_FS_POSIX_ACL
1158                 {"acl",.setmask = 1 << REISERFS_POSIXACL},
1159                 {"noacl",.clrmask = 1 << REISERFS_POSIXACL},
1160 #else
1161                 {"acl",.setmask = 1 << REISERFS_UNSUPPORTED_OPT},
1162                 {"noacl",.clrmask = 1 << REISERFS_UNSUPPORTED_OPT},
1163 #endif
1164                 {.option_name = "nolog"},
1165                 {"replayonly",.setmask = 1 << REPLAYONLY},
1166                 {"block-allocator",.arg_required = 'a',.values = balloc},
1167                 {"data",.arg_required = 'd',.values = logging_mode},
1168                 {"barrier",.arg_required = 'b',.values = barrier_mode},
1169                 {"resize",.arg_required = 'r',.values = NULL},
1170                 {"jdev",.arg_required = 'j',.values = NULL},
1171                 {"nolargeio",.arg_required = 'w',.values = NULL},
1172                 {"commit",.arg_required = 'c',.values = NULL},
1173                 {"usrquota",.setmask = 1 << REISERFS_USRQUOTA},
1174                 {"grpquota",.setmask = 1 << REISERFS_GRPQUOTA},
1175                 {"noquota",.clrmask = 1 << REISERFS_USRQUOTA | 1 << REISERFS_GRPQUOTA},
1176                 {"errors",.arg_required = 'e',.values = error_actions},
1177                 {"usrjquota",.arg_required =
1178                  'u' | (1 << REISERFS_OPT_ALLOWEMPTY),.values = NULL},
1179                 {"grpjquota",.arg_required =
1180                  'g' | (1 << REISERFS_OPT_ALLOWEMPTY),.values = NULL},
1181                 {"jqfmt",.arg_required = 'f',.values = NULL},
1182                 {.option_name = NULL}
1183         };
1184
1185         *blocks = 0;
1186         if (!options || !*options)
1187                 /*
1188                  * use default configuration: create tails, journaling on, no
1189                  * conversion to newest format
1190                  */
1191                 return 1;
1192
1193         for (pos = options; pos;) {
1194                 c = reiserfs_getopt(s, &pos, opts, &arg, mount_options);
1195                 if (c == -1)
1196                         /* wrong option is given */
1197                         return 0;
1198
1199                 if (c == 'r') {
1200                         char *p;
1201
1202                         p = NULL;
1203                         /* "resize=NNN" or "resize=auto" */
1204
1205                         if (!strcmp(arg, "auto")) {
1206                                 /* From JFS code, to auto-get the size. */
1207                                 *blocks =
1208                                     i_size_read(s->s_bdev->bd_inode) >> s->
1209                                     s_blocksize_bits;
1210                         } else {
1211                                 *blocks = simple_strtoul(arg, &p, 0);
1212                                 if (*p != '\0') {
1213                                         /* NNN does not look like a number */
1214                                         reiserfs_warning(s, "super-6507",
1215                                                          "bad value %s for "
1216                                                          "-oresize\n", arg);
1217                                         return 0;
1218                                 }
1219                         }
1220                 }
1221
1222                 if (c == 'c') {
1223                         char *p = NULL;
1224                         unsigned long val = simple_strtoul(arg, &p, 0);
1225                         /* commit=NNN (time in seconds) */
1226                         if (*p != '\0' || val >= (unsigned int)-1) {
1227                                 reiserfs_warning(s, "super-6508",
1228                                                  "bad value %s for -ocommit\n",
1229                                                  arg);
1230                                 return 0;
1231                         }
1232                         *commit_max_age = (unsigned int)val;
1233                 }
1234
1235                 if (c == 'w') {
1236                         reiserfs_warning(s, "super-6509", "nolargeio option "
1237                                          "is no longer supported");
1238                         return 0;
1239                 }
1240
1241                 if (c == 'j') {
1242                         if (arg && *arg && jdev_name) {
1243                                 /* Hm, already assigned? */
1244                                 if (*jdev_name) {
1245                                         reiserfs_warning(s, "super-6510",
1246                                                          "journal device was "
1247                                                          "already specified to "
1248                                                          "be %s", *jdev_name);
1249                                         return 0;
1250                                 }
1251                                 *jdev_name = arg;
1252                         }
1253                 }
1254 #ifdef CONFIG_QUOTA
1255                 if (c == 'u' || c == 'g') {
1256                         int qtype = c == 'u' ? USRQUOTA : GRPQUOTA;
1257
1258                         if (sb_any_quota_loaded(s) &&
1259                             (!*arg != !REISERFS_SB(s)->s_qf_names[qtype])) {
1260                                 reiserfs_warning(s, "super-6511",
1261                                                  "cannot change journaled "
1262                                                  "quota options when quota "
1263                                                  "turned on.");
1264                                 return 0;
1265                         }
1266                         if (*arg) {     /* Some filename specified? */
1267                                 if (REISERFS_SB(s)->s_qf_names[qtype]
1268                                     && strcmp(REISERFS_SB(s)->s_qf_names[qtype],
1269                                               arg)) {
1270                                         reiserfs_warning(s, "super-6512",
1271                                                          "%s quota file "
1272                                                          "already specified.",
1273                                                          QTYPE2NAME(qtype));
1274                                         return 0;
1275                                 }
1276                                 if (strchr(arg, '/')) {
1277                                         reiserfs_warning(s, "super-6513",
1278                                                          "quotafile must be "
1279                                                          "on filesystem root.");
1280                                         return 0;
1281                                 }
1282                                 qf_names[qtype] = kstrdup(arg, GFP_KERNEL);
1283                                 if (!qf_names[qtype]) {
1284                                         reiserfs_warning(s, "reiserfs-2502",
1285                                                          "not enough memory "
1286                                                          "for storing "
1287                                                          "quotafile name.");
1288                                         return 0;
1289                                 }
1290                                 if (qtype == USRQUOTA)
1291                                         *mount_options |= 1 << REISERFS_USRQUOTA;
1292                                 else
1293                                         *mount_options |= 1 << REISERFS_GRPQUOTA;
1294                         } else {
1295                                 if (qf_names[qtype] !=
1296                                     REISERFS_SB(s)->s_qf_names[qtype])
1297                                         kfree(qf_names[qtype]);
1298                                 qf_names[qtype] = NULL;
1299                                 if (qtype == USRQUOTA)
1300                                         *mount_options &= ~(1 << REISERFS_USRQUOTA);
1301                                 else
1302                                         *mount_options &= ~(1 << REISERFS_GRPQUOTA);
1303                         }
1304                 }
1305                 if (c == 'f') {
1306                         if (!strcmp(arg, "vfsold"))
1307                                 *qfmt = QFMT_VFS_OLD;
1308                         else if (!strcmp(arg, "vfsv0"))
1309                                 *qfmt = QFMT_VFS_V0;
1310                         else {
1311                                 reiserfs_warning(s, "super-6514",
1312                                                  "unknown quota format "
1313                                                  "specified.");
1314                                 return 0;
1315                         }
1316                         if (sb_any_quota_loaded(s) &&
1317                             *qfmt != REISERFS_SB(s)->s_jquota_fmt) {
1318                                 reiserfs_warning(s, "super-6515",
1319                                                  "cannot change journaled "
1320                                                  "quota options when quota "
1321                                                  "turned on.");
1322                                 return 0;
1323                         }
1324                 }
1325 #else
1326                 if (c == 'u' || c == 'g' || c == 'f') {
1327                         reiserfs_warning(s, "reiserfs-2503", "journaled "
1328                                          "quota options not supported.");
1329                         return 0;
1330                 }
1331 #endif
1332         }
1333
1334 #ifdef CONFIG_QUOTA
1335         if (!REISERFS_SB(s)->s_jquota_fmt && !*qfmt
1336             && (qf_names[USRQUOTA] || qf_names[GRPQUOTA])) {
1337                 reiserfs_warning(s, "super-6515",
1338                                  "journaled quota format not specified.");
1339                 return 0;
1340         }
1341         if ((!(*mount_options & (1 << REISERFS_USRQUOTA)) &&
1342                sb_has_quota_loaded(s, USRQUOTA)) ||
1343             (!(*mount_options & (1 << REISERFS_GRPQUOTA)) &&
1344                sb_has_quota_loaded(s, GRPQUOTA))) {
1345                 reiserfs_warning(s, "super-6516", "quota options must "
1346                                  "be present when quota is turned on.");
1347                 return 0;
1348         }
1349 #endif
1350
1351         return 1;
1352 }
1353
1354 static void switch_data_mode(struct super_block *s, unsigned long mode)
1355 {
1356         REISERFS_SB(s)->s_mount_opt &= ~((1 << REISERFS_DATA_LOG) |
1357                                          (1 << REISERFS_DATA_ORDERED) |
1358                                          (1 << REISERFS_DATA_WRITEBACK));
1359         REISERFS_SB(s)->s_mount_opt |= (1 << mode);
1360 }
1361
1362 static void handle_data_mode(struct super_block *s, unsigned long mount_options)
1363 {
1364         if (mount_options & (1 << REISERFS_DATA_LOG)) {
1365                 if (!reiserfs_data_log(s)) {
1366                         switch_data_mode(s, REISERFS_DATA_LOG);
1367                         reiserfs_info(s, "switching to journaled data mode\n");
1368                 }
1369         } else if (mount_options & (1 << REISERFS_DATA_ORDERED)) {
1370                 if (!reiserfs_data_ordered(s)) {
1371                         switch_data_mode(s, REISERFS_DATA_ORDERED);
1372                         reiserfs_info(s, "switching to ordered data mode\n");
1373                 }
1374         } else if (mount_options & (1 << REISERFS_DATA_WRITEBACK)) {
1375                 if (!reiserfs_data_writeback(s)) {
1376                         switch_data_mode(s, REISERFS_DATA_WRITEBACK);
1377                         reiserfs_info(s, "switching to writeback data mode\n");
1378                 }
1379         }
1380 }
1381
1382 static void handle_barrier_mode(struct super_block *s, unsigned long bits)
1383 {
1384         int flush = (1 << REISERFS_BARRIER_FLUSH);
1385         int none = (1 << REISERFS_BARRIER_NONE);
1386         int all_barrier = flush | none;
1387
1388         if (bits & all_barrier) {
1389                 REISERFS_SB(s)->s_mount_opt &= ~all_barrier;
1390                 if (bits & flush) {
1391                         REISERFS_SB(s)->s_mount_opt |= flush;
1392                         printk("reiserfs: enabling write barrier flush mode\n");
1393                 } else if (bits & none) {
1394                         REISERFS_SB(s)->s_mount_opt |= none;
1395                         printk("reiserfs: write barriers turned off\n");
1396                 }
1397         }
1398 }
1399
1400 static void handle_attrs(struct super_block *s)
1401 {
1402         struct reiserfs_super_block *rs = SB_DISK_SUPER_BLOCK(s);
1403
1404         if (reiserfs_attrs(s)) {
1405                 if (old_format_only(s)) {
1406                         reiserfs_warning(s, "super-6517", "cannot support "
1407                                          "attributes on 3.5.x disk format");
1408                         REISERFS_SB(s)->s_mount_opt &= ~(1 << REISERFS_ATTRS);
1409                         return;
1410                 }
1411                 if (!(le32_to_cpu(rs->s_flags) & reiserfs_attrs_cleared)) {
1412                         reiserfs_warning(s, "super-6518", "cannot support "
1413                                          "attributes until flag is set in "
1414                                          "super-block");
1415                         REISERFS_SB(s)->s_mount_opt &= ~(1 << REISERFS_ATTRS);
1416                 }
1417         }
1418 }
1419
1420 #ifdef CONFIG_QUOTA
1421 static void handle_quota_files(struct super_block *s, char **qf_names,
1422                                unsigned int *qfmt)
1423 {
1424         int i;
1425
1426         for (i = 0; i < REISERFS_MAXQUOTAS; i++) {
1427                 if (qf_names[i] != REISERFS_SB(s)->s_qf_names[i])
1428                         kfree(REISERFS_SB(s)->s_qf_names[i]);
1429                 REISERFS_SB(s)->s_qf_names[i] = qf_names[i];
1430         }
1431         if (*qfmt)
1432                 REISERFS_SB(s)->s_jquota_fmt = *qfmt;
1433 }
1434 #endif
1435
1436 static int reiserfs_remount(struct super_block *s, int *mount_flags, char *arg)
1437 {
1438         struct reiserfs_super_block *rs;
1439         struct reiserfs_transaction_handle th;
1440         unsigned long blocks;
1441         unsigned long mount_options = REISERFS_SB(s)->s_mount_opt;
1442         unsigned long safe_mask = 0;
1443         unsigned int commit_max_age = (unsigned int)-1;
1444         struct reiserfs_journal *journal = SB_JOURNAL(s);
1445         char *new_opts;
1446         int err;
1447         char *qf_names[REISERFS_MAXQUOTAS];
1448         unsigned int qfmt = 0;
1449 #ifdef CONFIG_QUOTA
1450         int i;
1451 #endif
1452
1453         new_opts = kstrdup(arg, GFP_KERNEL);
1454         if (arg && !new_opts)
1455                 return -ENOMEM;
1456
1457         sync_filesystem(s);
1458         reiserfs_write_lock(s);
1459
1460 #ifdef CONFIG_QUOTA
1461         memcpy(qf_names, REISERFS_SB(s)->s_qf_names, sizeof(qf_names));
1462 #endif
1463
1464         rs = SB_DISK_SUPER_BLOCK(s);
1465
1466         if (!reiserfs_parse_options
1467             (s, arg, &mount_options, &blocks, NULL, &commit_max_age,
1468             qf_names, &qfmt)) {
1469 #ifdef CONFIG_QUOTA
1470                 for (i = 0; i < REISERFS_MAXQUOTAS; i++)
1471                         if (qf_names[i] != REISERFS_SB(s)->s_qf_names[i])
1472                                 kfree(qf_names[i]);
1473 #endif
1474                 err = -EINVAL;
1475                 goto out_err_unlock;
1476         }
1477 #ifdef CONFIG_QUOTA
1478         handle_quota_files(s, qf_names, &qfmt);
1479 #endif
1480
1481         handle_attrs(s);
1482
1483         /* Add options that are safe here */
1484         safe_mask |= 1 << REISERFS_SMALLTAIL;
1485         safe_mask |= 1 << REISERFS_LARGETAIL;
1486         safe_mask |= 1 << REISERFS_NO_BORDER;
1487         safe_mask |= 1 << REISERFS_NO_UNHASHED_RELOCATION;
1488         safe_mask |= 1 << REISERFS_HASHED_RELOCATION;
1489         safe_mask |= 1 << REISERFS_TEST4;
1490         safe_mask |= 1 << REISERFS_ATTRS;
1491         safe_mask |= 1 << REISERFS_XATTRS_USER;
1492         safe_mask |= 1 << REISERFS_POSIXACL;
1493         safe_mask |= 1 << REISERFS_BARRIER_FLUSH;
1494         safe_mask |= 1 << REISERFS_BARRIER_NONE;
1495         safe_mask |= 1 << REISERFS_ERROR_RO;
1496         safe_mask |= 1 << REISERFS_ERROR_CONTINUE;
1497         safe_mask |= 1 << REISERFS_ERROR_PANIC;
1498         safe_mask |= 1 << REISERFS_USRQUOTA;
1499         safe_mask |= 1 << REISERFS_GRPQUOTA;
1500
1501         /*
1502          * Update the bitmask, taking care to keep
1503          * the bits we're not allowed to change here
1504          */
1505         REISERFS_SB(s)->s_mount_opt =
1506             (REISERFS_SB(s)->
1507              s_mount_opt & ~safe_mask) | (mount_options & safe_mask);
1508
1509         if (commit_max_age != 0 && commit_max_age != (unsigned int)-1) {
1510                 journal->j_max_commit_age = commit_max_age;
1511                 journal->j_max_trans_age = commit_max_age;
1512         } else if (commit_max_age == 0) {
1513                 /* 0 means restore defaults. */
1514                 journal->j_max_commit_age = journal->j_default_max_commit_age;
1515                 journal->j_max_trans_age = JOURNAL_MAX_TRANS_AGE;
1516         }
1517
1518         if (blocks) {
1519                 err = reiserfs_resize(s, blocks);
1520                 if (err != 0)
1521                         goto out_err_unlock;
1522         }
1523
1524         if (*mount_flags & MS_RDONLY) {
1525                 reiserfs_write_unlock(s);
1526                 reiserfs_xattr_init(s, *mount_flags);
1527                 /* remount read-only */
1528                 if (s->s_flags & MS_RDONLY)
1529                         /* it is read-only already */
1530                         goto out_ok_unlocked;
1531
1532                 err = dquot_suspend(s, -1);
1533                 if (err < 0)
1534                         goto out_err;
1535
1536                 /* try to remount file system with read-only permissions */
1537                 if (sb_umount_state(rs) == REISERFS_VALID_FS
1538                     || REISERFS_SB(s)->s_mount_state != REISERFS_VALID_FS) {
1539                         goto out_ok_unlocked;
1540                 }
1541
1542                 reiserfs_write_lock(s);
1543
1544                 err = journal_begin(&th, s, 10);
1545                 if (err)
1546                         goto out_err_unlock;
1547
1548                 /* Mounting a rw partition read-only. */
1549                 reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 1);
1550                 set_sb_umount_state(rs, REISERFS_SB(s)->s_mount_state);
1551                 journal_mark_dirty(&th, SB_BUFFER_WITH_SB(s));
1552         } else {
1553                 /* remount read-write */
1554                 if (!(s->s_flags & MS_RDONLY)) {
1555                         reiserfs_write_unlock(s);
1556                         reiserfs_xattr_init(s, *mount_flags);
1557                         goto out_ok_unlocked;   /* We are read-write already */
1558                 }
1559
1560                 if (reiserfs_is_journal_aborted(journal)) {
1561                         err = journal->j_errno;
1562                         goto out_err_unlock;
1563                 }
1564
1565                 handle_data_mode(s, mount_options);
1566                 handle_barrier_mode(s, mount_options);
1567                 REISERFS_SB(s)->s_mount_state = sb_umount_state(rs);
1568
1569                 /* now it is safe to call journal_begin */
1570                 s->s_flags &= ~MS_RDONLY;
1571                 err = journal_begin(&th, s, 10);
1572                 if (err)
1573                         goto out_err_unlock;
1574
1575                 /* Mount a partition which is read-only, read-write */
1576                 reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 1);
1577                 REISERFS_SB(s)->s_mount_state = sb_umount_state(rs);
1578                 s->s_flags &= ~MS_RDONLY;
1579                 set_sb_umount_state(rs, REISERFS_ERROR_FS);
1580                 if (!old_format_only(s))
1581                         set_sb_mnt_count(rs, sb_mnt_count(rs) + 1);
1582                 /* mark_buffer_dirty (SB_BUFFER_WITH_SB (s), 1); */
1583                 journal_mark_dirty(&th, SB_BUFFER_WITH_SB(s));
1584                 REISERFS_SB(s)->s_mount_state = REISERFS_VALID_FS;
1585         }
1586         /* this will force a full flush of all journal lists */
1587         SB_JOURNAL(s)->j_must_wait = 1;
1588         err = journal_end(&th);
1589         if (err)
1590                 goto out_err_unlock;
1591
1592         reiserfs_write_unlock(s);
1593         if (!(*mount_flags & MS_RDONLY)) {
1594                 dquot_resume(s, -1);
1595                 reiserfs_write_lock(s);
1596                 finish_unfinished(s);
1597                 reiserfs_write_unlock(s);
1598                 reiserfs_xattr_init(s, *mount_flags);
1599         }
1600
1601 out_ok_unlocked:
1602         if (new_opts)
1603                 replace_mount_options(s, new_opts);
1604         return 0;
1605
1606 out_err_unlock:
1607         reiserfs_write_unlock(s);
1608 out_err:
1609         kfree(new_opts);
1610         return err;
1611 }
1612
1613 static int read_super_block(struct super_block *s, int offset)
1614 {
1615         struct buffer_head *bh;
1616         struct reiserfs_super_block *rs;
1617         int fs_blocksize;
1618
1619         bh = sb_bread(s, offset / s->s_blocksize);
1620         if (!bh) {
1621                 reiserfs_warning(s, "sh-2006",
1622                                  "bread failed (dev %s, block %lu, size %lu)",
1623                                  s->s_id, offset / s->s_blocksize,
1624                                  s->s_blocksize);
1625                 return 1;
1626         }
1627
1628         rs = (struct reiserfs_super_block *)bh->b_data;
1629         if (!is_any_reiserfs_magic_string(rs)) {
1630                 brelse(bh);
1631                 return 1;
1632         }
1633         /*
1634          * ok, reiserfs signature (old or new) found in at the given offset
1635          */
1636         fs_blocksize = sb_blocksize(rs);
1637         brelse(bh);
1638         sb_set_blocksize(s, fs_blocksize);
1639
1640         bh = sb_bread(s, offset / s->s_blocksize);
1641         if (!bh) {
1642                 reiserfs_warning(s, "sh-2007",
1643                                  "bread failed (dev %s, block %lu, size %lu)",
1644                                  s->s_id, offset / s->s_blocksize,
1645                                  s->s_blocksize);
1646                 return 1;
1647         }
1648
1649         rs = (struct reiserfs_super_block *)bh->b_data;
1650         if (sb_blocksize(rs) != s->s_blocksize) {
1651                 reiserfs_warning(s, "sh-2011", "can't find a reiserfs "
1652                                  "filesystem on (dev %s, block %llu, size %lu)",
1653                                  s->s_id,
1654                                  (unsigned long long)bh->b_blocknr,
1655                                  s->s_blocksize);
1656                 brelse(bh);
1657                 return 1;
1658         }
1659
1660         if (rs->s_v1.s_root_block == cpu_to_le32(-1)) {
1661                 brelse(bh);
1662                 reiserfs_warning(s, "super-6519", "Unfinished reiserfsck "
1663                                  "--rebuild-tree run detected. Please run\n"
1664                                  "reiserfsck --rebuild-tree and wait for a "
1665                                  "completion. If that fails\n"
1666                                  "get newer reiserfsprogs package");
1667                 return 1;
1668         }
1669
1670         SB_BUFFER_WITH_SB(s) = bh;
1671         SB_DISK_SUPER_BLOCK(s) = rs;
1672
1673         /*
1674          * magic is of non-standard journal filesystem, look at s_version to
1675          * find which format is in use
1676          */
1677         if (is_reiserfs_jr(rs)) {
1678                 if (sb_version(rs) == REISERFS_VERSION_2)
1679                         reiserfs_info(s, "found reiserfs format \"3.6\""
1680                                       " with non-standard journal\n");
1681                 else if (sb_version(rs) == REISERFS_VERSION_1)
1682                         reiserfs_info(s, "found reiserfs format \"3.5\""
1683                                       " with non-standard journal\n");
1684                 else {
1685                         reiserfs_warning(s, "sh-2012", "found unknown "
1686                                          "format \"%u\" of reiserfs with "
1687                                          "non-standard magic", sb_version(rs));
1688                         return 1;
1689                 }
1690         } else
1691                 /*
1692                  * s_version of standard format may contain incorrect
1693                  * information, so we just look at the magic string
1694                  */
1695                 reiserfs_info(s,
1696                               "found reiserfs format \"%s\" with standard journal\n",
1697                               is_reiserfs_3_5(rs) ? "3.5" : "3.6");
1698
1699         s->s_op = &reiserfs_sops;
1700         s->s_export_op = &reiserfs_export_ops;
1701 #ifdef CONFIG_QUOTA
1702         s->s_qcop = &reiserfs_qctl_operations;
1703         s->dq_op = &reiserfs_quota_operations;
1704         s->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1705 #endif
1706
1707         /*
1708          * new format is limited by the 32 bit wide i_blocks field, want to
1709          * be one full block below that.
1710          */
1711         s->s_maxbytes = (512LL << 32) - s->s_blocksize;
1712         return 0;
1713 }
1714
1715 /* after journal replay, reread all bitmap and super blocks */
1716 static int reread_meta_blocks(struct super_block *s)
1717 {
1718         ll_rw_block(REQ_OP_READ, 0, 1, &SB_BUFFER_WITH_SB(s));
1719         wait_on_buffer(SB_BUFFER_WITH_SB(s));
1720         if (!buffer_uptodate(SB_BUFFER_WITH_SB(s))) {
1721                 reiserfs_warning(s, "reiserfs-2504", "error reading the super");
1722                 return 1;
1723         }
1724
1725         return 0;
1726 }
1727
1728 /* hash detection stuff */
1729
1730 /*
1731  * if root directory is empty - we set default - Yura's - hash and
1732  * warn about it
1733  * FIXME: we look for only one name in a directory. If tea and yura
1734  * both have the same value - we ask user to send report to the
1735  * mailing list
1736  */
1737 static __u32 find_hash_out(struct super_block *s)
1738 {
1739         int retval;
1740         struct inode *inode;
1741         struct cpu_key key;
1742         INITIALIZE_PATH(path);
1743         struct reiserfs_dir_entry de;
1744         struct reiserfs_de_head *deh;
1745         __u32 hash = DEFAULT_HASH;
1746         __u32 deh_hashval, teahash, r5hash, yurahash;
1747
1748         inode = d_inode(s->s_root);
1749
1750         make_cpu_key(&key, inode, ~0, TYPE_DIRENTRY, 3);
1751         retval = search_by_entry_key(s, &key, &path, &de);
1752         if (retval == IO_ERROR) {
1753                 pathrelse(&path);
1754                 return UNSET_HASH;
1755         }
1756         if (retval == NAME_NOT_FOUND)
1757                 de.de_entry_num--;
1758
1759         set_de_name_and_namelen(&de);
1760         deh = de.de_deh + de.de_entry_num;
1761
1762         if (deh_offset(deh) == DOT_DOT_OFFSET) {
1763                 /* allow override in this case */
1764                 if (reiserfs_rupasov_hash(s))
1765                         hash = YURA_HASH;
1766                 reiserfs_info(s, "FS seems to be empty, autodetect is using the default hash\n");
1767                 goto out;
1768         }
1769
1770         deh_hashval = GET_HASH_VALUE(deh_offset(deh));
1771         r5hash = GET_HASH_VALUE(r5_hash(de.de_name, de.de_namelen));
1772         teahash = GET_HASH_VALUE(keyed_hash(de.de_name, de.de_namelen));
1773         yurahash = GET_HASH_VALUE(yura_hash(de.de_name, de.de_namelen));
1774
1775         if ((teahash == r5hash && deh_hashval == r5hash) ||
1776             (teahash == yurahash && deh_hashval == yurahash) ||
1777             (r5hash == yurahash && deh_hashval == yurahash)) {
1778                 reiserfs_warning(s, "reiserfs-2506",
1779                                  "Unable to automatically detect hash "
1780                                  "function. Please mount with -o "
1781                                  "hash={tea,rupasov,r5}");
1782                 hash = UNSET_HASH;
1783                 goto out;
1784         }
1785
1786         if (deh_hashval == yurahash)
1787                 hash = YURA_HASH;
1788         else if (deh_hashval == teahash)
1789                 hash = TEA_HASH;
1790         else if (deh_hashval == r5hash)
1791                 hash = R5_HASH;
1792         else {
1793                 reiserfs_warning(s, "reiserfs-2506",
1794                                  "Unrecognised hash function");
1795                 hash = UNSET_HASH;
1796         }
1797 out:
1798         pathrelse(&path);
1799         return hash;
1800 }
1801
1802 /* finds out which hash names are sorted with */
1803 static int what_hash(struct super_block *s)
1804 {
1805         __u32 code;
1806
1807         code = sb_hash_function_code(SB_DISK_SUPER_BLOCK(s));
1808
1809         /*
1810          * reiserfs_hash_detect() == true if any of the hash mount options
1811          * were used.  We must check them to make sure the user isn't
1812          * using a bad hash value
1813          */
1814         if (code == UNSET_HASH || reiserfs_hash_detect(s))
1815                 code = find_hash_out(s);
1816
1817         if (code != UNSET_HASH && reiserfs_hash_detect(s)) {
1818                 /*
1819                  * detection has found the hash, and we must check against the
1820                  * mount options
1821                  */
1822                 if (reiserfs_rupasov_hash(s) && code != YURA_HASH) {
1823                         reiserfs_warning(s, "reiserfs-2507",
1824                                          "Error, %s hash detected, "
1825                                          "unable to force rupasov hash",
1826                                          reiserfs_hashname(code));
1827                         code = UNSET_HASH;
1828                 } else if (reiserfs_tea_hash(s) && code != TEA_HASH) {
1829                         reiserfs_warning(s, "reiserfs-2508",
1830                                          "Error, %s hash detected, "
1831                                          "unable to force tea hash",
1832                                          reiserfs_hashname(code));
1833                         code = UNSET_HASH;
1834                 } else if (reiserfs_r5_hash(s) && code != R5_HASH) {
1835                         reiserfs_warning(s, "reiserfs-2509",
1836                                          "Error, %s hash detected, "
1837                                          "unable to force r5 hash",
1838                                          reiserfs_hashname(code));
1839                         code = UNSET_HASH;
1840                 }
1841         } else {
1842                 /*
1843                  * find_hash_out was not called or
1844                  * could not determine the hash
1845                  */
1846                 if (reiserfs_rupasov_hash(s)) {
1847                         code = YURA_HASH;
1848                 } else if (reiserfs_tea_hash(s)) {
1849                         code = TEA_HASH;
1850                 } else if (reiserfs_r5_hash(s)) {
1851                         code = R5_HASH;
1852                 }
1853         }
1854
1855         /*
1856          * if we are mounted RW, and we have a new valid hash code, update
1857          * the super
1858          */
1859         if (code != UNSET_HASH &&
1860             !(s->s_flags & MS_RDONLY) &&
1861             code != sb_hash_function_code(SB_DISK_SUPER_BLOCK(s))) {
1862                 set_sb_hash_function_code(SB_DISK_SUPER_BLOCK(s), code);
1863         }
1864         return code;
1865 }
1866
1867 /* return pointer to appropriate function */
1868 static hashf_t hash_function(struct super_block *s)
1869 {
1870         switch (what_hash(s)) {
1871         case TEA_HASH:
1872                 reiserfs_info(s, "Using tea hash to sort names\n");
1873                 return keyed_hash;
1874         case YURA_HASH:
1875                 reiserfs_info(s, "Using rupasov hash to sort names\n");
1876                 return yura_hash;
1877         case R5_HASH:
1878                 reiserfs_info(s, "Using r5 hash to sort names\n");
1879                 return r5_hash;
1880         }
1881         return NULL;
1882 }
1883
1884 /* this is used to set up correct value for old partitions */
1885 static int function2code(hashf_t func)
1886 {
1887         if (func == keyed_hash)
1888                 return TEA_HASH;
1889         if (func == yura_hash)
1890                 return YURA_HASH;
1891         if (func == r5_hash)
1892                 return R5_HASH;
1893
1894         BUG();                  /* should never happen */
1895
1896         return 0;
1897 }
1898
1899 #define SWARN(silent, s, id, ...)                       \
1900         if (!(silent))                          \
1901                 reiserfs_warning(s, id, __VA_ARGS__)
1902
1903 static int reiserfs_fill_super(struct super_block *s, void *data, int silent)
1904 {
1905         struct inode *root_inode;
1906         struct reiserfs_transaction_handle th;
1907         int old_format = 0;
1908         unsigned long blocks;
1909         unsigned int commit_max_age = 0;
1910         int jinit_done = 0;
1911         struct reiserfs_iget_args args;
1912         struct reiserfs_super_block *rs;
1913         char *jdev_name;
1914         struct reiserfs_sb_info *sbi;
1915         int errval = -EINVAL;
1916         char *qf_names[REISERFS_MAXQUOTAS] = {};
1917         unsigned int qfmt = 0;
1918
1919         save_mount_options(s, data);
1920
1921         sbi = kzalloc(sizeof(struct reiserfs_sb_info), GFP_KERNEL);
1922         if (!sbi)
1923                 return -ENOMEM;
1924         s->s_fs_info = sbi;
1925         /* Set default values for options: non-aggressive tails, RO on errors */
1926         sbi->s_mount_opt |= (1 << REISERFS_SMALLTAIL);
1927         sbi->s_mount_opt |= (1 << REISERFS_ERROR_RO);
1928         sbi->s_mount_opt |= (1 << REISERFS_BARRIER_FLUSH);
1929         /* no preallocation minimum, be smart in reiserfs_file_write instead */
1930         sbi->s_alloc_options.preallocmin = 0;
1931         /* Preallocate by 16 blocks (17-1) at once */
1932         sbi->s_alloc_options.preallocsize = 17;
1933         /* setup default block allocator options */
1934         reiserfs_init_alloc_options(s);
1935
1936         spin_lock_init(&sbi->old_work_lock);
1937         INIT_DELAYED_WORK(&sbi->old_work, flush_old_commits);
1938         mutex_init(&sbi->lock);
1939         sbi->lock_depth = -1;
1940
1941         sbi->commit_wq = alloc_workqueue("reiserfs/%s", WQ_MEM_RECLAIM, 0,
1942                                          s->s_id);
1943         if (!sbi->commit_wq) {
1944                 SWARN(silent, s, "", "Cannot allocate commit workqueue");
1945                 errval = -ENOMEM;
1946                 goto error_unlocked;
1947         }
1948
1949         jdev_name = NULL;
1950         if (reiserfs_parse_options
1951             (s, (char *)data, &sbi->s_mount_opt, &blocks, &jdev_name,
1952              &commit_max_age, qf_names, &qfmt) == 0) {
1953                 goto error_unlocked;
1954         }
1955         if (jdev_name && jdev_name[0]) {
1956                 sbi->s_jdev = kstrdup(jdev_name, GFP_KERNEL);
1957                 if (!sbi->s_jdev) {
1958                         SWARN(silent, s, "", "Cannot allocate memory for "
1959                                 "journal device name");
1960                         goto error;
1961                 }
1962         }
1963 #ifdef CONFIG_QUOTA
1964         handle_quota_files(s, qf_names, &qfmt);
1965 #endif
1966
1967         if (blocks) {
1968                 SWARN(silent, s, "jmacd-7", "resize option for remount only");
1969                 goto error_unlocked;
1970         }
1971
1972         /*
1973          * try old format (undistributed bitmap, super block in 8-th 1k
1974          * block of a device)
1975          */
1976         if (!read_super_block(s, REISERFS_OLD_DISK_OFFSET_IN_BYTES))
1977                 old_format = 1;
1978
1979         /*
1980          * try new format (64-th 1k block), which can contain reiserfs
1981          * super block
1982          */
1983         else if (read_super_block(s, REISERFS_DISK_OFFSET_IN_BYTES)) {
1984                 SWARN(silent, s, "sh-2021", "can not find reiserfs on %s",
1985                       s->s_id);
1986                 goto error_unlocked;
1987         }
1988
1989         rs = SB_DISK_SUPER_BLOCK(s);
1990         /*
1991          * Let's do basic sanity check to verify that underlying device is not
1992          * smaller than the filesystem. If the check fails then abort and
1993          * scream, because bad stuff will happen otherwise.
1994          */
1995         if (s->s_bdev && s->s_bdev->bd_inode
1996             && i_size_read(s->s_bdev->bd_inode) <
1997             sb_block_count(rs) * sb_blocksize(rs)) {
1998                 SWARN(silent, s, "", "Filesystem cannot be "
1999                       "mounted because it is bigger than the device");
2000                 SWARN(silent, s, "", "You may need to run fsck "
2001                       "or increase size of your LVM partition");
2002                 SWARN(silent, s, "", "Or may be you forgot to "
2003                       "reboot after fdisk when it told you to");
2004                 goto error_unlocked;
2005         }
2006
2007         sbi->s_mount_state = SB_REISERFS_STATE(s);
2008         sbi->s_mount_state = REISERFS_VALID_FS;
2009
2010         if ((errval = reiserfs_init_bitmap_cache(s))) {
2011                 SWARN(silent, s, "jmacd-8", "unable to read bitmap");
2012                 goto error_unlocked;
2013         }
2014
2015         errval = -EINVAL;
2016 #ifdef CONFIG_REISERFS_CHECK
2017         SWARN(silent, s, "", "CONFIG_REISERFS_CHECK is set ON");
2018         SWARN(silent, s, "", "- it is slow mode for debugging.");
2019 #endif
2020
2021         /* make data=ordered the default */
2022         if (!reiserfs_data_log(s) && !reiserfs_data_ordered(s) &&
2023             !reiserfs_data_writeback(s)) {
2024                 sbi->s_mount_opt |= (1 << REISERFS_DATA_ORDERED);
2025         }
2026
2027         if (reiserfs_data_log(s)) {
2028                 reiserfs_info(s, "using journaled data mode\n");
2029         } else if (reiserfs_data_ordered(s)) {
2030                 reiserfs_info(s, "using ordered data mode\n");
2031         } else {
2032                 reiserfs_info(s, "using writeback data mode\n");
2033         }
2034         if (reiserfs_barrier_flush(s)) {
2035                 printk("reiserfs: using flush barriers\n");
2036         }
2037
2038         if (journal_init(s, jdev_name, old_format, commit_max_age)) {
2039                 SWARN(silent, s, "sh-2022",
2040                       "unable to initialize journal space");
2041                 goto error_unlocked;
2042         } else {
2043                 /*
2044                  * once this is set, journal_release must be called
2045                  * if we error out of the mount
2046                  */
2047                 jinit_done = 1;
2048         }
2049
2050         if (reread_meta_blocks(s)) {
2051                 SWARN(silent, s, "jmacd-9",
2052                       "unable to reread meta blocks after journal init");
2053                 goto error_unlocked;
2054         }
2055
2056         if (replay_only(s))
2057                 goto error_unlocked;
2058
2059         if (bdev_read_only(s->s_bdev) && !(s->s_flags & MS_RDONLY)) {
2060                 SWARN(silent, s, "clm-7000",
2061                       "Detected readonly device, marking FS readonly");
2062                 s->s_flags |= MS_RDONLY;
2063         }
2064         args.objectid = REISERFS_ROOT_OBJECTID;
2065         args.dirid = REISERFS_ROOT_PARENT_OBJECTID;
2066         root_inode =
2067             iget5_locked(s, REISERFS_ROOT_OBJECTID, reiserfs_find_actor,
2068                          reiserfs_init_locked_inode, (void *)&args);
2069         if (!root_inode) {
2070                 SWARN(silent, s, "jmacd-10", "get root inode failed");
2071                 goto error_unlocked;
2072         }
2073
2074         /*
2075          * This path assumed to be called with the BKL in the old times.
2076          * Now we have inherited the big reiserfs lock from it and many
2077          * reiserfs helpers called in the mount path and elsewhere require
2078          * this lock to be held even if it's not always necessary. Let's be
2079          * conservative and hold it early. The window can be reduced after
2080          * careful review of the code.
2081          */
2082         reiserfs_write_lock(s);
2083
2084         if (root_inode->i_state & I_NEW) {
2085                 reiserfs_read_locked_inode(root_inode, &args);
2086                 unlock_new_inode(root_inode);
2087         }
2088
2089         s->s_root = d_make_root(root_inode);
2090         if (!s->s_root)
2091                 goto error;
2092         /* define and initialize hash function */
2093         sbi->s_hash_function = hash_function(s);
2094         if (sbi->s_hash_function == NULL) {
2095                 dput(s->s_root);
2096                 s->s_root = NULL;
2097                 goto error;
2098         }
2099
2100         if (is_reiserfs_3_5(rs)
2101             || (is_reiserfs_jr(rs) && SB_VERSION(s) == REISERFS_VERSION_1))
2102                 set_bit(REISERFS_3_5, &sbi->s_properties);
2103         else if (old_format)
2104                 set_bit(REISERFS_OLD_FORMAT, &sbi->s_properties);
2105         else
2106                 set_bit(REISERFS_3_6, &sbi->s_properties);
2107
2108         if (!(s->s_flags & MS_RDONLY)) {
2109
2110                 errval = journal_begin(&th, s, 1);
2111                 if (errval) {
2112                         dput(s->s_root);
2113                         s->s_root = NULL;
2114                         goto error;
2115                 }
2116                 reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 1);
2117
2118                 set_sb_umount_state(rs, REISERFS_ERROR_FS);
2119                 set_sb_fs_state(rs, 0);
2120
2121                 /*
2122                  * Clear out s_bmap_nr if it would wrap. We can handle this
2123                  * case, but older revisions can't. This will cause the
2124                  * file system to fail mount on those older implementations,
2125                  * avoiding corruption. -jeffm
2126                  */
2127                 if (bmap_would_wrap(reiserfs_bmap_count(s)) &&
2128                     sb_bmap_nr(rs) != 0) {
2129                         reiserfs_warning(s, "super-2030", "This file system "
2130                                         "claims to use %u bitmap blocks in "
2131                                         "its super block, but requires %u. "
2132                                         "Clearing to zero.", sb_bmap_nr(rs),
2133                                         reiserfs_bmap_count(s));
2134
2135                         set_sb_bmap_nr(rs, 0);
2136                 }
2137
2138                 if (old_format_only(s)) {
2139                         /*
2140                          * filesystem of format 3.5 either with standard
2141                          * or non-standard journal
2142                          */
2143                         if (convert_reiserfs(s)) {
2144                                 /* and -o conv is given */
2145                                 if (!silent)
2146                                         reiserfs_info(s,
2147                                                       "converting 3.5 filesystem to the 3.6 format");
2148
2149                                 if (is_reiserfs_3_5(rs))
2150                                         /*
2151                                          * put magic string of 3.6 format.
2152                                          * 2.2 will not be able to
2153                                          * mount this filesystem anymore
2154                                          */
2155                                         memcpy(rs->s_v1.s_magic,
2156                                                reiserfs_3_6_magic_string,
2157                                                sizeof
2158                                                (reiserfs_3_6_magic_string));
2159
2160                                 set_sb_version(rs, REISERFS_VERSION_2);
2161                                 reiserfs_convert_objectid_map_v1(s);
2162                                 set_bit(REISERFS_3_6, &sbi->s_properties);
2163                                 clear_bit(REISERFS_3_5, &sbi->s_properties);
2164                         } else if (!silent) {
2165                                 reiserfs_info(s, "using 3.5.x disk format\n");
2166                         }
2167                 } else
2168                         set_sb_mnt_count(rs, sb_mnt_count(rs) + 1);
2169
2170
2171                 journal_mark_dirty(&th, SB_BUFFER_WITH_SB(s));
2172                 errval = journal_end(&th);
2173                 if (errval) {
2174                         dput(s->s_root);
2175                         s->s_root = NULL;
2176                         goto error;
2177                 }
2178
2179                 reiserfs_write_unlock(s);
2180                 if ((errval = reiserfs_lookup_privroot(s)) ||
2181                     (errval = reiserfs_xattr_init(s, s->s_flags))) {
2182                         dput(s->s_root);
2183                         s->s_root = NULL;
2184                         goto error_unlocked;
2185                 }
2186                 reiserfs_write_lock(s);
2187
2188                 /*
2189                  * look for files which were to be removed in previous session
2190                  */
2191                 finish_unfinished(s);
2192         } else {
2193                 if (old_format_only(s) && !silent) {
2194                         reiserfs_info(s, "using 3.5.x disk format\n");
2195                 }
2196
2197                 reiserfs_write_unlock(s);
2198                 if ((errval = reiserfs_lookup_privroot(s)) ||
2199                     (errval = reiserfs_xattr_init(s, s->s_flags))) {
2200                         dput(s->s_root);
2201                         s->s_root = NULL;
2202                         goto error_unlocked;
2203                 }
2204                 reiserfs_write_lock(s);
2205         }
2206         /*
2207          * mark hash in super block: it could be unset. overwrite should be ok
2208          */
2209         set_sb_hash_function_code(rs, function2code(sbi->s_hash_function));
2210
2211         handle_attrs(s);
2212
2213         reiserfs_proc_info_init(s);
2214
2215         init_waitqueue_head(&(sbi->s_wait));
2216         spin_lock_init(&sbi->bitmap_lock);
2217
2218         reiserfs_write_unlock(s);
2219
2220         return (0);
2221
2222 error:
2223         reiserfs_write_unlock(s);
2224
2225 error_unlocked:
2226         /* kill the commit thread, free journal ram */
2227         if (jinit_done) {
2228                 reiserfs_write_lock(s);
2229                 journal_release_error(NULL, s);
2230                 reiserfs_write_unlock(s);
2231         }
2232
2233         if (sbi->commit_wq)
2234                 destroy_workqueue(sbi->commit_wq);
2235
2236         reiserfs_cancel_old_flush(s);
2237
2238         reiserfs_free_bitmap_cache(s);
2239         if (SB_BUFFER_WITH_SB(s))
2240                 brelse(SB_BUFFER_WITH_SB(s));
2241 #ifdef CONFIG_QUOTA
2242         {
2243                 int j;
2244                 for (j = 0; j < REISERFS_MAXQUOTAS; j++)
2245                         kfree(qf_names[j]);
2246         }
2247 #endif
2248         kfree(sbi);
2249
2250         s->s_fs_info = NULL;
2251         return errval;
2252 }
2253
2254 static int reiserfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2255 {
2256         struct reiserfs_super_block *rs = SB_DISK_SUPER_BLOCK(dentry->d_sb);
2257
2258         buf->f_namelen = (REISERFS_MAX_NAME(s->s_blocksize));
2259         buf->f_bfree = sb_free_blocks(rs);
2260         buf->f_bavail = buf->f_bfree;
2261         buf->f_blocks = sb_block_count(rs) - sb_bmap_nr(rs) - 1;
2262         buf->f_bsize = dentry->d_sb->s_blocksize;
2263         /* changed to accommodate gcc folks. */
2264         buf->f_type = REISERFS_SUPER_MAGIC;
2265         buf->f_fsid.val[0] = (u32)crc32_le(0, rs->s_uuid, sizeof(rs->s_uuid)/2);
2266         buf->f_fsid.val[1] = (u32)crc32_le(0, rs->s_uuid + sizeof(rs->s_uuid)/2,
2267                                 sizeof(rs->s_uuid)/2);
2268
2269         return 0;
2270 }
2271
2272 #ifdef CONFIG_QUOTA
2273 static int reiserfs_write_dquot(struct dquot *dquot)
2274 {
2275         struct reiserfs_transaction_handle th;
2276         int ret, err;
2277         int depth;
2278
2279         reiserfs_write_lock(dquot->dq_sb);
2280         ret =
2281             journal_begin(&th, dquot->dq_sb,
2282                           REISERFS_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2283         if (ret)
2284                 goto out;
2285         depth = reiserfs_write_unlock_nested(dquot->dq_sb);
2286         ret = dquot_commit(dquot);
2287         reiserfs_write_lock_nested(dquot->dq_sb, depth);
2288         err = journal_end(&th);
2289         if (!ret && err)
2290                 ret = err;
2291 out:
2292         reiserfs_write_unlock(dquot->dq_sb);
2293         return ret;
2294 }
2295
2296 static int reiserfs_acquire_dquot(struct dquot *dquot)
2297 {
2298         struct reiserfs_transaction_handle th;
2299         int ret, err;
2300         int depth;
2301
2302         reiserfs_write_lock(dquot->dq_sb);
2303         ret =
2304             journal_begin(&th, dquot->dq_sb,
2305                           REISERFS_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2306         if (ret)
2307                 goto out;
2308         depth = reiserfs_write_unlock_nested(dquot->dq_sb);
2309         ret = dquot_acquire(dquot);
2310         reiserfs_write_lock_nested(dquot->dq_sb, depth);
2311         err = journal_end(&th);
2312         if (!ret && err)
2313                 ret = err;
2314 out:
2315         reiserfs_write_unlock(dquot->dq_sb);
2316         return ret;
2317 }
2318
2319 static int reiserfs_release_dquot(struct dquot *dquot)
2320 {
2321         struct reiserfs_transaction_handle th;
2322         int ret, err;
2323
2324         reiserfs_write_lock(dquot->dq_sb);
2325         ret =
2326             journal_begin(&th, dquot->dq_sb,
2327                           REISERFS_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2328         reiserfs_write_unlock(dquot->dq_sb);
2329         if (ret) {
2330                 /* Release dquot anyway to avoid endless cycle in dqput() */
2331                 dquot_release(dquot);
2332                 goto out;
2333         }
2334         ret = dquot_release(dquot);
2335         reiserfs_write_lock(dquot->dq_sb);
2336         err = journal_end(&th);
2337         if (!ret && err)
2338                 ret = err;
2339         reiserfs_write_unlock(dquot->dq_sb);
2340 out:
2341         return ret;
2342 }
2343
2344 static int reiserfs_mark_dquot_dirty(struct dquot *dquot)
2345 {
2346         /* Are we journaling quotas? */
2347         if (REISERFS_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2348             REISERFS_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2349                 dquot_mark_dquot_dirty(dquot);
2350                 return reiserfs_write_dquot(dquot);
2351         } else
2352                 return dquot_mark_dquot_dirty(dquot);
2353 }
2354
2355 static int reiserfs_write_info(struct super_block *sb, int type)
2356 {
2357         struct reiserfs_transaction_handle th;
2358         int ret, err;
2359         int depth;
2360
2361         /* Data block + inode block */
2362         reiserfs_write_lock(sb);
2363         ret = journal_begin(&th, sb, 2);
2364         if (ret)
2365                 goto out;
2366         depth = reiserfs_write_unlock_nested(sb);
2367         ret = dquot_commit_info(sb, type);
2368         reiserfs_write_lock_nested(sb, depth);
2369         err = journal_end(&th);
2370         if (!ret && err)
2371                 ret = err;
2372 out:
2373         reiserfs_write_unlock(sb);
2374         return ret;
2375 }
2376
2377 /*
2378  * Turn on quotas during mount time - we need to find the quota file and such...
2379  */
2380 static int reiserfs_quota_on_mount(struct super_block *sb, int type)
2381 {
2382         return dquot_quota_on_mount(sb, REISERFS_SB(sb)->s_qf_names[type],
2383                                         REISERFS_SB(sb)->s_jquota_fmt, type);
2384 }
2385
2386 /*
2387  * Standard function to be called on quota_on
2388  */
2389 static int reiserfs_quota_on(struct super_block *sb, int type, int format_id,
2390                              const struct path *path)
2391 {
2392         int err;
2393         struct inode *inode;
2394         struct reiserfs_transaction_handle th;
2395         int opt = type == USRQUOTA ? REISERFS_USRQUOTA : REISERFS_GRPQUOTA;
2396
2397         reiserfs_write_lock(sb);
2398         if (!(REISERFS_SB(sb)->s_mount_opt & (1 << opt))) {
2399                 err = -EINVAL;
2400                 goto out;
2401         }
2402
2403         /* Quotafile not on the same filesystem? */
2404         if (path->dentry->d_sb != sb) {
2405                 err = -EXDEV;
2406                 goto out;
2407         }
2408         inode = d_inode(path->dentry);
2409         /*
2410          * We must not pack tails for quota files on reiserfs for quota
2411          * IO to work
2412          */
2413         if (!(REISERFS_I(inode)->i_flags & i_nopack_mask)) {
2414                 err = reiserfs_unpack(inode, NULL);
2415                 if (err) {
2416                         reiserfs_warning(sb, "super-6520",
2417                                 "Unpacking tail of quota file failed"
2418                                 " (%d). Cannot turn on quotas.", err);
2419                         err = -EINVAL;
2420                         goto out;
2421                 }
2422                 mark_inode_dirty(inode);
2423         }
2424         /* Journaling quota? */
2425         if (REISERFS_SB(sb)->s_qf_names[type]) {
2426                 /* Quotafile not of fs root? */
2427                 if (path->dentry->d_parent != sb->s_root)
2428                         reiserfs_warning(sb, "super-6521",
2429                                  "Quota file not on filesystem root. "
2430                                  "Journalled quota will not work.");
2431         }
2432
2433         /*
2434          * When we journal data on quota file, we have to flush journal to see
2435          * all updates to the file when we bypass pagecache...
2436          */
2437         if (reiserfs_file_data_log(inode)) {
2438                 /* Just start temporary transaction and finish it */
2439                 err = journal_begin(&th, sb, 1);
2440                 if (err)
2441                         goto out;
2442                 err = journal_end_sync(&th);
2443                 if (err)
2444                         goto out;
2445         }
2446         reiserfs_write_unlock(sb);
2447         err = dquot_quota_on(sb, type, format_id, path);
2448         if (!err) {
2449                 inode_lock(inode);
2450                 REISERFS_I(inode)->i_attrs |= REISERFS_IMMUTABLE_FL |
2451                                               REISERFS_NOATIME_FL;
2452                 inode_set_flags(inode, S_IMMUTABLE | S_NOATIME,
2453                                 S_IMMUTABLE | S_NOATIME);
2454                 inode_unlock(inode);
2455                 mark_inode_dirty(inode);
2456         }
2457         return err;
2458 out:
2459         reiserfs_write_unlock(sb);
2460         return err;
2461 }
2462
2463 static int reiserfs_quota_off(struct super_block *sb, int type)
2464 {
2465         int err;
2466         struct inode *inode = sb_dqopt(sb)->files[type];
2467
2468         if (!inode || !igrab(inode))
2469                 goto out;
2470
2471         err = dquot_quota_off(sb, type);
2472         if (err)
2473                 goto out_put;
2474
2475         inode_lock(inode);
2476         REISERFS_I(inode)->i_attrs &= ~(REISERFS_IMMUTABLE_FL |
2477                                         REISERFS_NOATIME_FL);
2478         inode_set_flags(inode, 0, S_IMMUTABLE | S_NOATIME);
2479         inode_unlock(inode);
2480         mark_inode_dirty(inode);
2481 out_put:
2482         iput(inode);
2483         return err;
2484 out:
2485         return dquot_quota_off(sb, type);
2486 }
2487
2488 /*
2489  * Read data from quotafile - avoid pagecache and such because we cannot afford
2490  * acquiring the locks... As quota files are never truncated and quota code
2491  * itself serializes the operations (and no one else should touch the files)
2492  * we don't have to be afraid of races
2493  */
2494 static ssize_t reiserfs_quota_read(struct super_block *sb, int type, char *data,
2495                                    size_t len, loff_t off)
2496 {
2497         struct inode *inode = sb_dqopt(sb)->files[type];
2498         unsigned long blk = off >> sb->s_blocksize_bits;
2499         int err = 0, offset = off & (sb->s_blocksize - 1), tocopy;
2500         size_t toread;
2501         struct buffer_head tmp_bh, *bh;
2502         loff_t i_size = i_size_read(inode);
2503
2504         if (off > i_size)
2505                 return 0;
2506         if (off + len > i_size)
2507                 len = i_size - off;
2508         toread = len;
2509         while (toread > 0) {
2510                 tocopy =
2511                     sb->s_blocksize - offset <
2512                     toread ? sb->s_blocksize - offset : toread;
2513                 tmp_bh.b_state = 0;
2514                 /*
2515                  * Quota files are without tails so we can safely
2516                  * use this function
2517                  */
2518                 reiserfs_write_lock(sb);
2519                 err = reiserfs_get_block(inode, blk, &tmp_bh, 0);
2520                 reiserfs_write_unlock(sb);
2521                 if (err)
2522                         return err;
2523                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
2524                         memset(data, 0, tocopy);
2525                 else {
2526                         bh = sb_bread(sb, tmp_bh.b_blocknr);
2527                         if (!bh)
2528                                 return -EIO;
2529                         memcpy(data, bh->b_data + offset, tocopy);
2530                         brelse(bh);
2531                 }
2532                 offset = 0;
2533                 toread -= tocopy;
2534                 data += tocopy;
2535                 blk++;
2536         }
2537         return len;
2538 }
2539
2540 /*
2541  * Write to quotafile (we know the transaction is already started and has
2542  * enough credits)
2543  */
2544 static ssize_t reiserfs_quota_write(struct super_block *sb, int type,
2545                                     const char *data, size_t len, loff_t off)
2546 {
2547         struct inode *inode = sb_dqopt(sb)->files[type];
2548         unsigned long blk = off >> sb->s_blocksize_bits;
2549         int err = 0, offset = off & (sb->s_blocksize - 1), tocopy;
2550         int journal_quota = REISERFS_SB(sb)->s_qf_names[type] != NULL;
2551         size_t towrite = len;
2552         struct buffer_head tmp_bh, *bh;
2553
2554         if (!current->journal_info) {
2555                 printk(KERN_WARNING "reiserfs: Quota write (off=%llu, len=%llu) cancelled because transaction is not started.\n",
2556                         (unsigned long long)off, (unsigned long long)len);
2557                 return -EIO;
2558         }
2559         while (towrite > 0) {
2560                 tocopy = sb->s_blocksize - offset < towrite ?
2561                     sb->s_blocksize - offset : towrite;
2562                 tmp_bh.b_state = 0;
2563                 reiserfs_write_lock(sb);
2564                 err = reiserfs_get_block(inode, blk, &tmp_bh, GET_BLOCK_CREATE);
2565                 reiserfs_write_unlock(sb);
2566                 if (err)
2567                         goto out;
2568                 if (offset || tocopy != sb->s_blocksize)
2569                         bh = sb_bread(sb, tmp_bh.b_blocknr);
2570                 else
2571                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
2572                 if (!bh) {
2573                         err = -EIO;
2574                         goto out;
2575                 }
2576                 lock_buffer(bh);
2577                 memcpy(bh->b_data + offset, data, tocopy);
2578                 flush_dcache_page(bh->b_page);
2579                 set_buffer_uptodate(bh);
2580                 unlock_buffer(bh);
2581                 reiserfs_write_lock(sb);
2582                 reiserfs_prepare_for_journal(sb, bh, 1);
2583                 journal_mark_dirty(current->journal_info, bh);
2584                 if (!journal_quota)
2585                         reiserfs_add_ordered_list(inode, bh);
2586                 reiserfs_write_unlock(sb);
2587                 brelse(bh);
2588                 offset = 0;
2589                 towrite -= tocopy;
2590                 data += tocopy;
2591                 blk++;
2592         }
2593 out:
2594         if (len == towrite)
2595                 return err;
2596         if (inode->i_size < off + len - towrite)
2597                 i_size_write(inode, off + len - towrite);
2598         inode->i_version++;
2599         inode->i_mtime = inode->i_ctime = current_time(inode);
2600         mark_inode_dirty(inode);
2601         return len - towrite;
2602 }
2603
2604 #endif
2605
2606 static struct dentry *get_super_block(struct file_system_type *fs_type,
2607                            int flags, const char *dev_name,
2608                            void *data)
2609 {
2610         return mount_bdev(fs_type, flags, dev_name, data, reiserfs_fill_super);
2611 }
2612
2613 static int __init init_reiserfs_fs(void)
2614 {
2615         int ret;
2616
2617         ret = init_inodecache();
2618         if (ret)
2619                 return ret;
2620
2621         reiserfs_proc_info_global_init();
2622
2623         ret = register_filesystem(&reiserfs_fs_type);
2624         if (ret)
2625                 goto out;
2626
2627         return 0;
2628 out:
2629         reiserfs_proc_info_global_done();
2630         destroy_inodecache();
2631
2632         return ret;
2633 }
2634
2635 static void __exit exit_reiserfs_fs(void)
2636 {
2637         reiserfs_proc_info_global_done();
2638         unregister_filesystem(&reiserfs_fs_type);
2639         destroy_inodecache();
2640 }
2641
2642 struct file_system_type reiserfs_fs_type = {
2643         .owner = THIS_MODULE,
2644         .name = "reiserfs",
2645         .mount = get_super_block,
2646         .kill_sb = reiserfs_kill_sb,
2647         .fs_flags = FS_REQUIRES_DEV,
2648 };
2649 MODULE_ALIAS_FS("reiserfs");
2650
2651 MODULE_DESCRIPTION("ReiserFS journaled filesystem");
2652 MODULE_AUTHOR("Hans Reiser <reiser@namesys.com>");
2653 MODULE_LICENSE("GPL");
2654
2655 module_init(init_reiserfs_fs);
2656 module_exit(exit_reiserfs_fs);