1/*
2 * Copyright (C) 2007 Oracle.  All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/err.h>
20#include <linux/uuid.h>
21#include "ctree.h"
22#include "transaction.h"
23#include "disk-io.h"
24#include "print-tree.h"
25
26/*
27 * Read a root item from the tree. In case we detect a root item smaller then
28 * sizeof(root_item), we know it's an old version of the root structure and
29 * initialize all new fields to zero. The same happens if we detect mismatching
30 * generation numbers as then we know the root was once mounted with an older
31 * kernel that was not aware of the root item structure change.
32 */
33static void btrfs_read_root_item(struct extent_buffer *eb, int slot,
34				struct btrfs_root_item *item)
35{
36	uuid_le uuid;
37	int len;
38	int need_reset = 0;
39
40	len = btrfs_item_size_nr(eb, slot);
41	read_extent_buffer(eb, item, btrfs_item_ptr_offset(eb, slot),
42			min_t(int, len, (int)sizeof(*item)));
43	if (len < sizeof(*item))
44		need_reset = 1;
45	if (!need_reset && btrfs_root_generation(item)
46		!= btrfs_root_generation_v2(item)) {
47		if (btrfs_root_generation_v2(item) != 0) {
48			printk(KERN_WARNING "BTRFS: mismatching "
49					"generation and generation_v2 "
50					"found in root item. This root "
51					"was probably mounted with an "
52					"older kernel. Resetting all "
53					"new fields.\n");
54		}
55		need_reset = 1;
56	}
57	if (need_reset) {
58		memset(&item->generation_v2, 0,
59			sizeof(*item) - offsetof(struct btrfs_root_item,
60					generation_v2));
61
62		uuid_le_gen(&uuid);
63		memcpy(item->uuid, uuid.b, BTRFS_UUID_SIZE);
64	}
65}
66
67/*
68 * btrfs_find_root - lookup the root by the key.
69 * root: the root of the root tree
70 * search_key: the key to search
71 * path: the path we search
72 * root_item: the root item of the tree we look for
73 * root_key: the reak key of the tree we look for
74 *
75 * If ->offset of 'seach_key' is -1ULL, it means we are not sure the offset
76 * of the search key, just lookup the root with the highest offset for a
77 * given objectid.
78 *
79 * If we find something return 0, otherwise > 0, < 0 on error.
80 */
81int btrfs_find_root(struct btrfs_root *root, struct btrfs_key *search_key,
82		    struct btrfs_path *path, struct btrfs_root_item *root_item,
83		    struct btrfs_key *root_key)
84{
85	struct btrfs_key found_key;
86	struct extent_buffer *l;
87	int ret;
88	int slot;
89
90	ret = btrfs_search_slot(NULL, root, search_key, path, 0, 0);
91	if (ret < 0)
92		return ret;
93
94	if (search_key->offset != -1ULL) {	/* the search key is exact */
95		if (ret > 0)
96			goto out;
97	} else {
98		BUG_ON(ret == 0);		/* Logical error */
99		if (path->slots[0] == 0)
100			goto out;
101		path->slots[0]--;
102		ret = 0;
103	}
104
105	l = path->nodes[0];
106	slot = path->slots[0];
107
108	btrfs_item_key_to_cpu(l, &found_key, slot);
109	if (found_key.objectid != search_key->objectid ||
110	    found_key.type != BTRFS_ROOT_ITEM_KEY) {
111		ret = 1;
112		goto out;
113	}
114
115	if (root_item)
116		btrfs_read_root_item(l, slot, root_item);
117	if (root_key)
118		memcpy(root_key, &found_key, sizeof(found_key));
119out:
120	btrfs_release_path(path);
121	return ret;
122}
123
124void btrfs_set_root_node(struct btrfs_root_item *item,
125			 struct extent_buffer *node)
126{
127	btrfs_set_root_bytenr(item, node->start);
128	btrfs_set_root_level(item, btrfs_header_level(node));
129	btrfs_set_root_generation(item, btrfs_header_generation(node));
130}
131
132/*
133 * copy the data in 'item' into the btree
134 */
135int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
136		      *root, struct btrfs_key *key, struct btrfs_root_item
137		      *item)
138{
139	struct btrfs_path *path;
140	struct extent_buffer *l;
141	int ret;
142	int slot;
143	unsigned long ptr;
144	int old_len;
145
146	path = btrfs_alloc_path();
147	if (!path)
148		return -ENOMEM;
149
150	ret = btrfs_search_slot(trans, root, key, path, 0, 1);
151	if (ret < 0) {
152		btrfs_abort_transaction(trans, root, ret);
153		goto out;
154	}
155
156	if (ret != 0) {
157		btrfs_print_leaf(root, path->nodes[0]);
158		btrfs_crit(root->fs_info, "unable to update root key %llu %u %llu",
159		       key->objectid, key->type, key->offset);
160		BUG_ON(1);
161	}
162
163	l = path->nodes[0];
164	slot = path->slots[0];
165	ptr = btrfs_item_ptr_offset(l, slot);
166	old_len = btrfs_item_size_nr(l, slot);
167
168	/*
169	 * If this is the first time we update the root item which originated
170	 * from an older kernel, we need to enlarge the item size to make room
171	 * for the added fields.
172	 */
173	if (old_len < sizeof(*item)) {
174		btrfs_release_path(path);
175		ret = btrfs_search_slot(trans, root, key, path,
176				-1, 1);
177		if (ret < 0) {
178			btrfs_abort_transaction(trans, root, ret);
179			goto out;
180		}
181
182		ret = btrfs_del_item(trans, root, path);
183		if (ret < 0) {
184			btrfs_abort_transaction(trans, root, ret);
185			goto out;
186		}
187		btrfs_release_path(path);
188		ret = btrfs_insert_empty_item(trans, root, path,
189				key, sizeof(*item));
190		if (ret < 0) {
191			btrfs_abort_transaction(trans, root, ret);
192			goto out;
193		}
194		l = path->nodes[0];
195		slot = path->slots[0];
196		ptr = btrfs_item_ptr_offset(l, slot);
197	}
198
199	/*
200	 * Update generation_v2 so at the next mount we know the new root
201	 * fields are valid.
202	 */
203	btrfs_set_root_generation_v2(item, btrfs_root_generation(item));
204
205	write_extent_buffer(l, item, ptr, sizeof(*item));
206	btrfs_mark_buffer_dirty(path->nodes[0]);
207out:
208	btrfs_free_path(path);
209	return ret;
210}
211
212int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
213		      struct btrfs_key *key, struct btrfs_root_item *item)
214{
215	/*
216	 * Make sure generation v1 and v2 match. See update_root for details.
217	 */
218	btrfs_set_root_generation_v2(item, btrfs_root_generation(item));
219	return btrfs_insert_item(trans, root, key, item, sizeof(*item));
220}
221
222int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
223{
224	struct extent_buffer *leaf;
225	struct btrfs_path *path;
226	struct btrfs_key key;
227	struct btrfs_key root_key;
228	struct btrfs_root *root;
229	int err = 0;
230	int ret;
231	bool can_recover = true;
232
233	if (tree_root->fs_info->sb->s_flags & MS_RDONLY)
234		can_recover = false;
235
236	path = btrfs_alloc_path();
237	if (!path)
238		return -ENOMEM;
239
240	key.objectid = BTRFS_ORPHAN_OBJECTID;
241	key.type = BTRFS_ORPHAN_ITEM_KEY;
242	key.offset = 0;
243
244	root_key.type = BTRFS_ROOT_ITEM_KEY;
245	root_key.offset = (u64)-1;
246
247	while (1) {
248		ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
249		if (ret < 0) {
250			err = ret;
251			break;
252		}
253
254		leaf = path->nodes[0];
255		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
256			ret = btrfs_next_leaf(tree_root, path);
257			if (ret < 0)
258				err = ret;
259			if (ret != 0)
260				break;
261			leaf = path->nodes[0];
262		}
263
264		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
265		btrfs_release_path(path);
266
267		if (key.objectid != BTRFS_ORPHAN_OBJECTID ||
268		    key.type != BTRFS_ORPHAN_ITEM_KEY)
269			break;
270
271		root_key.objectid = key.offset;
272		key.offset++;
273
274		root = btrfs_read_fs_root(tree_root, &root_key);
275		err = PTR_ERR_OR_ZERO(root);
276		if (err && err != -ENOENT) {
277			break;
278		} else if (err == -ENOENT) {
279			struct btrfs_trans_handle *trans;
280
281			btrfs_release_path(path);
282
283			trans = btrfs_join_transaction(tree_root);
284			if (IS_ERR(trans)) {
285				err = PTR_ERR(trans);
286				btrfs_error(tree_root->fs_info, err,
287					    "Failed to start trans to delete "
288					    "orphan item");
289				break;
290			}
291			err = btrfs_del_orphan_item(trans, tree_root,
292						    root_key.objectid);
293			btrfs_end_transaction(trans, tree_root);
294			if (err) {
295				btrfs_error(tree_root->fs_info, err,
296					    "Failed to delete root orphan "
297					    "item");
298				break;
299			}
300			continue;
301		}
302
303		err = btrfs_init_fs_root(root);
304		if (err) {
305			btrfs_free_fs_root(root);
306			break;
307		}
308
309		set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
310
311		err = btrfs_insert_fs_root(root->fs_info, root);
312		if (err) {
313			BUG_ON(err == -EEXIST);
314			btrfs_free_fs_root(root);
315			break;
316		}
317
318		if (btrfs_root_refs(&root->root_item) == 0)
319			btrfs_add_dead_root(root);
320	}
321
322	btrfs_free_path(path);
323	return err;
324}
325
326/* drop the root item for 'key' from 'root' */
327int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
328		   struct btrfs_key *key)
329{
330	struct btrfs_path *path;
331	int ret;
332
333	path = btrfs_alloc_path();
334	if (!path)
335		return -ENOMEM;
336	ret = btrfs_search_slot(trans, root, key, path, -1, 1);
337	if (ret < 0)
338		goto out;
339
340	BUG_ON(ret != 0);
341
342	ret = btrfs_del_item(trans, root, path);
343out:
344	btrfs_free_path(path);
345	return ret;
346}
347
348int btrfs_del_root_ref(struct btrfs_trans_handle *trans,
349		       struct btrfs_root *tree_root,
350		       u64 root_id, u64 ref_id, u64 dirid, u64 *sequence,
351		       const char *name, int name_len)
352
353{
354	struct btrfs_path *path;
355	struct btrfs_root_ref *ref;
356	struct extent_buffer *leaf;
357	struct btrfs_key key;
358	unsigned long ptr;
359	int err = 0;
360	int ret;
361
362	path = btrfs_alloc_path();
363	if (!path)
364		return -ENOMEM;
365
366	key.objectid = root_id;
367	key.type = BTRFS_ROOT_BACKREF_KEY;
368	key.offset = ref_id;
369again:
370	ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
371	BUG_ON(ret < 0);
372	if (ret == 0) {
373		leaf = path->nodes[0];
374		ref = btrfs_item_ptr(leaf, path->slots[0],
375				     struct btrfs_root_ref);
376
377		WARN_ON(btrfs_root_ref_dirid(leaf, ref) != dirid);
378		WARN_ON(btrfs_root_ref_name_len(leaf, ref) != name_len);
379		ptr = (unsigned long)(ref + 1);
380		WARN_ON(memcmp_extent_buffer(leaf, name, ptr, name_len));
381		*sequence = btrfs_root_ref_sequence(leaf, ref);
382
383		ret = btrfs_del_item(trans, tree_root, path);
384		if (ret) {
385			err = ret;
386			goto out;
387		}
388	} else
389		err = -ENOENT;
390
391	if (key.type == BTRFS_ROOT_BACKREF_KEY) {
392		btrfs_release_path(path);
393		key.objectid = ref_id;
394		key.type = BTRFS_ROOT_REF_KEY;
395		key.offset = root_id;
396		goto again;
397	}
398
399out:
400	btrfs_free_path(path);
401	return err;
402}
403
404/*
405 * add a btrfs_root_ref item.  type is either BTRFS_ROOT_REF_KEY
406 * or BTRFS_ROOT_BACKREF_KEY.
407 *
408 * The dirid, sequence, name and name_len refer to the directory entry
409 * that is referencing the root.
410 *
411 * For a forward ref, the root_id is the id of the tree referencing
412 * the root and ref_id is the id of the subvol  or snapshot.
413 *
414 * For a back ref the root_id is the id of the subvol or snapshot and
415 * ref_id is the id of the tree referencing it.
416 *
417 * Will return 0, -ENOMEM, or anything from the CoW path
418 */
419int btrfs_add_root_ref(struct btrfs_trans_handle *trans,
420		       struct btrfs_root *tree_root,
421		       u64 root_id, u64 ref_id, u64 dirid, u64 sequence,
422		       const char *name, int name_len)
423{
424	struct btrfs_key key;
425	int ret;
426	struct btrfs_path *path;
427	struct btrfs_root_ref *ref;
428	struct extent_buffer *leaf;
429	unsigned long ptr;
430
431	path = btrfs_alloc_path();
432	if (!path)
433		return -ENOMEM;
434
435	key.objectid = root_id;
436	key.type = BTRFS_ROOT_BACKREF_KEY;
437	key.offset = ref_id;
438again:
439	ret = btrfs_insert_empty_item(trans, tree_root, path, &key,
440				      sizeof(*ref) + name_len);
441	if (ret) {
442		btrfs_abort_transaction(trans, tree_root, ret);
443		btrfs_free_path(path);
444		return ret;
445	}
446
447	leaf = path->nodes[0];
448	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
449	btrfs_set_root_ref_dirid(leaf, ref, dirid);
450	btrfs_set_root_ref_sequence(leaf, ref, sequence);
451	btrfs_set_root_ref_name_len(leaf, ref, name_len);
452	ptr = (unsigned long)(ref + 1);
453	write_extent_buffer(leaf, name, ptr, name_len);
454	btrfs_mark_buffer_dirty(leaf);
455
456	if (key.type == BTRFS_ROOT_BACKREF_KEY) {
457		btrfs_release_path(path);
458		key.objectid = ref_id;
459		key.type = BTRFS_ROOT_REF_KEY;
460		key.offset = root_id;
461		goto again;
462	}
463
464	btrfs_free_path(path);
465	return 0;
466}
467
468/*
469 * Old btrfs forgets to init root_item->flags and root_item->byte_limit
470 * for subvolumes. To work around this problem, we steal a bit from
471 * root_item->inode_item->flags, and use it to indicate if those fields
472 * have been properly initialized.
473 */
474void btrfs_check_and_init_root_item(struct btrfs_root_item *root_item)
475{
476	u64 inode_flags = btrfs_stack_inode_flags(&root_item->inode);
477
478	if (!(inode_flags & BTRFS_INODE_ROOT_ITEM_INIT)) {
479		inode_flags |= BTRFS_INODE_ROOT_ITEM_INIT;
480		btrfs_set_stack_inode_flags(&root_item->inode, inode_flags);
481		btrfs_set_root_flags(root_item, 0);
482		btrfs_set_root_limit(root_item, 0);
483	}
484}
485
486void btrfs_update_root_times(struct btrfs_trans_handle *trans,
487			     struct btrfs_root *root)
488{
489	struct btrfs_root_item *item = &root->root_item;
490	struct timespec ct = CURRENT_TIME;
491
492	spin_lock(&root->root_item_lock);
493	btrfs_set_root_ctransid(item, trans->transid);
494	btrfs_set_stack_timespec_sec(&item->ctime, ct.tv_sec);
495	btrfs_set_stack_timespec_nsec(&item->ctime, ct.tv_nsec);
496	spin_unlock(&root->root_item_lock);
497}
498