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