super.c revision af591ad896ef75585752ac2eab4fba9437f23322
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 "reiserfs.h"
20#include "acl.h"
21#include "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
33struct file_system_type reiserfs_fs_type;
34
35static const char reiserfs_3_5_magic_string[] = REISERFS_SUPER_MAGIC_STRING;
36static const char reiserfs_3_6_magic_string[] = REISER2FS_SUPER_MAGIC_STRING;
37static const char reiserfs_jr_magic_string[] = REISER2FS_JR_SUPER_MAGIC_STRING;
38
39int 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
45int 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
51int 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
57static 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
63static int reiserfs_remount(struct super_block *s, int *flags, char *data);
64static int reiserfs_statfs(struct dentry *dentry, struct kstatfs *buf);
65void show_alloc_options(struct seq_file *seq, struct super_block *s);
66
67static int reiserfs_sync_fs(struct super_block *s, int wait)
68{
69	struct reiserfs_transaction_handle th;
70
71	/*
72	 * Writeback quota in non-journalled quota case - journalled quota has
73	 * no dirty dquots
74	 */
75	dquot_writeback_dquots(s, -1);
76	reiserfs_write_lock(s);
77	if (!journal_begin(&th, s, 1))
78		if (!journal_end_sync(&th, s, 1))
79			reiserfs_flush_old_commits(s);
80	reiserfs_write_unlock(s);
81	return 0;
82}
83
84static void flush_old_commits(struct work_struct *work)
85{
86	struct reiserfs_sb_info *sbi;
87	struct super_block *s;
88
89	sbi = container_of(work, struct reiserfs_sb_info, old_work.work);
90	s = sbi->s_journal->j_work_sb;
91
92	spin_lock(&sbi->old_work_lock);
93	sbi->work_queued = 0;
94	spin_unlock(&sbi->old_work_lock);
95
96	reiserfs_sync_fs(s, 1);
97}
98
99void reiserfs_schedule_old_flush(struct super_block *s)
100{
101	struct reiserfs_sb_info *sbi = REISERFS_SB(s);
102	unsigned long delay;
103
104	if (s->s_flags & MS_RDONLY)
105		return;
106
107	spin_lock(&sbi->old_work_lock);
108	if (!sbi->work_queued) {
109		delay = msecs_to_jiffies(dirty_writeback_interval * 10);
110		queue_delayed_work(system_long_wq, &sbi->old_work, delay);
111		sbi->work_queued = 1;
112	}
113	spin_unlock(&sbi->old_work_lock);
114}
115
116static void cancel_old_flush(struct super_block *s)
117{
118	struct reiserfs_sb_info *sbi = REISERFS_SB(s);
119
120	cancel_delayed_work_sync(&REISERFS_SB(s)->old_work);
121	spin_lock(&sbi->old_work_lock);
122	sbi->work_queued = 0;
123	spin_unlock(&sbi->old_work_lock);
124}
125
126static int reiserfs_freeze(struct super_block *s)
127{
128	struct reiserfs_transaction_handle th;
129
130	cancel_old_flush(s);
131
132	reiserfs_write_lock(s);
133	if (!(s->s_flags & MS_RDONLY)) {
134		int err = journal_begin(&th, s, 1);
135		if (err) {
136			reiserfs_block_writes(&th);
137		} else {
138			reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s),
139						     1);
140			journal_mark_dirty(&th, s, SB_BUFFER_WITH_SB(s));
141			reiserfs_block_writes(&th);
142			journal_end_sync(&th, s, 1);
143		}
144	}
145	reiserfs_write_unlock(s);
146	return 0;
147}
148
149static int reiserfs_unfreeze(struct super_block *s)
150{
151	reiserfs_allow_writes(s);
152	return 0;
153}
154
155extern const struct in_core_key MAX_IN_CORE_KEY;
156
157/* this is used to delete "save link" when there are no items of a
158   file it points to. It can either happen if unlink is completed but
159   "save unlink" removal, or if file has both unlink and truncate
160   pending and as unlink completes first (because key of "save link"
161   protecting unlink is bigger that a key lf "save link" which
162   protects truncate), so there left no items to make truncate
163   completion on */
164static int remove_save_link_only(struct super_block *s,
165				 struct reiserfs_key *key, int oid_free)
166{
167	struct reiserfs_transaction_handle th;
168	int err;
169
170	/* we are going to do one balancing */
171	err = journal_begin(&th, s, JOURNAL_PER_BALANCE_CNT);
172	if (err)
173		return err;
174
175	reiserfs_delete_solid_item(&th, NULL, key);
176	if (oid_free)
177		/* removals are protected by direct items */
178		reiserfs_release_objectid(&th, le32_to_cpu(key->k_objectid));
179
180	return journal_end(&th, s, JOURNAL_PER_BALANCE_CNT);
181}
182
183#ifdef CONFIG_QUOTA
184static int reiserfs_quota_on_mount(struct super_block *, int);
185#endif
186
187/* look for uncompleted unlinks and truncates and complete them */
188static int finish_unfinished(struct super_block *s)
189{
190	INITIALIZE_PATH(path);
191	struct cpu_key max_cpu_key, obj_key;
192	struct reiserfs_key save_link_key, last_inode_key;
193	int retval = 0;
194	struct item_head *ih;
195	struct buffer_head *bh;
196	int item_pos;
197	char *item;
198	int done;
199	struct inode *inode;
200	int truncate;
201#ifdef CONFIG_QUOTA
202	int i;
203	int ms_active_set;
204	int quota_enabled[MAXQUOTAS];
205#endif
206
207	/* compose key to look for "save" links */
208	max_cpu_key.version = KEY_FORMAT_3_5;
209	max_cpu_key.on_disk_key.k_dir_id = ~0U;
210	max_cpu_key.on_disk_key.k_objectid = ~0U;
211	set_cpu_key_k_offset(&max_cpu_key, ~0U);
212	max_cpu_key.key_length = 3;
213
214	memset(&last_inode_key, 0, sizeof(last_inode_key));
215
216#ifdef CONFIG_QUOTA
217	/* Needed for iput() to work correctly and not trash data */
218	if (s->s_flags & MS_ACTIVE) {
219		ms_active_set = 0;
220	} else {
221		ms_active_set = 1;
222		s->s_flags |= MS_ACTIVE;
223	}
224	/* Turn on quotas so that they are updated correctly */
225	for (i = 0; i < MAXQUOTAS; i++) {
226		quota_enabled[i] = 1;
227		if (REISERFS_SB(s)->s_qf_names[i]) {
228			int ret;
229
230			if (sb_has_quota_active(s, i)) {
231				quota_enabled[i] = 0;
232				continue;
233			}
234			ret = reiserfs_quota_on_mount(s, i);
235			if (ret < 0)
236				reiserfs_warning(s, "reiserfs-2500",
237						 "cannot turn on journaled "
238						 "quota: error %d", ret);
239		}
240	}
241#endif
242
243	done = 0;
244	REISERFS_SB(s)->s_is_unlinked_ok = 1;
245	while (!retval) {
246		retval = search_item(s, &max_cpu_key, &path);
247		if (retval != ITEM_NOT_FOUND) {
248			reiserfs_error(s, "vs-2140",
249				       "search_by_key returned %d", retval);
250			break;
251		}
252
253		bh = get_last_bh(&path);
254		item_pos = get_item_pos(&path);
255		if (item_pos != B_NR_ITEMS(bh)) {
256			reiserfs_warning(s, "vs-2060",
257					 "wrong position found");
258			break;
259		}
260		item_pos--;
261		ih = B_N_PITEM_HEAD(bh, item_pos);
262
263		if (le32_to_cpu(ih->ih_key.k_dir_id) != MAX_KEY_OBJECTID)
264			/* there are no "save" links anymore */
265			break;
266
267		save_link_key = ih->ih_key;
268		if (is_indirect_le_ih(ih))
269			truncate = 1;
270		else
271			truncate = 0;
272
273		/* reiserfs_iget needs k_dirid and k_objectid only */
274		item = B_I_PITEM(bh, ih);
275		obj_key.on_disk_key.k_dir_id = le32_to_cpu(*(__le32 *) item);
276		obj_key.on_disk_key.k_objectid =
277		    le32_to_cpu(ih->ih_key.k_objectid);
278		obj_key.on_disk_key.k_offset = 0;
279		obj_key.on_disk_key.k_type = 0;
280
281		pathrelse(&path);
282
283		inode = reiserfs_iget(s, &obj_key);
284		if (!inode) {
285			/* the unlink almost completed, it just did not manage to remove
286			   "save" link and release objectid */
287			reiserfs_warning(s, "vs-2180", "iget failed for %K",
288					 &obj_key);
289			retval = remove_save_link_only(s, &save_link_key, 1);
290			continue;
291		}
292
293		if (!truncate && inode->i_nlink) {
294			/* file is not unlinked */
295			reiserfs_warning(s, "vs-2185",
296					 "file %K is not unlinked",
297					 &obj_key);
298			retval = remove_save_link_only(s, &save_link_key, 0);
299			continue;
300		}
301		reiserfs_write_unlock(s);
302		dquot_initialize(inode);
303		reiserfs_write_lock(s);
304
305		if (truncate && S_ISDIR(inode->i_mode)) {
306			/* We got a truncate request for a dir which is impossible.
307			   The only imaginable way is to execute unfinished truncate request
308			   then boot into old kernel, remove the file and create dir with
309			   the same key. */
310			reiserfs_warning(s, "green-2101",
311					 "impossible truncate on a "
312					 "directory %k. Please report",
313					 INODE_PKEY(inode));
314			retval = remove_save_link_only(s, &save_link_key, 0);
315			truncate = 0;
316			iput(inode);
317			continue;
318		}
319
320		if (truncate) {
321			REISERFS_I(inode)->i_flags |=
322			    i_link_saved_truncate_mask;
323			/* not completed truncate found. New size was committed together
324			   with "save" link */
325			reiserfs_info(s, "Truncating %k to %Ld ..",
326				      INODE_PKEY(inode), inode->i_size);
327			reiserfs_truncate_file(inode,
328					       0
329					       /*don't update modification time */
330					       );
331			retval = remove_save_link(inode, truncate);
332		} else {
333			REISERFS_I(inode)->i_flags |= i_link_saved_unlink_mask;
334			/* not completed unlink (rmdir) found */
335			reiserfs_info(s, "Removing %k..", INODE_PKEY(inode));
336			if (memcmp(&last_inode_key, INODE_PKEY(inode),
337					sizeof(last_inode_key))){
338				last_inode_key = *INODE_PKEY(inode);
339				/* removal gets completed in iput */
340				retval = 0;
341			} else {
342				reiserfs_warning(s, "super-2189", "Dead loop "
343						 "in finish_unfinished "
344						 "detected, just remove "
345						 "save link\n");
346				retval = remove_save_link_only(s,
347							&save_link_key, 0);
348			}
349		}
350
351		iput(inode);
352		printk("done\n");
353		done++;
354	}
355	REISERFS_SB(s)->s_is_unlinked_ok = 0;
356
357#ifdef CONFIG_QUOTA
358	/* Turn quotas off */
359	for (i = 0; i < MAXQUOTAS; i++) {
360		if (sb_dqopt(s)->files[i] && quota_enabled[i])
361			dquot_quota_off(s, i);
362	}
363	if (ms_active_set)
364		/* Restore the flag back */
365		s->s_flags &= ~MS_ACTIVE;
366#endif
367	pathrelse(&path);
368	if (done)
369		reiserfs_info(s, "There were %d uncompleted unlinks/truncates. "
370			      "Completed\n", done);
371	return retval;
372}
373
374/* to protect file being unlinked from getting lost we "safe" link files
375   being unlinked. This link will be deleted in the same transaction with last
376   item of file. mounting the filesystem we scan all these links and remove
377   files which almost got lost */
378void add_save_link(struct reiserfs_transaction_handle *th,
379		   struct inode *inode, int truncate)
380{
381	INITIALIZE_PATH(path);
382	int retval;
383	struct cpu_key key;
384	struct item_head ih;
385	__le32 link;
386
387	BUG_ON(!th->t_trans_id);
388
389	/* file can only get one "save link" of each kind */
390	RFALSE(truncate &&
391	       (REISERFS_I(inode)->i_flags & i_link_saved_truncate_mask),
392	       "saved link already exists for truncated inode %lx",
393	       (long)inode->i_ino);
394	RFALSE(!truncate &&
395	       (REISERFS_I(inode)->i_flags & i_link_saved_unlink_mask),
396	       "saved link already exists for unlinked inode %lx",
397	       (long)inode->i_ino);
398
399	/* setup key of "save" link */
400	key.version = KEY_FORMAT_3_5;
401	key.on_disk_key.k_dir_id = MAX_KEY_OBJECTID;
402	key.on_disk_key.k_objectid = inode->i_ino;
403	if (!truncate) {
404		/* unlink, rmdir, rename */
405		set_cpu_key_k_offset(&key, 1 + inode->i_sb->s_blocksize);
406		set_cpu_key_k_type(&key, TYPE_DIRECT);
407
408		/* item head of "safe" link */
409		make_le_item_head(&ih, &key, key.version,
410				  1 + inode->i_sb->s_blocksize, TYPE_DIRECT,
411				  4 /*length */ , 0xffff /*free space */ );
412	} else {
413		/* truncate */
414		if (S_ISDIR(inode->i_mode))
415			reiserfs_warning(inode->i_sb, "green-2102",
416					 "Adding a truncate savelink for "
417					 "a directory %k! Please report",
418					 INODE_PKEY(inode));
419		set_cpu_key_k_offset(&key, 1);
420		set_cpu_key_k_type(&key, TYPE_INDIRECT);
421
422		/* item head of "safe" link */
423		make_le_item_head(&ih, &key, key.version, 1, TYPE_INDIRECT,
424				  4 /*length */ , 0 /*free space */ );
425	}
426	key.key_length = 3;
427
428	/* look for its place in the tree */
429	retval = search_item(inode->i_sb, &key, &path);
430	if (retval != ITEM_NOT_FOUND) {
431		if (retval != -ENOSPC)
432			reiserfs_error(inode->i_sb, "vs-2100",
433				       "search_by_key (%K) returned %d", &key,
434				       retval);
435		pathrelse(&path);
436		return;
437	}
438
439	/* body of "save" link */
440	link = INODE_PKEY(inode)->k_dir_id;
441
442	/* put "save" link into tree, don't charge quota to anyone */
443	retval =
444	    reiserfs_insert_item(th, &path, &key, &ih, NULL, (char *)&link);
445	if (retval) {
446		if (retval != -ENOSPC)
447			reiserfs_error(inode->i_sb, "vs-2120",
448				       "insert_item returned %d", retval);
449	} else {
450		if (truncate)
451			REISERFS_I(inode)->i_flags |=
452			    i_link_saved_truncate_mask;
453		else
454			REISERFS_I(inode)->i_flags |= i_link_saved_unlink_mask;
455	}
456}
457
458/* this opens transaction unlike add_save_link */
459int remove_save_link(struct inode *inode, int truncate)
460{
461	struct reiserfs_transaction_handle th;
462	struct reiserfs_key key;
463	int err;
464
465	/* we are going to do one balancing only */
466	err = journal_begin(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT);
467	if (err)
468		return err;
469
470	/* setup key of "save" link */
471	key.k_dir_id = cpu_to_le32(MAX_KEY_OBJECTID);
472	key.k_objectid = INODE_PKEY(inode)->k_objectid;
473	if (!truncate) {
474		/* unlink, rmdir, rename */
475		set_le_key_k_offset(KEY_FORMAT_3_5, &key,
476				    1 + inode->i_sb->s_blocksize);
477		set_le_key_k_type(KEY_FORMAT_3_5, &key, TYPE_DIRECT);
478	} else {
479		/* truncate */
480		set_le_key_k_offset(KEY_FORMAT_3_5, &key, 1);
481		set_le_key_k_type(KEY_FORMAT_3_5, &key, TYPE_INDIRECT);
482	}
483
484	if ((truncate &&
485	     (REISERFS_I(inode)->i_flags & i_link_saved_truncate_mask)) ||
486	    (!truncate &&
487	     (REISERFS_I(inode)->i_flags & i_link_saved_unlink_mask)))
488		/* don't take quota bytes from anywhere */
489		reiserfs_delete_solid_item(&th, NULL, &key);
490	if (!truncate) {
491		reiserfs_release_objectid(&th, inode->i_ino);
492		REISERFS_I(inode)->i_flags &= ~i_link_saved_unlink_mask;
493	} else
494		REISERFS_I(inode)->i_flags &= ~i_link_saved_truncate_mask;
495
496	return journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT);
497}
498
499static void reiserfs_kill_sb(struct super_block *s)
500{
501	if (REISERFS_SB(s)) {
502		/*
503		 * Force any pending inode evictions to occur now. Any
504		 * inodes to be removed that have extended attributes
505		 * associated with them need to clean them up before
506		 * we can release the extended attribute root dentries.
507		 * shrink_dcache_for_umount will BUG if we don't release
508		 * those before it's called so ->put_super is too late.
509		 */
510		shrink_dcache_sb(s);
511
512		dput(REISERFS_SB(s)->xattr_root);
513		REISERFS_SB(s)->xattr_root = NULL;
514		dput(REISERFS_SB(s)->priv_root);
515		REISERFS_SB(s)->priv_root = NULL;
516	}
517
518	kill_block_super(s);
519}
520
521static void reiserfs_put_super(struct super_block *s)
522{
523	struct reiserfs_transaction_handle th;
524	th.t_trans_id = 0;
525
526	dquot_disable(s, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
527
528	reiserfs_write_lock(s);
529
530	/* change file system state to current state if it was mounted with read-write permissions */
531	if (!(s->s_flags & MS_RDONLY)) {
532		if (!journal_begin(&th, s, 10)) {
533			reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s),
534						     1);
535			set_sb_umount_state(SB_DISK_SUPER_BLOCK(s),
536					    REISERFS_SB(s)->s_mount_state);
537			journal_mark_dirty(&th, s, SB_BUFFER_WITH_SB(s));
538		}
539	}
540
541	/* note, journal_release checks for readonly mount, and can decide not
542	 ** to do a journal_end
543	 */
544	journal_release(&th, s);
545
546	reiserfs_free_bitmap_cache(s);
547
548	brelse(SB_BUFFER_WITH_SB(s));
549
550	print_statistics(s);
551
552	if (REISERFS_SB(s)->reserved_blocks != 0) {
553		reiserfs_warning(s, "green-2005", "reserved blocks left %d",
554				 REISERFS_SB(s)->reserved_blocks);
555	}
556
557	reiserfs_proc_info_done(s);
558
559	reiserfs_write_unlock(s);
560	mutex_destroy(&REISERFS_SB(s)->lock);
561	kfree(s->s_fs_info);
562	s->s_fs_info = NULL;
563}
564
565static struct kmem_cache *reiserfs_inode_cachep;
566
567static struct inode *reiserfs_alloc_inode(struct super_block *sb)
568{
569	struct reiserfs_inode_info *ei;
570	ei = (struct reiserfs_inode_info *)
571	    kmem_cache_alloc(reiserfs_inode_cachep, GFP_KERNEL);
572	if (!ei)
573		return NULL;
574	atomic_set(&ei->openers, 0);
575	mutex_init(&ei->tailpack);
576	return &ei->vfs_inode;
577}
578
579static void reiserfs_i_callback(struct rcu_head *head)
580{
581	struct inode *inode = container_of(head, struct inode, i_rcu);
582	kmem_cache_free(reiserfs_inode_cachep, REISERFS_I(inode));
583}
584
585static void reiserfs_destroy_inode(struct inode *inode)
586{
587	call_rcu(&inode->i_rcu, reiserfs_i_callback);
588}
589
590static void init_once(void *foo)
591{
592	struct reiserfs_inode_info *ei = (struct reiserfs_inode_info *)foo;
593
594	INIT_LIST_HEAD(&ei->i_prealloc_list);
595	inode_init_once(&ei->vfs_inode);
596}
597
598static int init_inodecache(void)
599{
600	reiserfs_inode_cachep = kmem_cache_create("reiser_inode_cache",
601						  sizeof(struct
602							 reiserfs_inode_info),
603						  0, (SLAB_RECLAIM_ACCOUNT|
604							SLAB_MEM_SPREAD),
605						  init_once);
606	if (reiserfs_inode_cachep == NULL)
607		return -ENOMEM;
608	return 0;
609}
610
611static void destroy_inodecache(void)
612{
613	/*
614	 * Make sure all delayed rcu free inodes are flushed before we
615	 * destroy cache.
616	 */
617	rcu_barrier();
618	kmem_cache_destroy(reiserfs_inode_cachep);
619}
620
621/* we don't mark inodes dirty, we just log them */
622static void reiserfs_dirty_inode(struct inode *inode, int flags)
623{
624	struct reiserfs_transaction_handle th;
625
626	int err = 0;
627	int lock_depth;
628
629	if (inode->i_sb->s_flags & MS_RDONLY) {
630		reiserfs_warning(inode->i_sb, "clm-6006",
631				 "writing inode %lu on readonly FS",
632				 inode->i_ino);
633		return;
634	}
635	lock_depth = reiserfs_write_lock_once(inode->i_sb);
636
637	/* this is really only used for atime updates, so they don't have
638	 ** to be included in O_SYNC or fsync
639	 */
640	err = journal_begin(&th, inode->i_sb, 1);
641	if (err)
642		goto out;
643
644	reiserfs_update_sd(&th, inode);
645	journal_end(&th, inode->i_sb, 1);
646
647out:
648	reiserfs_write_unlock_once(inode->i_sb, lock_depth);
649}
650
651static int reiserfs_show_options(struct seq_file *seq, struct dentry *root)
652{
653	struct super_block *s = root->d_sb;
654	struct reiserfs_journal *journal = SB_JOURNAL(s);
655	long opts = REISERFS_SB(s)->s_mount_opt;
656
657	if (opts & (1 << REISERFS_LARGETAIL))
658		seq_puts(seq, ",tails=on");
659	else if (!(opts & (1 << REISERFS_SMALLTAIL)))
660		seq_puts(seq, ",notail");
661	/* tails=small is default so we don't show it */
662
663	if (!(opts & (1 << REISERFS_BARRIER_FLUSH)))
664		seq_puts(seq, ",barrier=none");
665	/* barrier=flush is default so we don't show it */
666
667	if (opts & (1 << REISERFS_ERROR_CONTINUE))
668		seq_puts(seq, ",errors=continue");
669	else if (opts & (1 << REISERFS_ERROR_PANIC))
670		seq_puts(seq, ",errors=panic");
671	/* errors=ro is default so we don't show it */
672
673	if (opts & (1 << REISERFS_DATA_LOG))
674		seq_puts(seq, ",data=journal");
675	else if (opts & (1 << REISERFS_DATA_WRITEBACK))
676		seq_puts(seq, ",data=writeback");
677	/* data=ordered is default so we don't show it */
678
679	if (opts & (1 << REISERFS_ATTRS))
680		seq_puts(seq, ",attrs");
681
682	if (opts & (1 << REISERFS_XATTRS_USER))
683		seq_puts(seq, ",user_xattr");
684
685	if (opts & (1 << REISERFS_EXPOSE_PRIVROOT))
686		seq_puts(seq, ",expose_privroot");
687
688	if (opts & (1 << REISERFS_POSIXACL))
689		seq_puts(seq, ",acl");
690
691	if (REISERFS_SB(s)->s_jdev)
692		seq_printf(seq, ",jdev=%s", REISERFS_SB(s)->s_jdev);
693
694	if (journal->j_max_commit_age != journal->j_default_max_commit_age)
695		seq_printf(seq, ",commit=%d", journal->j_max_commit_age);
696
697#ifdef CONFIG_QUOTA
698	if (REISERFS_SB(s)->s_qf_names[USRQUOTA])
699		seq_printf(seq, ",usrjquota=%s", REISERFS_SB(s)->s_qf_names[USRQUOTA]);
700	else if (opts & (1 << REISERFS_USRQUOTA))
701		seq_puts(seq, ",usrquota");
702	if (REISERFS_SB(s)->s_qf_names[GRPQUOTA])
703		seq_printf(seq, ",grpjquota=%s", REISERFS_SB(s)->s_qf_names[GRPQUOTA]);
704	else if (opts & (1 << REISERFS_GRPQUOTA))
705		seq_puts(seq, ",grpquota");
706	if (REISERFS_SB(s)->s_jquota_fmt) {
707		if (REISERFS_SB(s)->s_jquota_fmt == QFMT_VFS_OLD)
708			seq_puts(seq, ",jqfmt=vfsold");
709		else if (REISERFS_SB(s)->s_jquota_fmt == QFMT_VFS_V0)
710			seq_puts(seq, ",jqfmt=vfsv0");
711	}
712#endif
713
714	/* Block allocator options */
715	if (opts & (1 << REISERFS_NO_BORDER))
716		seq_puts(seq, ",block-allocator=noborder");
717	if (opts & (1 << REISERFS_NO_UNHASHED_RELOCATION))
718		seq_puts(seq, ",block-allocator=no_unhashed_relocation");
719	if (opts & (1 << REISERFS_HASHED_RELOCATION))
720		seq_puts(seq, ",block-allocator=hashed_relocation");
721	if (opts & (1 << REISERFS_TEST4))
722		seq_puts(seq, ",block-allocator=test4");
723	show_alloc_options(seq, s);
724	return 0;
725}
726
727#ifdef CONFIG_QUOTA
728static ssize_t reiserfs_quota_write(struct super_block *, int, const char *,
729				    size_t, loff_t);
730static ssize_t reiserfs_quota_read(struct super_block *, int, char *, size_t,
731				   loff_t);
732#endif
733
734static const struct super_operations reiserfs_sops = {
735	.alloc_inode = reiserfs_alloc_inode,
736	.destroy_inode = reiserfs_destroy_inode,
737	.write_inode = reiserfs_write_inode,
738	.dirty_inode = reiserfs_dirty_inode,
739	.evict_inode = reiserfs_evict_inode,
740	.put_super = reiserfs_put_super,
741	.sync_fs = reiserfs_sync_fs,
742	.freeze_fs = reiserfs_freeze,
743	.unfreeze_fs = reiserfs_unfreeze,
744	.statfs = reiserfs_statfs,
745	.remount_fs = reiserfs_remount,
746	.show_options = reiserfs_show_options,
747#ifdef CONFIG_QUOTA
748	.quota_read = reiserfs_quota_read,
749	.quota_write = reiserfs_quota_write,
750#endif
751};
752
753#ifdef CONFIG_QUOTA
754#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
755
756static int reiserfs_write_dquot(struct dquot *);
757static int reiserfs_acquire_dquot(struct dquot *);
758static int reiserfs_release_dquot(struct dquot *);
759static int reiserfs_mark_dquot_dirty(struct dquot *);
760static int reiserfs_write_info(struct super_block *, int);
761static int reiserfs_quota_on(struct super_block *, int, int, struct path *);
762
763static const struct dquot_operations reiserfs_quota_operations = {
764	.write_dquot = reiserfs_write_dquot,
765	.acquire_dquot = reiserfs_acquire_dquot,
766	.release_dquot = reiserfs_release_dquot,
767	.mark_dirty = reiserfs_mark_dquot_dirty,
768	.write_info = reiserfs_write_info,
769	.alloc_dquot	= dquot_alloc,
770	.destroy_dquot	= dquot_destroy,
771};
772
773static const struct quotactl_ops reiserfs_qctl_operations = {
774	.quota_on = reiserfs_quota_on,
775	.quota_off = dquot_quota_off,
776	.quota_sync = dquot_quota_sync,
777	.get_info = dquot_get_dqinfo,
778	.set_info = dquot_set_dqinfo,
779	.get_dqblk = dquot_get_dqblk,
780	.set_dqblk = dquot_set_dqblk,
781};
782#endif
783
784static const struct export_operations reiserfs_export_ops = {
785	.encode_fh = reiserfs_encode_fh,
786	.fh_to_dentry = reiserfs_fh_to_dentry,
787	.fh_to_parent = reiserfs_fh_to_parent,
788	.get_parent = reiserfs_get_parent,
789};
790
791/* this struct is used in reiserfs_getopt () for containing the value for those
792   mount options that have values rather than being toggles. */
793typedef struct {
794	char *value;
795	int setmask;		/* bitmask which is to set on mount_options bitmask when this
796				   value is found, 0 is no bits are to be changed. */
797	int clrmask;		/* bitmask which is to clear on mount_options bitmask when  this
798				   value is found, 0 is no bits are to be changed. This is
799				   applied BEFORE setmask */
800} arg_desc_t;
801
802/* Set this bit in arg_required to allow empty arguments */
803#define REISERFS_OPT_ALLOWEMPTY 31
804
805/* this struct is used in reiserfs_getopt() for describing the set of reiserfs
806   mount options */
807typedef struct {
808	char *option_name;
809	int arg_required;	/* 0 if argument is not required, not 0 otherwise */
810	const arg_desc_t *values;	/* list of values accepted by an option */
811	int setmask;		/* bitmask which is to set on mount_options bitmask when this
812				   value is found, 0 is no bits are to be changed. */
813	int clrmask;		/* bitmask which is to clear on mount_options bitmask when  this
814				   value is found, 0 is no bits are to be changed. This is
815				   applied BEFORE setmask */
816} opt_desc_t;
817
818/* possible values for -o data= */
819static const arg_desc_t logging_mode[] = {
820	{"ordered", 1 << REISERFS_DATA_ORDERED,
821	 (1 << REISERFS_DATA_LOG | 1 << REISERFS_DATA_WRITEBACK)},
822	{"journal", 1 << REISERFS_DATA_LOG,
823	 (1 << REISERFS_DATA_ORDERED | 1 << REISERFS_DATA_WRITEBACK)},
824	{"writeback", 1 << REISERFS_DATA_WRITEBACK,
825	 (1 << REISERFS_DATA_ORDERED | 1 << REISERFS_DATA_LOG)},
826	{.value = NULL}
827};
828
829/* possible values for -o barrier= */
830static const arg_desc_t barrier_mode[] = {
831	{"none", 1 << REISERFS_BARRIER_NONE, 1 << REISERFS_BARRIER_FLUSH},
832	{"flush", 1 << REISERFS_BARRIER_FLUSH, 1 << REISERFS_BARRIER_NONE},
833	{.value = NULL}
834};
835
836/* possible values for "-o block-allocator=" and bits which are to be set in
837   s_mount_opt of reiserfs specific part of in-core super block */
838static const arg_desc_t balloc[] = {
839	{"noborder", 1 << REISERFS_NO_BORDER, 0},
840	{"border", 0, 1 << REISERFS_NO_BORDER},
841	{"no_unhashed_relocation", 1 << REISERFS_NO_UNHASHED_RELOCATION, 0},
842	{"hashed_relocation", 1 << REISERFS_HASHED_RELOCATION, 0},
843	{"test4", 1 << REISERFS_TEST4, 0},
844	{"notest4", 0, 1 << REISERFS_TEST4},
845	{NULL, 0, 0}
846};
847
848static const arg_desc_t tails[] = {
849	{"on", 1 << REISERFS_LARGETAIL, 1 << REISERFS_SMALLTAIL},
850	{"off", 0, (1 << REISERFS_LARGETAIL) | (1 << REISERFS_SMALLTAIL)},
851	{"small", 1 << REISERFS_SMALLTAIL, 1 << REISERFS_LARGETAIL},
852	{NULL, 0, 0}
853};
854
855static const arg_desc_t error_actions[] = {
856	{"panic", 1 << REISERFS_ERROR_PANIC,
857	 (1 << REISERFS_ERROR_RO | 1 << REISERFS_ERROR_CONTINUE)},
858	{"ro-remount", 1 << REISERFS_ERROR_RO,
859	 (1 << REISERFS_ERROR_PANIC | 1 << REISERFS_ERROR_CONTINUE)},
860#ifdef REISERFS_JOURNAL_ERROR_ALLOWS_NO_LOG
861	{"continue", 1 << REISERFS_ERROR_CONTINUE,
862	 (1 << REISERFS_ERROR_PANIC | 1 << REISERFS_ERROR_RO)},
863#endif
864	{NULL, 0, 0},
865};
866
867/* proceed only one option from a list *cur - string containing of mount options
868   opts - array of options which are accepted
869   opt_arg - if option is found and requires an argument and if it is specifed
870   in the input - pointer to the argument is stored here
871   bit_flags - if option requires to set a certain bit - it is set here
872   return -1 if unknown option is found, opt->arg_required otherwise */
873static int reiserfs_getopt(struct super_block *s, char **cur, opt_desc_t * opts,
874			   char **opt_arg, unsigned long *bit_flags)
875{
876	char *p;
877	/* foo=bar,
878	   ^   ^  ^
879	   |   |  +-- option_end
880	   |   +-- arg_start
881	   +-- option_start
882	 */
883	const opt_desc_t *opt;
884	const arg_desc_t *arg;
885
886	p = *cur;
887
888	/* assume argument cannot contain commas */
889	*cur = strchr(p, ',');
890	if (*cur) {
891		*(*cur) = '\0';
892		(*cur)++;
893	}
894
895	if (!strncmp(p, "alloc=", 6)) {
896		/* Ugly special case, probably we should redo options parser so that
897		   it can understand several arguments for some options, also so that
898		   it can fill several bitfields with option values. */
899		if (reiserfs_parse_alloc_options(s, p + 6)) {
900			return -1;
901		} else {
902			return 0;
903		}
904	}
905
906	/* for every option in the list */
907	for (opt = opts; opt->option_name; opt++) {
908		if (!strncmp(p, opt->option_name, strlen(opt->option_name))) {
909			if (bit_flags) {
910				if (opt->clrmask ==
911				    (1 << REISERFS_UNSUPPORTED_OPT))
912					reiserfs_warning(s, "super-6500",
913							 "%s not supported.\n",
914							 p);
915				else
916					*bit_flags &= ~opt->clrmask;
917				if (opt->setmask ==
918				    (1 << REISERFS_UNSUPPORTED_OPT))
919					reiserfs_warning(s, "super-6501",
920							 "%s not supported.\n",
921							 p);
922				else
923					*bit_flags |= opt->setmask;
924			}
925			break;
926		}
927	}
928	if (!opt->option_name) {
929		reiserfs_warning(s, "super-6502",
930				 "unknown mount option \"%s\"", p);
931		return -1;
932	}
933
934	p += strlen(opt->option_name);
935	switch (*p) {
936	case '=':
937		if (!opt->arg_required) {
938			reiserfs_warning(s, "super-6503",
939					 "the option \"%s\" does not "
940					 "require an argument\n",
941					 opt->option_name);
942			return -1;
943		}
944		break;
945
946	case 0:
947		if (opt->arg_required) {
948			reiserfs_warning(s, "super-6504",
949					 "the option \"%s\" requires an "
950					 "argument\n", opt->option_name);
951			return -1;
952		}
953		break;
954	default:
955		reiserfs_warning(s, "super-6505",
956				 "head of option \"%s\" is only correct\n",
957				 opt->option_name);
958		return -1;
959	}
960
961	/* move to the argument, or to next option if argument is not required */
962	p++;
963
964	if (opt->arg_required
965	    && !(opt->arg_required & (1 << REISERFS_OPT_ALLOWEMPTY))
966	    && !strlen(p)) {
967		/* this catches "option=," if not allowed */
968		reiserfs_warning(s, "super-6506",
969				 "empty argument for \"%s\"\n",
970				 opt->option_name);
971		return -1;
972	}
973
974	if (!opt->values) {
975		/* *=NULLopt_arg contains pointer to argument */
976		*opt_arg = p;
977		return opt->arg_required & ~(1 << REISERFS_OPT_ALLOWEMPTY);
978	}
979
980	/* values possible for this option are listed in opt->values */
981	for (arg = opt->values; arg->value; arg++) {
982		if (!strcmp(p, arg->value)) {
983			if (bit_flags) {
984				*bit_flags &= ~arg->clrmask;
985				*bit_flags |= arg->setmask;
986			}
987			return opt->arg_required;
988		}
989	}
990
991	reiserfs_warning(s, "super-6506",
992			 "bad value \"%s\" for option \"%s\"\n", p,
993			 opt->option_name);
994	return -1;
995}
996
997/* returns 0 if something is wrong in option string, 1 - otherwise */
998static int reiserfs_parse_options(struct super_block *s, char *options,	/* string given via mount's -o */
999				  unsigned long *mount_options,
1000				  /* after the parsing phase, contains the
1001				     collection of bitflags defining what
1002				     mount options were selected. */
1003				  unsigned long *blocks,	/* strtol-ed from NNN of resize=NNN */
1004				  char **jdev_name,
1005				  unsigned int *commit_max_age,
1006				  char **qf_names,
1007				  unsigned int *qfmt)
1008{
1009	int c;
1010	char *arg = NULL;
1011	char *pos;
1012	opt_desc_t opts[] = {
1013		/* Compatibility stuff, so that -o notail for old setups still work */
1014		{"tails",.arg_required = 't',.values = tails},
1015		{"notail",.clrmask =
1016		 (1 << REISERFS_LARGETAIL) | (1 << REISERFS_SMALLTAIL)},
1017		{"conv",.setmask = 1 << REISERFS_CONVERT},
1018		{"attrs",.setmask = 1 << REISERFS_ATTRS},
1019		{"noattrs",.clrmask = 1 << REISERFS_ATTRS},
1020		{"expose_privroot", .setmask = 1 << REISERFS_EXPOSE_PRIVROOT},
1021#ifdef CONFIG_REISERFS_FS_XATTR
1022		{"user_xattr",.setmask = 1 << REISERFS_XATTRS_USER},
1023		{"nouser_xattr",.clrmask = 1 << REISERFS_XATTRS_USER},
1024#else
1025		{"user_xattr",.setmask = 1 << REISERFS_UNSUPPORTED_OPT},
1026		{"nouser_xattr",.clrmask = 1 << REISERFS_UNSUPPORTED_OPT},
1027#endif
1028#ifdef CONFIG_REISERFS_FS_POSIX_ACL
1029		{"acl",.setmask = 1 << REISERFS_POSIXACL},
1030		{"noacl",.clrmask = 1 << REISERFS_POSIXACL},
1031#else
1032		{"acl",.setmask = 1 << REISERFS_UNSUPPORTED_OPT},
1033		{"noacl",.clrmask = 1 << REISERFS_UNSUPPORTED_OPT},
1034#endif
1035		{.option_name = "nolog"},
1036		{"replayonly",.setmask = 1 << REPLAYONLY},
1037		{"block-allocator",.arg_required = 'a',.values = balloc},
1038		{"data",.arg_required = 'd',.values = logging_mode},
1039		{"barrier",.arg_required = 'b',.values = barrier_mode},
1040		{"resize",.arg_required = 'r',.values = NULL},
1041		{"jdev",.arg_required = 'j',.values = NULL},
1042		{"nolargeio",.arg_required = 'w',.values = NULL},
1043		{"commit",.arg_required = 'c',.values = NULL},
1044		{"usrquota",.setmask = 1 << REISERFS_USRQUOTA},
1045		{"grpquota",.setmask = 1 << REISERFS_GRPQUOTA},
1046		{"noquota",.clrmask = 1 << REISERFS_USRQUOTA | 1 << REISERFS_GRPQUOTA},
1047		{"errors",.arg_required = 'e',.values = error_actions},
1048		{"usrjquota",.arg_required =
1049		 'u' | (1 << REISERFS_OPT_ALLOWEMPTY),.values = NULL},
1050		{"grpjquota",.arg_required =
1051		 'g' | (1 << REISERFS_OPT_ALLOWEMPTY),.values = NULL},
1052		{"jqfmt",.arg_required = 'f',.values = NULL},
1053		{.option_name = NULL}
1054	};
1055
1056	*blocks = 0;
1057	if (!options || !*options)
1058		/* use default configuration: create tails, journaling on, no
1059		   conversion to newest format */
1060		return 1;
1061
1062	for (pos = options; pos;) {
1063		c = reiserfs_getopt(s, &pos, opts, &arg, mount_options);
1064		if (c == -1)
1065			/* wrong option is given */
1066			return 0;
1067
1068		if (c == 'r') {
1069			char *p;
1070
1071			p = NULL;
1072			/* "resize=NNN" or "resize=auto" */
1073
1074			if (!strcmp(arg, "auto")) {
1075				/* From JFS code, to auto-get the size. */
1076				*blocks =
1077				    s->s_bdev->bd_inode->i_size >> s->
1078				    s_blocksize_bits;
1079			} else {
1080				*blocks = simple_strtoul(arg, &p, 0);
1081				if (*p != '\0') {
1082					/* NNN does not look like a number */
1083					reiserfs_warning(s, "super-6507",
1084							 "bad value %s for "
1085							 "-oresize\n", arg);
1086					return 0;
1087				}
1088			}
1089		}
1090
1091		if (c == 'c') {
1092			char *p = NULL;
1093			unsigned long val = simple_strtoul(arg, &p, 0);
1094			/* commit=NNN (time in seconds) */
1095			if (*p != '\0' || val >= (unsigned int)-1) {
1096				reiserfs_warning(s, "super-6508",
1097						 "bad value %s for -ocommit\n",
1098						 arg);
1099				return 0;
1100			}
1101			*commit_max_age = (unsigned int)val;
1102		}
1103
1104		if (c == 'w') {
1105			reiserfs_warning(s, "super-6509", "nolargeio option "
1106					 "is no longer supported");
1107			return 0;
1108		}
1109
1110		if (c == 'j') {
1111			if (arg && *arg && jdev_name) {
1112				if (*jdev_name) {	//Hm, already assigned?
1113					reiserfs_warning(s, "super-6510",
1114							 "journal device was "
1115							 "already specified to "
1116							 "be %s", *jdev_name);
1117					return 0;
1118				}
1119				*jdev_name = arg;
1120			}
1121		}
1122#ifdef CONFIG_QUOTA
1123		if (c == 'u' || c == 'g') {
1124			int qtype = c == 'u' ? USRQUOTA : GRPQUOTA;
1125
1126			if (sb_any_quota_loaded(s) &&
1127			    (!*arg != !REISERFS_SB(s)->s_qf_names[qtype])) {
1128				reiserfs_warning(s, "super-6511",
1129						 "cannot change journaled "
1130						 "quota options when quota "
1131						 "turned on.");
1132				return 0;
1133			}
1134			if (*arg) {	/* Some filename specified? */
1135				if (REISERFS_SB(s)->s_qf_names[qtype]
1136				    && strcmp(REISERFS_SB(s)->s_qf_names[qtype],
1137					      arg)) {
1138					reiserfs_warning(s, "super-6512",
1139							 "%s quota file "
1140							 "already specified.",
1141							 QTYPE2NAME(qtype));
1142					return 0;
1143				}
1144				if (strchr(arg, '/')) {
1145					reiserfs_warning(s, "super-6513",
1146							 "quotafile must be "
1147							 "on filesystem root.");
1148					return 0;
1149				}
1150				qf_names[qtype] = kstrdup(arg, GFP_KERNEL);
1151				if (!qf_names[qtype]) {
1152					reiserfs_warning(s, "reiserfs-2502",
1153							 "not enough memory "
1154							 "for storing "
1155							 "quotafile name.");
1156					return 0;
1157				}
1158				if (qtype == USRQUOTA)
1159					*mount_options |= 1 << REISERFS_USRQUOTA;
1160				else
1161					*mount_options |= 1 << REISERFS_GRPQUOTA;
1162			} else {
1163				if (qf_names[qtype] !=
1164				    REISERFS_SB(s)->s_qf_names[qtype])
1165					kfree(qf_names[qtype]);
1166				qf_names[qtype] = NULL;
1167				if (qtype == USRQUOTA)
1168					*mount_options &= ~(1 << REISERFS_USRQUOTA);
1169				else
1170					*mount_options &= ~(1 << REISERFS_GRPQUOTA);
1171			}
1172		}
1173		if (c == 'f') {
1174			if (!strcmp(arg, "vfsold"))
1175				*qfmt = QFMT_VFS_OLD;
1176			else if (!strcmp(arg, "vfsv0"))
1177				*qfmt = QFMT_VFS_V0;
1178			else {
1179				reiserfs_warning(s, "super-6514",
1180						 "unknown quota format "
1181						 "specified.");
1182				return 0;
1183			}
1184			if (sb_any_quota_loaded(s) &&
1185			    *qfmt != REISERFS_SB(s)->s_jquota_fmt) {
1186				reiserfs_warning(s, "super-6515",
1187						 "cannot change journaled "
1188						 "quota options when quota "
1189						 "turned on.");
1190				return 0;
1191			}
1192		}
1193#else
1194		if (c == 'u' || c == 'g' || c == 'f') {
1195			reiserfs_warning(s, "reiserfs-2503", "journaled "
1196					 "quota options not supported.");
1197			return 0;
1198		}
1199#endif
1200	}
1201
1202#ifdef CONFIG_QUOTA
1203	if (!REISERFS_SB(s)->s_jquota_fmt && !*qfmt
1204	    && (qf_names[USRQUOTA] || qf_names[GRPQUOTA])) {
1205		reiserfs_warning(s, "super-6515",
1206				 "journaled quota format not specified.");
1207		return 0;
1208	}
1209	if ((!(*mount_options & (1 << REISERFS_USRQUOTA)) &&
1210	       sb_has_quota_loaded(s, USRQUOTA)) ||
1211	    (!(*mount_options & (1 << REISERFS_GRPQUOTA)) &&
1212	       sb_has_quota_loaded(s, GRPQUOTA))) {
1213		reiserfs_warning(s, "super-6516", "quota options must "
1214				 "be present when quota is turned on.");
1215		return 0;
1216	}
1217#endif
1218
1219	return 1;
1220}
1221
1222static void switch_data_mode(struct super_block *s, unsigned long mode)
1223{
1224	REISERFS_SB(s)->s_mount_opt &= ~((1 << REISERFS_DATA_LOG) |
1225					 (1 << REISERFS_DATA_ORDERED) |
1226					 (1 << REISERFS_DATA_WRITEBACK));
1227	REISERFS_SB(s)->s_mount_opt |= (1 << mode);
1228}
1229
1230static void handle_data_mode(struct super_block *s, unsigned long mount_options)
1231{
1232	if (mount_options & (1 << REISERFS_DATA_LOG)) {
1233		if (!reiserfs_data_log(s)) {
1234			switch_data_mode(s, REISERFS_DATA_LOG);
1235			reiserfs_info(s, "switching to journaled data mode\n");
1236		}
1237	} else if (mount_options & (1 << REISERFS_DATA_ORDERED)) {
1238		if (!reiserfs_data_ordered(s)) {
1239			switch_data_mode(s, REISERFS_DATA_ORDERED);
1240			reiserfs_info(s, "switching to ordered data mode\n");
1241		}
1242	} else if (mount_options & (1 << REISERFS_DATA_WRITEBACK)) {
1243		if (!reiserfs_data_writeback(s)) {
1244			switch_data_mode(s, REISERFS_DATA_WRITEBACK);
1245			reiserfs_info(s, "switching to writeback data mode\n");
1246		}
1247	}
1248}
1249
1250static void handle_barrier_mode(struct super_block *s, unsigned long bits)
1251{
1252	int flush = (1 << REISERFS_BARRIER_FLUSH);
1253	int none = (1 << REISERFS_BARRIER_NONE);
1254	int all_barrier = flush | none;
1255
1256	if (bits & all_barrier) {
1257		REISERFS_SB(s)->s_mount_opt &= ~all_barrier;
1258		if (bits & flush) {
1259			REISERFS_SB(s)->s_mount_opt |= flush;
1260			printk("reiserfs: enabling write barrier flush mode\n");
1261		} else if (bits & none) {
1262			REISERFS_SB(s)->s_mount_opt |= none;
1263			printk("reiserfs: write barriers turned off\n");
1264		}
1265	}
1266}
1267
1268static void handle_attrs(struct super_block *s)
1269{
1270	struct reiserfs_super_block *rs = SB_DISK_SUPER_BLOCK(s);
1271
1272	if (reiserfs_attrs(s)) {
1273		if (old_format_only(s)) {
1274			reiserfs_warning(s, "super-6517", "cannot support "
1275					 "attributes on 3.5.x disk format");
1276			REISERFS_SB(s)->s_mount_opt &= ~(1 << REISERFS_ATTRS);
1277			return;
1278		}
1279		if (!(le32_to_cpu(rs->s_flags) & reiserfs_attrs_cleared)) {
1280			reiserfs_warning(s, "super-6518", "cannot support "
1281					 "attributes until flag is set in "
1282					 "super-block");
1283			REISERFS_SB(s)->s_mount_opt &= ~(1 << REISERFS_ATTRS);
1284		}
1285	}
1286}
1287
1288#ifdef CONFIG_QUOTA
1289static void handle_quota_files(struct super_block *s, char **qf_names,
1290			       unsigned int *qfmt)
1291{
1292	int i;
1293
1294	for (i = 0; i < MAXQUOTAS; i++) {
1295		if (qf_names[i] != REISERFS_SB(s)->s_qf_names[i])
1296			kfree(REISERFS_SB(s)->s_qf_names[i]);
1297		REISERFS_SB(s)->s_qf_names[i] = qf_names[i];
1298	}
1299	if (*qfmt)
1300		REISERFS_SB(s)->s_jquota_fmt = *qfmt;
1301}
1302#endif
1303
1304static int reiserfs_remount(struct super_block *s, int *mount_flags, char *arg)
1305{
1306	struct reiserfs_super_block *rs;
1307	struct reiserfs_transaction_handle th;
1308	unsigned long blocks;
1309	unsigned long mount_options = REISERFS_SB(s)->s_mount_opt;
1310	unsigned long safe_mask = 0;
1311	unsigned int commit_max_age = (unsigned int)-1;
1312	struct reiserfs_journal *journal = SB_JOURNAL(s);
1313	char *new_opts = kstrdup(arg, GFP_KERNEL);
1314	int err;
1315	char *qf_names[MAXQUOTAS];
1316	unsigned int qfmt = 0;
1317#ifdef CONFIG_QUOTA
1318	int i;
1319#endif
1320
1321	reiserfs_write_lock(s);
1322
1323#ifdef CONFIG_QUOTA
1324	memcpy(qf_names, REISERFS_SB(s)->s_qf_names, sizeof(qf_names));
1325#endif
1326
1327	rs = SB_DISK_SUPER_BLOCK(s);
1328
1329	if (!reiserfs_parse_options
1330	    (s, arg, &mount_options, &blocks, NULL, &commit_max_age,
1331	    qf_names, &qfmt)) {
1332#ifdef CONFIG_QUOTA
1333		for (i = 0; i < MAXQUOTAS; i++)
1334			if (qf_names[i] != REISERFS_SB(s)->s_qf_names[i])
1335				kfree(qf_names[i]);
1336#endif
1337		err = -EINVAL;
1338		goto out_unlock;
1339	}
1340#ifdef CONFIG_QUOTA
1341	handle_quota_files(s, qf_names, &qfmt);
1342#endif
1343
1344	handle_attrs(s);
1345
1346	/* Add options that are safe here */
1347	safe_mask |= 1 << REISERFS_SMALLTAIL;
1348	safe_mask |= 1 << REISERFS_LARGETAIL;
1349	safe_mask |= 1 << REISERFS_NO_BORDER;
1350	safe_mask |= 1 << REISERFS_NO_UNHASHED_RELOCATION;
1351	safe_mask |= 1 << REISERFS_HASHED_RELOCATION;
1352	safe_mask |= 1 << REISERFS_TEST4;
1353	safe_mask |= 1 << REISERFS_ATTRS;
1354	safe_mask |= 1 << REISERFS_XATTRS_USER;
1355	safe_mask |= 1 << REISERFS_POSIXACL;
1356	safe_mask |= 1 << REISERFS_BARRIER_FLUSH;
1357	safe_mask |= 1 << REISERFS_BARRIER_NONE;
1358	safe_mask |= 1 << REISERFS_ERROR_RO;
1359	safe_mask |= 1 << REISERFS_ERROR_CONTINUE;
1360	safe_mask |= 1 << REISERFS_ERROR_PANIC;
1361	safe_mask |= 1 << REISERFS_USRQUOTA;
1362	safe_mask |= 1 << REISERFS_GRPQUOTA;
1363
1364	/* Update the bitmask, taking care to keep
1365	 * the bits we're not allowed to change here */
1366	REISERFS_SB(s)->s_mount_opt =
1367	    (REISERFS_SB(s)->
1368	     s_mount_opt & ~safe_mask) | (mount_options & safe_mask);
1369
1370	if (commit_max_age != 0 && commit_max_age != (unsigned int)-1) {
1371		journal->j_max_commit_age = commit_max_age;
1372		journal->j_max_trans_age = commit_max_age;
1373	} else if (commit_max_age == 0) {
1374		/* 0 means restore defaults. */
1375		journal->j_max_commit_age = journal->j_default_max_commit_age;
1376		journal->j_max_trans_age = JOURNAL_MAX_TRANS_AGE;
1377	}
1378
1379	if (blocks) {
1380		err = reiserfs_resize(s, blocks);
1381		if (err != 0)
1382			goto out_unlock;
1383	}
1384
1385	if (*mount_flags & MS_RDONLY) {
1386		reiserfs_xattr_init(s, *mount_flags);
1387		/* remount read-only */
1388		if (s->s_flags & MS_RDONLY)
1389			/* it is read-only already */
1390			goto out_ok;
1391
1392		/*
1393		 * Drop write lock. Quota will retake it when needed and lock
1394		 * ordering requires calling dquot_suspend() without it.
1395		 */
1396		reiserfs_write_unlock(s);
1397		err = dquot_suspend(s, -1);
1398		if (err < 0)
1399			goto out_err;
1400		reiserfs_write_lock(s);
1401
1402		/* try to remount file system with read-only permissions */
1403		if (sb_umount_state(rs) == REISERFS_VALID_FS
1404		    || REISERFS_SB(s)->s_mount_state != REISERFS_VALID_FS) {
1405			goto out_ok;
1406		}
1407
1408		err = journal_begin(&th, s, 10);
1409		if (err)
1410			goto out_unlock;
1411
1412		/* Mounting a rw partition read-only. */
1413		reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 1);
1414		set_sb_umount_state(rs, REISERFS_SB(s)->s_mount_state);
1415		journal_mark_dirty(&th, s, SB_BUFFER_WITH_SB(s));
1416	} else {
1417		/* remount read-write */
1418		if (!(s->s_flags & MS_RDONLY)) {
1419			reiserfs_xattr_init(s, *mount_flags);
1420			goto out_ok;	/* We are read-write already */
1421		}
1422
1423		if (reiserfs_is_journal_aborted(journal)) {
1424			err = journal->j_errno;
1425			goto out_unlock;
1426		}
1427
1428		handle_data_mode(s, mount_options);
1429		handle_barrier_mode(s, mount_options);
1430		REISERFS_SB(s)->s_mount_state = sb_umount_state(rs);
1431		s->s_flags &= ~MS_RDONLY;	/* now it is safe to call journal_begin */
1432		err = journal_begin(&th, s, 10);
1433		if (err)
1434			goto out_unlock;
1435
1436		/* Mount a partition which is read-only, read-write */
1437		reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 1);
1438		REISERFS_SB(s)->s_mount_state = sb_umount_state(rs);
1439		s->s_flags &= ~MS_RDONLY;
1440		set_sb_umount_state(rs, REISERFS_ERROR_FS);
1441		if (!old_format_only(s))
1442			set_sb_mnt_count(rs, sb_mnt_count(rs) + 1);
1443		/* mark_buffer_dirty (SB_BUFFER_WITH_SB (s), 1); */
1444		journal_mark_dirty(&th, s, SB_BUFFER_WITH_SB(s));
1445		REISERFS_SB(s)->s_mount_state = REISERFS_VALID_FS;
1446	}
1447	/* this will force a full flush of all journal lists */
1448	SB_JOURNAL(s)->j_must_wait = 1;
1449	err = journal_end(&th, s, 10);
1450	if (err)
1451		goto out_unlock;
1452
1453	if (!(*mount_flags & MS_RDONLY)) {
1454		/*
1455		 * Drop write lock. Quota will retake it when needed and lock
1456		 * ordering requires calling dquot_resume() without it.
1457		 */
1458		reiserfs_write_unlock(s);
1459		dquot_resume(s, -1);
1460		reiserfs_write_lock(s);
1461		finish_unfinished(s);
1462		reiserfs_xattr_init(s, *mount_flags);
1463	}
1464
1465out_ok:
1466	replace_mount_options(s, new_opts);
1467	reiserfs_write_unlock(s);
1468	return 0;
1469
1470out_unlock:
1471	reiserfs_write_unlock(s);
1472out_err:
1473	kfree(new_opts);
1474	return err;
1475}
1476
1477static int read_super_block(struct super_block *s, int offset)
1478{
1479	struct buffer_head *bh;
1480	struct reiserfs_super_block *rs;
1481	int fs_blocksize;
1482
1483	bh = sb_bread(s, offset / s->s_blocksize);
1484	if (!bh) {
1485		reiserfs_warning(s, "sh-2006",
1486				 "bread failed (dev %s, block %lu, size %lu)",
1487				 reiserfs_bdevname(s), offset / s->s_blocksize,
1488				 s->s_blocksize);
1489		return 1;
1490	}
1491
1492	rs = (struct reiserfs_super_block *)bh->b_data;
1493	if (!is_any_reiserfs_magic_string(rs)) {
1494		brelse(bh);
1495		return 1;
1496	}
1497	//
1498	// ok, reiserfs signature (old or new) found in at the given offset
1499	//
1500	fs_blocksize = sb_blocksize(rs);
1501	brelse(bh);
1502	sb_set_blocksize(s, fs_blocksize);
1503
1504	bh = sb_bread(s, offset / s->s_blocksize);
1505	if (!bh) {
1506		reiserfs_warning(s, "sh-2007",
1507				 "bread failed (dev %s, block %lu, size %lu)",
1508				 reiserfs_bdevname(s), offset / s->s_blocksize,
1509				 s->s_blocksize);
1510		return 1;
1511	}
1512
1513	rs = (struct reiserfs_super_block *)bh->b_data;
1514	if (sb_blocksize(rs) != s->s_blocksize) {
1515		reiserfs_warning(s, "sh-2011", "can't find a reiserfs "
1516				 "filesystem on (dev %s, block %Lu, size %lu)",
1517				 reiserfs_bdevname(s),
1518				 (unsigned long long)bh->b_blocknr,
1519				 s->s_blocksize);
1520		brelse(bh);
1521		return 1;
1522	}
1523
1524	if (rs->s_v1.s_root_block == cpu_to_le32(-1)) {
1525		brelse(bh);
1526		reiserfs_warning(s, "super-6519", "Unfinished reiserfsck "
1527				 "--rebuild-tree run detected. Please run\n"
1528				 "reiserfsck --rebuild-tree and wait for a "
1529				 "completion. If that fails\n"
1530				 "get newer reiserfsprogs package");
1531		return 1;
1532	}
1533
1534	SB_BUFFER_WITH_SB(s) = bh;
1535	SB_DISK_SUPER_BLOCK(s) = rs;
1536
1537	if (is_reiserfs_jr(rs)) {
1538		/* magic is of non-standard journal filesystem, look at s_version to
1539		   find which format is in use */
1540		if (sb_version(rs) == REISERFS_VERSION_2)
1541			reiserfs_info(s, "found reiserfs format \"3.6\""
1542				      " with non-standard journal\n");
1543		else if (sb_version(rs) == REISERFS_VERSION_1)
1544			reiserfs_info(s, "found reiserfs format \"3.5\""
1545				      " with non-standard journal\n");
1546		else {
1547			reiserfs_warning(s, "sh-2012", "found unknown "
1548					 "format \"%u\" of reiserfs with "
1549					 "non-standard magic", sb_version(rs));
1550			return 1;
1551		}
1552	} else
1553		/* s_version of standard format may contain incorrect information,
1554		   so we just look at the magic string */
1555		reiserfs_info(s,
1556			      "found reiserfs format \"%s\" with standard journal\n",
1557			      is_reiserfs_3_5(rs) ? "3.5" : "3.6");
1558
1559	s->s_op = &reiserfs_sops;
1560	s->s_export_op = &reiserfs_export_ops;
1561#ifdef CONFIG_QUOTA
1562	s->s_qcop = &reiserfs_qctl_operations;
1563	s->dq_op = &reiserfs_quota_operations;
1564#endif
1565
1566	/* new format is limited by the 32 bit wide i_blocks field, want to
1567	 ** be one full block below that.
1568	 */
1569	s->s_maxbytes = (512LL << 32) - s->s_blocksize;
1570	return 0;
1571}
1572
1573/* after journal replay, reread all bitmap and super blocks */
1574static int reread_meta_blocks(struct super_block *s)
1575{
1576	ll_rw_block(READ, 1, &(SB_BUFFER_WITH_SB(s)));
1577	wait_on_buffer(SB_BUFFER_WITH_SB(s));
1578	if (!buffer_uptodate(SB_BUFFER_WITH_SB(s))) {
1579		reiserfs_warning(s, "reiserfs-2504", "error reading the super");
1580		return 1;
1581	}
1582
1583	return 0;
1584}
1585
1586/////////////////////////////////////////////////////
1587// hash detection stuff
1588
1589// if root directory is empty - we set default - Yura's - hash and
1590// warn about it
1591// FIXME: we look for only one name in a directory. If tea and yura
1592// bith have the same value - we ask user to send report to the
1593// mailing list
1594static __u32 find_hash_out(struct super_block *s)
1595{
1596	int retval;
1597	struct inode *inode;
1598	struct cpu_key key;
1599	INITIALIZE_PATH(path);
1600	struct reiserfs_dir_entry de;
1601	__u32 hash = DEFAULT_HASH;
1602
1603	inode = s->s_root->d_inode;
1604
1605	do {			// Some serious "goto"-hater was there ;)
1606		u32 teahash, r5hash, yurahash;
1607
1608		make_cpu_key(&key, inode, ~0, TYPE_DIRENTRY, 3);
1609		retval = search_by_entry_key(s, &key, &path, &de);
1610		if (retval == IO_ERROR) {
1611			pathrelse(&path);
1612			return UNSET_HASH;
1613		}
1614		if (retval == NAME_NOT_FOUND)
1615			de.de_entry_num--;
1616		set_de_name_and_namelen(&de);
1617		if (deh_offset(&(de.de_deh[de.de_entry_num])) == DOT_DOT_OFFSET) {
1618			/* allow override in this case */
1619			if (reiserfs_rupasov_hash(s)) {
1620				hash = YURA_HASH;
1621			}
1622			reiserfs_info(s, "FS seems to be empty, autodetect "
1623					 "is using the default hash\n");
1624			break;
1625		}
1626		r5hash = GET_HASH_VALUE(r5_hash(de.de_name, de.de_namelen));
1627		teahash = GET_HASH_VALUE(keyed_hash(de.de_name, de.de_namelen));
1628		yurahash = GET_HASH_VALUE(yura_hash(de.de_name, de.de_namelen));
1629		if (((teahash == r5hash)
1630		     &&
1631		     (GET_HASH_VALUE(deh_offset(&(de.de_deh[de.de_entry_num])))
1632		      == r5hash)) || ((teahash == yurahash)
1633				      && (yurahash ==
1634					  GET_HASH_VALUE(deh_offset
1635							 (&
1636							  (de.
1637							   de_deh[de.
1638								  de_entry_num])))))
1639		    || ((r5hash == yurahash)
1640			&& (yurahash ==
1641			    GET_HASH_VALUE(deh_offset
1642					   (&(de.de_deh[de.de_entry_num])))))) {
1643			reiserfs_warning(s, "reiserfs-2506", "Unable to "
1644					 "automatically detect hash function. "
1645					 "Please mount with -o "
1646					 "hash={tea,rupasov,r5}");
1647			hash = UNSET_HASH;
1648			break;
1649		}
1650		if (GET_HASH_VALUE(deh_offset(&(de.de_deh[de.de_entry_num]))) ==
1651		    yurahash)
1652			hash = YURA_HASH;
1653		else if (GET_HASH_VALUE
1654			 (deh_offset(&(de.de_deh[de.de_entry_num]))) == teahash)
1655			hash = TEA_HASH;
1656		else if (GET_HASH_VALUE
1657			 (deh_offset(&(de.de_deh[de.de_entry_num]))) == r5hash)
1658			hash = R5_HASH;
1659		else {
1660			reiserfs_warning(s, "reiserfs-2506",
1661					 "Unrecognised hash function");
1662			hash = UNSET_HASH;
1663		}
1664	} while (0);
1665
1666	pathrelse(&path);
1667	return hash;
1668}
1669
1670// finds out which hash names are sorted with
1671static int what_hash(struct super_block *s)
1672{
1673	__u32 code;
1674
1675	code = sb_hash_function_code(SB_DISK_SUPER_BLOCK(s));
1676
1677	/* reiserfs_hash_detect() == true if any of the hash mount options
1678	 ** were used.  We must check them to make sure the user isn't
1679	 ** using a bad hash value
1680	 */
1681	if (code == UNSET_HASH || reiserfs_hash_detect(s))
1682		code = find_hash_out(s);
1683
1684	if (code != UNSET_HASH && reiserfs_hash_detect(s)) {
1685		/* detection has found the hash, and we must check against the
1686		 ** mount options
1687		 */
1688		if (reiserfs_rupasov_hash(s) && code != YURA_HASH) {
1689			reiserfs_warning(s, "reiserfs-2507",
1690					 "Error, %s hash detected, "
1691					 "unable to force rupasov hash",
1692					 reiserfs_hashname(code));
1693			code = UNSET_HASH;
1694		} else if (reiserfs_tea_hash(s) && code != TEA_HASH) {
1695			reiserfs_warning(s, "reiserfs-2508",
1696					 "Error, %s hash detected, "
1697					 "unable to force tea hash",
1698					 reiserfs_hashname(code));
1699			code = UNSET_HASH;
1700		} else if (reiserfs_r5_hash(s) && code != R5_HASH) {
1701			reiserfs_warning(s, "reiserfs-2509",
1702					 "Error, %s hash detected, "
1703					 "unable to force r5 hash",
1704					 reiserfs_hashname(code));
1705			code = UNSET_HASH;
1706		}
1707	} else {
1708		/* find_hash_out was not called or could not determine the hash */
1709		if (reiserfs_rupasov_hash(s)) {
1710			code = YURA_HASH;
1711		} else if (reiserfs_tea_hash(s)) {
1712			code = TEA_HASH;
1713		} else if (reiserfs_r5_hash(s)) {
1714			code = R5_HASH;
1715		}
1716	}
1717
1718	/* if we are mounted RW, and we have a new valid hash code, update
1719	 ** the super
1720	 */
1721	if (code != UNSET_HASH &&
1722	    !(s->s_flags & MS_RDONLY) &&
1723	    code != sb_hash_function_code(SB_DISK_SUPER_BLOCK(s))) {
1724		set_sb_hash_function_code(SB_DISK_SUPER_BLOCK(s), code);
1725	}
1726	return code;
1727}
1728
1729// return pointer to appropriate function
1730static hashf_t hash_function(struct super_block *s)
1731{
1732	switch (what_hash(s)) {
1733	case TEA_HASH:
1734		reiserfs_info(s, "Using tea hash to sort names\n");
1735		return keyed_hash;
1736	case YURA_HASH:
1737		reiserfs_info(s, "Using rupasov hash to sort names\n");
1738		return yura_hash;
1739	case R5_HASH:
1740		reiserfs_info(s, "Using r5 hash to sort names\n");
1741		return r5_hash;
1742	}
1743	return NULL;
1744}
1745
1746// this is used to set up correct value for old partitions
1747static int function2code(hashf_t func)
1748{
1749	if (func == keyed_hash)
1750		return TEA_HASH;
1751	if (func == yura_hash)
1752		return YURA_HASH;
1753	if (func == r5_hash)
1754		return R5_HASH;
1755
1756	BUG();			// should never happen
1757
1758	return 0;
1759}
1760
1761#define SWARN(silent, s, id, ...)			\
1762	if (!(silent))				\
1763		reiserfs_warning(s, id, __VA_ARGS__)
1764
1765static int reiserfs_fill_super(struct super_block *s, void *data, int silent)
1766{
1767	struct inode *root_inode;
1768	struct reiserfs_transaction_handle th;
1769	int old_format = 0;
1770	unsigned long blocks;
1771	unsigned int commit_max_age = 0;
1772	int jinit_done = 0;
1773	struct reiserfs_iget_args args;
1774	struct reiserfs_super_block *rs;
1775	char *jdev_name;
1776	struct reiserfs_sb_info *sbi;
1777	int errval = -EINVAL;
1778	char *qf_names[MAXQUOTAS] = {};
1779	unsigned int qfmt = 0;
1780
1781	save_mount_options(s, data);
1782
1783	sbi = kzalloc(sizeof(struct reiserfs_sb_info), GFP_KERNEL);
1784	if (!sbi)
1785		return -ENOMEM;
1786	s->s_fs_info = sbi;
1787	/* Set default values for options: non-aggressive tails, RO on errors */
1788	sbi->s_mount_opt |= (1 << REISERFS_SMALLTAIL);
1789	sbi->s_mount_opt |= (1 << REISERFS_ERROR_RO);
1790	sbi->s_mount_opt |= (1 << REISERFS_BARRIER_FLUSH);
1791	/* no preallocation minimum, be smart in
1792	   reiserfs_file_write instead */
1793	sbi->s_alloc_options.preallocmin = 0;
1794	/* Preallocate by 16 blocks (17-1) at once */
1795	sbi->s_alloc_options.preallocsize = 17;
1796	/* setup default block allocator options */
1797	reiserfs_init_alloc_options(s);
1798
1799	spin_lock_init(&sbi->old_work_lock);
1800	INIT_DELAYED_WORK(&sbi->old_work, flush_old_commits);
1801	mutex_init(&sbi->lock);
1802	sbi->lock_depth = -1;
1803
1804	jdev_name = NULL;
1805	if (reiserfs_parse_options
1806	    (s, (char *)data, &(sbi->s_mount_opt), &blocks, &jdev_name,
1807	     &commit_max_age, qf_names, &qfmt) == 0) {
1808		goto error_unlocked;
1809	}
1810	if (jdev_name && jdev_name[0]) {
1811		sbi->s_jdev = kstrdup(jdev_name, GFP_KERNEL);
1812		if (!sbi->s_jdev) {
1813			SWARN(silent, s, "", "Cannot allocate memory for "
1814				"journal device name");
1815			goto error;
1816		}
1817	}
1818#ifdef CONFIG_QUOTA
1819	handle_quota_files(s, qf_names, &qfmt);
1820#endif
1821
1822	if (blocks) {
1823		SWARN(silent, s, "jmacd-7", "resize option for remount only");
1824		goto error_unlocked;
1825	}
1826
1827	/* try old format (undistributed bitmap, super block in 8-th 1k block of a device) */
1828	if (!read_super_block(s, REISERFS_OLD_DISK_OFFSET_IN_BYTES))
1829		old_format = 1;
1830	/* try new format (64-th 1k block), which can contain reiserfs super block */
1831	else if (read_super_block(s, REISERFS_DISK_OFFSET_IN_BYTES)) {
1832		SWARN(silent, s, "sh-2021", "can not find reiserfs on %s",
1833		      reiserfs_bdevname(s));
1834		goto error_unlocked;
1835	}
1836
1837	rs = SB_DISK_SUPER_BLOCK(s);
1838	/* Let's do basic sanity check to verify that underlying device is not
1839	   smaller than the filesystem. If the check fails then abort and scream,
1840	   because bad stuff will happen otherwise. */
1841	if (s->s_bdev && s->s_bdev->bd_inode
1842	    && i_size_read(s->s_bdev->bd_inode) <
1843	    sb_block_count(rs) * sb_blocksize(rs)) {
1844		SWARN(silent, s, "", "Filesystem cannot be "
1845		      "mounted because it is bigger than the device");
1846		SWARN(silent, s, "", "You may need to run fsck "
1847		      "or increase size of your LVM partition");
1848		SWARN(silent, s, "", "Or may be you forgot to "
1849		      "reboot after fdisk when it told you to");
1850		goto error_unlocked;
1851	}
1852
1853	sbi->s_mount_state = SB_REISERFS_STATE(s);
1854	sbi->s_mount_state = REISERFS_VALID_FS;
1855
1856	if ((errval = reiserfs_init_bitmap_cache(s))) {
1857		SWARN(silent, s, "jmacd-8", "unable to read bitmap");
1858		goto error_unlocked;
1859	}
1860
1861	errval = -EINVAL;
1862#ifdef CONFIG_REISERFS_CHECK
1863	SWARN(silent, s, "", "CONFIG_REISERFS_CHECK is set ON");
1864	SWARN(silent, s, "", "- it is slow mode for debugging.");
1865#endif
1866
1867	/* make data=ordered the default */
1868	if (!reiserfs_data_log(s) && !reiserfs_data_ordered(s) &&
1869	    !reiserfs_data_writeback(s)) {
1870		sbi->s_mount_opt |= (1 << REISERFS_DATA_ORDERED);
1871	}
1872
1873	if (reiserfs_data_log(s)) {
1874		reiserfs_info(s, "using journaled data mode\n");
1875	} else if (reiserfs_data_ordered(s)) {
1876		reiserfs_info(s, "using ordered data mode\n");
1877	} else {
1878		reiserfs_info(s, "using writeback data mode\n");
1879	}
1880	if (reiserfs_barrier_flush(s)) {
1881		printk("reiserfs: using flush barriers\n");
1882	}
1883
1884	// set_device_ro(s->s_dev, 1) ;
1885	if (journal_init(s, jdev_name, old_format, commit_max_age)) {
1886		SWARN(silent, s, "sh-2022",
1887		      "unable to initialize journal space");
1888		goto error_unlocked;
1889	} else {
1890		jinit_done = 1;	/* once this is set, journal_release must be called
1891				 ** if we error out of the mount
1892				 */
1893	}
1894
1895	if (reread_meta_blocks(s)) {
1896		SWARN(silent, s, "jmacd-9",
1897		      "unable to reread meta blocks after journal init");
1898		goto error_unlocked;
1899	}
1900
1901	if (replay_only(s))
1902		goto error_unlocked;
1903
1904	if (bdev_read_only(s->s_bdev) && !(s->s_flags & MS_RDONLY)) {
1905		SWARN(silent, s, "clm-7000",
1906		      "Detected readonly device, marking FS readonly");
1907		s->s_flags |= MS_RDONLY;
1908	}
1909	args.objectid = REISERFS_ROOT_OBJECTID;
1910	args.dirid = REISERFS_ROOT_PARENT_OBJECTID;
1911	root_inode =
1912	    iget5_locked(s, REISERFS_ROOT_OBJECTID, reiserfs_find_actor,
1913			 reiserfs_init_locked_inode, (void *)(&args));
1914	if (!root_inode) {
1915		SWARN(silent, s, "jmacd-10", "get root inode failed");
1916		goto error_unlocked;
1917	}
1918
1919	/*
1920	 * This path assumed to be called with the BKL in the old times.
1921	 * Now we have inherited the big reiserfs lock from it and many
1922	 * reiserfs helpers called in the mount path and elsewhere require
1923	 * this lock to be held even if it's not always necessary. Let's be
1924	 * conservative and hold it early. The window can be reduced after
1925	 * careful review of the code.
1926	 */
1927	reiserfs_write_lock(s);
1928
1929	if (root_inode->i_state & I_NEW) {
1930		reiserfs_read_locked_inode(root_inode, &args);
1931		unlock_new_inode(root_inode);
1932	}
1933
1934	s->s_root = d_make_root(root_inode);
1935	if (!s->s_root)
1936		goto error;
1937	// define and initialize hash function
1938	sbi->s_hash_function = hash_function(s);
1939	if (sbi->s_hash_function == NULL) {
1940		dput(s->s_root);
1941		s->s_root = NULL;
1942		goto error;
1943	}
1944
1945	if (is_reiserfs_3_5(rs)
1946	    || (is_reiserfs_jr(rs) && SB_VERSION(s) == REISERFS_VERSION_1))
1947		set_bit(REISERFS_3_5, &(sbi->s_properties));
1948	else if (old_format)
1949		set_bit(REISERFS_OLD_FORMAT, &(sbi->s_properties));
1950	else
1951		set_bit(REISERFS_3_6, &(sbi->s_properties));
1952
1953	if (!(s->s_flags & MS_RDONLY)) {
1954
1955		errval = journal_begin(&th, s, 1);
1956		if (errval) {
1957			dput(s->s_root);
1958			s->s_root = NULL;
1959			goto error;
1960		}
1961		reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 1);
1962
1963		set_sb_umount_state(rs, REISERFS_ERROR_FS);
1964		set_sb_fs_state(rs, 0);
1965
1966		/* Clear out s_bmap_nr if it would wrap. We can handle this
1967		 * case, but older revisions can't. This will cause the
1968		 * file system to fail mount on those older implementations,
1969		 * avoiding corruption. -jeffm */
1970		if (bmap_would_wrap(reiserfs_bmap_count(s)) &&
1971		    sb_bmap_nr(rs) != 0) {
1972			reiserfs_warning(s, "super-2030", "This file system "
1973					"claims to use %u bitmap blocks in "
1974					"its super block, but requires %u. "
1975					"Clearing to zero.", sb_bmap_nr(rs),
1976					reiserfs_bmap_count(s));
1977
1978			set_sb_bmap_nr(rs, 0);
1979		}
1980
1981		if (old_format_only(s)) {
1982			/* filesystem of format 3.5 either with standard or non-standard
1983			   journal */
1984			if (convert_reiserfs(s)) {
1985				/* and -o conv is given */
1986				if (!silent)
1987					reiserfs_info(s,
1988						      "converting 3.5 filesystem to the 3.6 format");
1989
1990				if (is_reiserfs_3_5(rs))
1991					/* put magic string of 3.6 format. 2.2 will not be able to
1992					   mount this filesystem anymore */
1993					memcpy(rs->s_v1.s_magic,
1994					       reiserfs_3_6_magic_string,
1995					       sizeof
1996					       (reiserfs_3_6_magic_string));
1997
1998				set_sb_version(rs, REISERFS_VERSION_2);
1999				reiserfs_convert_objectid_map_v1(s);
2000				set_bit(REISERFS_3_6, &(sbi->s_properties));
2001				clear_bit(REISERFS_3_5, &(sbi->s_properties));
2002			} else if (!silent) {
2003				reiserfs_info(s, "using 3.5.x disk format\n");
2004			}
2005		} else
2006			set_sb_mnt_count(rs, sb_mnt_count(rs) + 1);
2007
2008
2009		journal_mark_dirty(&th, s, SB_BUFFER_WITH_SB(s));
2010		errval = journal_end(&th, s, 1);
2011		if (errval) {
2012			dput(s->s_root);
2013			s->s_root = NULL;
2014			goto error;
2015		}
2016
2017		if ((errval = reiserfs_lookup_privroot(s)) ||
2018		    (errval = reiserfs_xattr_init(s, s->s_flags))) {
2019			dput(s->s_root);
2020			s->s_root = NULL;
2021			goto error;
2022		}
2023
2024		/* look for files which were to be removed in previous session */
2025		finish_unfinished(s);
2026	} else {
2027		if (old_format_only(s) && !silent) {
2028			reiserfs_info(s, "using 3.5.x disk format\n");
2029		}
2030
2031		if ((errval = reiserfs_lookup_privroot(s)) ||
2032		    (errval = reiserfs_xattr_init(s, s->s_flags))) {
2033			dput(s->s_root);
2034			s->s_root = NULL;
2035			goto error;
2036		}
2037	}
2038	// mark hash in super block: it could be unset. overwrite should be ok
2039	set_sb_hash_function_code(rs, function2code(sbi->s_hash_function));
2040
2041	handle_attrs(s);
2042
2043	reiserfs_proc_info_init(s);
2044
2045	init_waitqueue_head(&(sbi->s_wait));
2046	spin_lock_init(&sbi->bitmap_lock);
2047
2048	reiserfs_write_unlock(s);
2049
2050	return (0);
2051
2052error:
2053	reiserfs_write_unlock(s);
2054
2055error_unlocked:
2056	/* kill the commit thread, free journal ram */
2057	if (jinit_done) {
2058		reiserfs_write_lock(s);
2059		journal_release_error(NULL, s);
2060		reiserfs_write_unlock(s);
2061	}
2062
2063	cancel_delayed_work_sync(&REISERFS_SB(s)->old_work);
2064
2065	reiserfs_free_bitmap_cache(s);
2066	if (SB_BUFFER_WITH_SB(s))
2067		brelse(SB_BUFFER_WITH_SB(s));
2068#ifdef CONFIG_QUOTA
2069	{
2070		int j;
2071		for (j = 0; j < MAXQUOTAS; j++)
2072			kfree(qf_names[j]);
2073	}
2074#endif
2075	kfree(sbi);
2076
2077	s->s_fs_info = NULL;
2078	return errval;
2079}
2080
2081static int reiserfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2082{
2083	struct reiserfs_super_block *rs = SB_DISK_SUPER_BLOCK(dentry->d_sb);
2084
2085	buf->f_namelen = (REISERFS_MAX_NAME(s->s_blocksize));
2086	buf->f_bfree = sb_free_blocks(rs);
2087	buf->f_bavail = buf->f_bfree;
2088	buf->f_blocks = sb_block_count(rs) - sb_bmap_nr(rs) - 1;
2089	buf->f_bsize = dentry->d_sb->s_blocksize;
2090	/* changed to accommodate gcc folks. */
2091	buf->f_type = REISERFS_SUPER_MAGIC;
2092	buf->f_fsid.val[0] = (u32)crc32_le(0, rs->s_uuid, sizeof(rs->s_uuid)/2);
2093	buf->f_fsid.val[1] = (u32)crc32_le(0, rs->s_uuid + sizeof(rs->s_uuid)/2,
2094				sizeof(rs->s_uuid)/2);
2095
2096	return 0;
2097}
2098
2099#ifdef CONFIG_QUOTA
2100static int reiserfs_write_dquot(struct dquot *dquot)
2101{
2102	struct reiserfs_transaction_handle th;
2103	int ret, err;
2104
2105	reiserfs_write_lock(dquot->dq_sb);
2106	ret =
2107	    journal_begin(&th, dquot->dq_sb,
2108			  REISERFS_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2109	if (ret)
2110		goto out;
2111	reiserfs_write_unlock(dquot->dq_sb);
2112	ret = dquot_commit(dquot);
2113	reiserfs_write_lock(dquot->dq_sb);
2114	err =
2115	    journal_end(&th, dquot->dq_sb,
2116			REISERFS_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2117	if (!ret && err)
2118		ret = err;
2119out:
2120	reiserfs_write_unlock(dquot->dq_sb);
2121	return ret;
2122}
2123
2124static int reiserfs_acquire_dquot(struct dquot *dquot)
2125{
2126	struct reiserfs_transaction_handle th;
2127	int ret, err;
2128
2129	reiserfs_write_lock(dquot->dq_sb);
2130	ret =
2131	    journal_begin(&th, dquot->dq_sb,
2132			  REISERFS_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2133	if (ret)
2134		goto out;
2135	reiserfs_write_unlock(dquot->dq_sb);
2136	ret = dquot_acquire(dquot);
2137	reiserfs_write_lock(dquot->dq_sb);
2138	err =
2139	    journal_end(&th, dquot->dq_sb,
2140			REISERFS_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2141	if (!ret && err)
2142		ret = err;
2143out:
2144	reiserfs_write_unlock(dquot->dq_sb);
2145	return ret;
2146}
2147
2148static int reiserfs_release_dquot(struct dquot *dquot)
2149{
2150	struct reiserfs_transaction_handle th;
2151	int ret, err;
2152
2153	reiserfs_write_lock(dquot->dq_sb);
2154	ret =
2155	    journal_begin(&th, dquot->dq_sb,
2156			  REISERFS_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2157	reiserfs_write_unlock(dquot->dq_sb);
2158	if (ret) {
2159		/* Release dquot anyway to avoid endless cycle in dqput() */
2160		dquot_release(dquot);
2161		goto out;
2162	}
2163	ret = dquot_release(dquot);
2164	reiserfs_write_lock(dquot->dq_sb);
2165	err =
2166	    journal_end(&th, dquot->dq_sb,
2167			REISERFS_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2168	if (!ret && err)
2169		ret = err;
2170	reiserfs_write_unlock(dquot->dq_sb);
2171out:
2172	return ret;
2173}
2174
2175static int reiserfs_mark_dquot_dirty(struct dquot *dquot)
2176{
2177	/* Are we journaling quotas? */
2178	if (REISERFS_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2179	    REISERFS_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2180		dquot_mark_dquot_dirty(dquot);
2181		return reiserfs_write_dquot(dquot);
2182	} else
2183		return dquot_mark_dquot_dirty(dquot);
2184}
2185
2186static int reiserfs_write_info(struct super_block *sb, int type)
2187{
2188	struct reiserfs_transaction_handle th;
2189	int ret, err;
2190
2191	/* Data block + inode block */
2192	reiserfs_write_lock(sb);
2193	ret = journal_begin(&th, sb, 2);
2194	if (ret)
2195		goto out;
2196	reiserfs_write_unlock(sb);
2197	ret = dquot_commit_info(sb, type);
2198	reiserfs_write_lock(sb);
2199	err = journal_end(&th, sb, 2);
2200	if (!ret && err)
2201		ret = err;
2202out:
2203	reiserfs_write_unlock(sb);
2204	return ret;
2205}
2206
2207/*
2208 * Turn on quotas during mount time - we need to find the quota file and such...
2209 */
2210static int reiserfs_quota_on_mount(struct super_block *sb, int type)
2211{
2212	return dquot_quota_on_mount(sb, REISERFS_SB(sb)->s_qf_names[type],
2213					REISERFS_SB(sb)->s_jquota_fmt, type);
2214}
2215
2216/*
2217 * Standard function to be called on quota_on
2218 */
2219static int reiserfs_quota_on(struct super_block *sb, int type, int format_id,
2220			     struct path *path)
2221{
2222	int err;
2223	struct inode *inode;
2224	struct reiserfs_transaction_handle th;
2225	int opt = type == USRQUOTA ? REISERFS_USRQUOTA : REISERFS_GRPQUOTA;
2226
2227	reiserfs_write_lock(sb);
2228	if (!(REISERFS_SB(sb)->s_mount_opt & (1 << opt))) {
2229		err = -EINVAL;
2230		goto out;
2231	}
2232
2233	/* Quotafile not on the same filesystem? */
2234	if (path->dentry->d_sb != sb) {
2235		err = -EXDEV;
2236		goto out;
2237	}
2238	inode = path->dentry->d_inode;
2239	/* We must not pack tails for quota files on reiserfs for quota IO to work */
2240	if (!(REISERFS_I(inode)->i_flags & i_nopack_mask)) {
2241		err = reiserfs_unpack(inode, NULL);
2242		if (err) {
2243			reiserfs_warning(sb, "super-6520",
2244				"Unpacking tail of quota file failed"
2245				" (%d). Cannot turn on quotas.", err);
2246			err = -EINVAL;
2247			goto out;
2248		}
2249		mark_inode_dirty(inode);
2250	}
2251	/* Journaling quota? */
2252	if (REISERFS_SB(sb)->s_qf_names[type]) {
2253		/* Quotafile not of fs root? */
2254		if (path->dentry->d_parent != sb->s_root)
2255			reiserfs_warning(sb, "super-6521",
2256				 "Quota file not on filesystem root. "
2257				 "Journalled quota will not work.");
2258	}
2259
2260	/*
2261	 * When we journal data on quota file, we have to flush journal to see
2262	 * all updates to the file when we bypass pagecache...
2263	 */
2264	if (reiserfs_file_data_log(inode)) {
2265		/* Just start temporary transaction and finish it */
2266		err = journal_begin(&th, sb, 1);
2267		if (err)
2268			goto out;
2269		err = journal_end_sync(&th, sb, 1);
2270		if (err)
2271			goto out;
2272	}
2273	reiserfs_write_unlock(sb);
2274	return dquot_quota_on(sb, type, format_id, path);
2275out:
2276	reiserfs_write_unlock(sb);
2277	return err;
2278}
2279
2280/* Read data from quotafile - avoid pagecache and such because we cannot afford
2281 * acquiring the locks... As quota files are never truncated and quota code
2282 * itself serializes the operations (and no one else should touch the files)
2283 * we don't have to be afraid of races */
2284static ssize_t reiserfs_quota_read(struct super_block *sb, int type, char *data,
2285				   size_t len, loff_t off)
2286{
2287	struct inode *inode = sb_dqopt(sb)->files[type];
2288	unsigned long blk = off >> sb->s_blocksize_bits;
2289	int err = 0, offset = off & (sb->s_blocksize - 1), tocopy;
2290	size_t toread;
2291	struct buffer_head tmp_bh, *bh;
2292	loff_t i_size = i_size_read(inode);
2293
2294	if (off > i_size)
2295		return 0;
2296	if (off + len > i_size)
2297		len = i_size - off;
2298	toread = len;
2299	while (toread > 0) {
2300		tocopy =
2301		    sb->s_blocksize - offset <
2302		    toread ? sb->s_blocksize - offset : toread;
2303		tmp_bh.b_state = 0;
2304		/* Quota files are without tails so we can safely use this function */
2305		reiserfs_write_lock(sb);
2306		err = reiserfs_get_block(inode, blk, &tmp_bh, 0);
2307		reiserfs_write_unlock(sb);
2308		if (err)
2309			return err;
2310		if (!buffer_mapped(&tmp_bh))	/* A hole? */
2311			memset(data, 0, tocopy);
2312		else {
2313			bh = sb_bread(sb, tmp_bh.b_blocknr);
2314			if (!bh)
2315				return -EIO;
2316			memcpy(data, bh->b_data + offset, tocopy);
2317			brelse(bh);
2318		}
2319		offset = 0;
2320		toread -= tocopy;
2321		data += tocopy;
2322		blk++;
2323	}
2324	return len;
2325}
2326
2327/* Write to quotafile (we know the transaction is already started and has
2328 * enough credits) */
2329static ssize_t reiserfs_quota_write(struct super_block *sb, int type,
2330				    const char *data, size_t len, loff_t off)
2331{
2332	struct inode *inode = sb_dqopt(sb)->files[type];
2333	unsigned long blk = off >> sb->s_blocksize_bits;
2334	int err = 0, offset = off & (sb->s_blocksize - 1), tocopy;
2335	int journal_quota = REISERFS_SB(sb)->s_qf_names[type] != NULL;
2336	size_t towrite = len;
2337	struct buffer_head tmp_bh, *bh;
2338
2339	if (!current->journal_info) {
2340		printk(KERN_WARNING "reiserfs: Quota write (off=%Lu, len=%Lu)"
2341			" cancelled because transaction is not started.\n",
2342			(unsigned long long)off, (unsigned long long)len);
2343		return -EIO;
2344	}
2345	while (towrite > 0) {
2346		tocopy = sb->s_blocksize - offset < towrite ?
2347		    sb->s_blocksize - offset : towrite;
2348		tmp_bh.b_state = 0;
2349		reiserfs_write_lock(sb);
2350		err = reiserfs_get_block(inode, blk, &tmp_bh, GET_BLOCK_CREATE);
2351		reiserfs_write_unlock(sb);
2352		if (err)
2353			goto out;
2354		if (offset || tocopy != sb->s_blocksize)
2355			bh = sb_bread(sb, tmp_bh.b_blocknr);
2356		else
2357			bh = sb_getblk(sb, tmp_bh.b_blocknr);
2358		if (!bh) {
2359			err = -EIO;
2360			goto out;
2361		}
2362		lock_buffer(bh);
2363		memcpy(bh->b_data + offset, data, tocopy);
2364		flush_dcache_page(bh->b_page);
2365		set_buffer_uptodate(bh);
2366		unlock_buffer(bh);
2367		reiserfs_write_lock(sb);
2368		reiserfs_prepare_for_journal(sb, bh, 1);
2369		journal_mark_dirty(current->journal_info, sb, bh);
2370		if (!journal_quota)
2371			reiserfs_add_ordered_list(inode, bh);
2372		reiserfs_write_unlock(sb);
2373		brelse(bh);
2374		offset = 0;
2375		towrite -= tocopy;
2376		data += tocopy;
2377		blk++;
2378	}
2379out:
2380	if (len == towrite)
2381		return err;
2382	if (inode->i_size < off + len - towrite)
2383		i_size_write(inode, off + len - towrite);
2384	inode->i_version++;
2385	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2386	mark_inode_dirty(inode);
2387	return len - towrite;
2388}
2389
2390#endif
2391
2392static struct dentry *get_super_block(struct file_system_type *fs_type,
2393			   int flags, const char *dev_name,
2394			   void *data)
2395{
2396	return mount_bdev(fs_type, flags, dev_name, data, reiserfs_fill_super);
2397}
2398
2399static int __init init_reiserfs_fs(void)
2400{
2401	int ret;
2402
2403	if ((ret = init_inodecache())) {
2404		return ret;
2405	}
2406
2407	reiserfs_proc_info_global_init();
2408
2409	ret = register_filesystem(&reiserfs_fs_type);
2410
2411	if (ret == 0) {
2412		return 0;
2413	}
2414
2415	reiserfs_proc_info_global_done();
2416	destroy_inodecache();
2417
2418	return ret;
2419}
2420
2421static void __exit exit_reiserfs_fs(void)
2422{
2423	reiserfs_proc_info_global_done();
2424	unregister_filesystem(&reiserfs_fs_type);
2425	destroy_inodecache();
2426}
2427
2428struct file_system_type reiserfs_fs_type = {
2429	.owner = THIS_MODULE,
2430	.name = "reiserfs",
2431	.mount = get_super_block,
2432	.kill_sb = reiserfs_kill_sb,
2433	.fs_flags = FS_REQUIRES_DEV,
2434};
2435MODULE_ALIAS_FS("reiserfs");
2436
2437MODULE_DESCRIPTION("ReiserFS journaled filesystem");
2438MODULE_AUTHOR("Hans Reiser <reiser@namesys.com>");
2439MODULE_LICENSE("GPL");
2440
2441module_init(init_reiserfs_fs);
2442module_exit(exit_reiserfs_fs);
2443