xref: /linux/fs/btrfs/uuid-tree.c (revision e5267796482fad93ee7948c7cbc37f32046244f7)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) STRATO AG 2013.  All rights reserved.
4  */
5 
6 #include <linux/kthread.h>
7 #include <linux/uuid.h>
8 #include <linux/unaligned.h>
9 #include "messages.h"
10 #include "ctree.h"
11 #include "transaction.h"
12 #include "disk-io.h"
13 #include "fs.h"
14 #include "accessors.h"
15 #include "uuid-tree.h"
16 #include "ioctl.h"
17 
18 static void btrfs_uuid_to_key(const u8 *uuid, u8 type, struct btrfs_key *key)
19 {
20 	key->type = type;
21 	key->objectid = get_unaligned_le64(uuid);
22 	key->offset = get_unaligned_le64(uuid + sizeof(u64));
23 }
24 
25 /* return -ENOENT for !found, < 0 for errors, or 0 if an item was found */
26 static int btrfs_uuid_tree_lookup(struct btrfs_root *uuid_root, const u8 *uuid,
27 				  u8 type, u64 subid)
28 {
29 	int ret;
30 	BTRFS_PATH_AUTO_FREE(path);
31 	struct extent_buffer *eb;
32 	int slot;
33 	u32 item_size;
34 	unsigned long offset;
35 	struct btrfs_key key;
36 
37 	if (WARN_ON_ONCE(!uuid_root))
38 		return -EINVAL;
39 
40 	path = btrfs_alloc_path();
41 	if (!path)
42 		return -ENOMEM;
43 
44 	btrfs_uuid_to_key(uuid, type, &key);
45 	ret = btrfs_search_slot(NULL, uuid_root, &key, path, 0, 0);
46 	if (ret < 0)
47 		return ret;
48 	if (ret > 0)
49 		return -ENOENT;
50 
51 	eb = path->nodes[0];
52 	slot = path->slots[0];
53 	item_size = btrfs_item_size(eb, slot);
54 	offset = btrfs_item_ptr_offset(eb, slot);
55 	ret = -ENOENT;
56 
57 	if (!IS_ALIGNED(item_size, sizeof(u64))) {
58 		btrfs_warn(uuid_root->fs_info,
59 			   "uuid item with illegal size %lu!",
60 			   (unsigned long)item_size);
61 		return ret;
62 	}
63 	while (item_size) {
64 		__le64 data;
65 
66 		read_extent_buffer(eb, &data, offset, sizeof(data));
67 		if (le64_to_cpu(data) == subid) {
68 			ret = 0;
69 			break;
70 		}
71 		offset += sizeof(data);
72 		item_size -= sizeof(data);
73 	}
74 
75 	return ret;
76 }
77 
78 int btrfs_uuid_tree_add(struct btrfs_trans_handle *trans, const u8 *uuid, u8 type,
79 			u64 subid_cpu)
80 {
81 	struct btrfs_fs_info *fs_info = trans->fs_info;
82 	struct btrfs_root *uuid_root = fs_info->uuid_root;
83 	int ret;
84 	BTRFS_PATH_AUTO_FREE(path);
85 	struct btrfs_key key;
86 	struct extent_buffer *eb;
87 	int slot;
88 	unsigned long offset;
89 	__le64 subid_le;
90 
91 	ret = btrfs_uuid_tree_lookup(uuid_root, uuid, type, subid_cpu);
92 	if (ret != -ENOENT)
93 		return ret;
94 
95 	btrfs_uuid_to_key(uuid, type, &key);
96 
97 	path = btrfs_alloc_path();
98 	if (!path)
99 		return -ENOMEM;
100 
101 	ret = btrfs_insert_empty_item(trans, uuid_root, path, &key,
102 				      sizeof(subid_le));
103 	if (ret == 0) {
104 		/* Add an item for the type for the first time */
105 		eb = path->nodes[0];
106 		slot = path->slots[0];
107 		offset = btrfs_item_ptr_offset(eb, slot);
108 	} else if (ret == -EEXIST) {
109 		/*
110 		 * An item with that type already exists.
111 		 * Extend the item and store the new subid at the end.
112 		 */
113 		btrfs_extend_item(trans, path, sizeof(subid_le));
114 		eb = path->nodes[0];
115 		slot = path->slots[0];
116 		offset = btrfs_item_ptr_offset(eb, slot);
117 		offset += btrfs_item_size(eb, slot) - sizeof(subid_le);
118 	} else {
119 		btrfs_warn(fs_info,
120 			   "insert uuid item failed %d (0x%016llx, 0x%016llx) type %u!",
121 			   ret, key.objectid, key.offset, type);
122 		return ret;
123 	}
124 
125 	subid_le = cpu_to_le64(subid_cpu);
126 	write_extent_buffer(eb, &subid_le, offset, sizeof(subid_le));
127 	return 0;
128 }
129 
130 int btrfs_uuid_tree_remove(struct btrfs_trans_handle *trans, const u8 *uuid, u8 type,
131 			u64 subid)
132 {
133 	struct btrfs_fs_info *fs_info = trans->fs_info;
134 	struct btrfs_root *uuid_root = fs_info->uuid_root;
135 	int ret;
136 	BTRFS_PATH_AUTO_FREE(path);
137 	struct btrfs_key key;
138 	struct extent_buffer *eb;
139 	int slot;
140 	unsigned long offset;
141 	u32 item_size;
142 	unsigned long move_dst;
143 	unsigned long move_src;
144 	unsigned long move_len;
145 
146 	if (WARN_ON_ONCE(!uuid_root))
147 		return -EINVAL;
148 
149 	btrfs_uuid_to_key(uuid, type, &key);
150 
151 	path = btrfs_alloc_path();
152 	if (!path)
153 		return -ENOMEM;
154 
155 	ret = btrfs_search_slot(trans, uuid_root, &key, path, -1, 1);
156 	if (ret < 0) {
157 		btrfs_warn(fs_info, "error %d while searching for uuid item!",
158 			   ret);
159 		return ret;
160 	}
161 	if (ret > 0)
162 		return -ENOENT;
163 
164 	eb = path->nodes[0];
165 	slot = path->slots[0];
166 	offset = btrfs_item_ptr_offset(eb, slot);
167 	item_size = btrfs_item_size(eb, slot);
168 	if (!IS_ALIGNED(item_size, sizeof(u64))) {
169 		btrfs_warn(fs_info, "uuid item with illegal size %lu!",
170 			   (unsigned long)item_size);
171 		return -ENOENT;
172 	}
173 	while (item_size) {
174 		__le64 read_subid;
175 
176 		read_extent_buffer(eb, &read_subid, offset, sizeof(read_subid));
177 		if (le64_to_cpu(read_subid) == subid)
178 			break;
179 		offset += sizeof(read_subid);
180 		item_size -= sizeof(read_subid);
181 	}
182 
183 	if (!item_size)
184 		return -ENOENT;
185 
186 	item_size = btrfs_item_size(eb, slot);
187 	if (item_size == sizeof(subid))
188 		return btrfs_del_item(trans, uuid_root, path);
189 
190 	move_dst = offset;
191 	move_src = offset + sizeof(subid);
192 	move_len = item_size - (move_src - btrfs_item_ptr_offset(eb, slot));
193 	memmove_extent_buffer(eb, move_dst, move_src, move_len);
194 	btrfs_truncate_item(trans, path, item_size - sizeof(subid), 1);
195 
196 	return 0;
197 }
198 
199 /*
200  * Check if we can add one root ID to a UUID key.
201  * If the key does not yet exists, we can, otherwise only if extended item does
202  * not exceeds the maximum item size permitted by the leaf size.
203  *
204  * Returns 0 on success, negative value on error.
205  */
206 int btrfs_uuid_tree_check_overflow(struct btrfs_fs_info *fs_info,
207 				   const u8 *uuid, u8 type)
208 {
209 	BTRFS_PATH_AUTO_FREE(path);
210 	int ret;
211 	u32 item_size;
212 	struct btrfs_key key;
213 
214 	if (WARN_ON_ONCE(!fs_info->uuid_root))
215 		return -EINVAL;
216 
217 	path = btrfs_alloc_path();
218 	if (!path)
219 		return -ENOMEM;
220 
221 	btrfs_uuid_to_key(uuid, type, &key);
222 	ret = btrfs_search_slot(NULL, fs_info->uuid_root, &key, path, 0, 0);
223 	if (ret < 0)
224 		return ret;
225 	if (ret > 0)
226 		return 0;
227 
228 	item_size = btrfs_item_size(path->nodes[0], path->slots[0]);
229 
230 	if (sizeof(struct btrfs_item) + item_size + sizeof(u64) >
231 	    BTRFS_LEAF_DATA_SIZE(fs_info))
232 		return -EOVERFLOW;
233 
234 	return 0;
235 }
236 
237 static int btrfs_uuid_iter_rem(struct btrfs_root *uuid_root, u8 *uuid, u8 type,
238 			       u64 subid)
239 {
240 	struct btrfs_trans_handle *trans;
241 	int ret;
242 
243 	/* 1 - for the uuid item */
244 	trans = btrfs_start_transaction(uuid_root, 1);
245 	if (IS_ERR(trans))
246 		return PTR_ERR(trans);
247 
248 	ret = btrfs_uuid_tree_remove(trans, uuid, type, subid);
249 	btrfs_end_transaction(trans);
250 	return ret;
251 }
252 
253 /*
254  * Check if there's an matching subvolume for given UUID
255  *
256  * Return:
257  * 0	check succeeded, the entry is not outdated
258  * > 0	if the check failed, the caller should remove the entry
259  * < 0	if an error occurred
260  */
261 static int btrfs_check_uuid_tree_entry(struct btrfs_fs_info *fs_info,
262 				       const u8 *uuid, u8 type, u64 subvolid)
263 {
264 	int ret = 0;
265 	struct btrfs_root *subvol_root;
266 
267 	if (type != BTRFS_UUID_KEY_SUBVOL &&
268 	    type != BTRFS_UUID_KEY_RECEIVED_SUBVOL)
269 		return 0;
270 
271 	subvol_root = btrfs_get_fs_root(fs_info, subvolid, true);
272 	if (IS_ERR(subvol_root)) {
273 		ret = PTR_ERR(subvol_root);
274 		if (ret == -ENOENT)
275 			return 1;
276 		return ret;
277 	}
278 
279 	switch (type) {
280 	case BTRFS_UUID_KEY_SUBVOL:
281 		if (memcmp(uuid, subvol_root->root_item.uuid, BTRFS_UUID_SIZE))
282 			ret = 1;
283 		break;
284 	case BTRFS_UUID_KEY_RECEIVED_SUBVOL:
285 		if (memcmp(uuid, subvol_root->root_item.received_uuid,
286 			   BTRFS_UUID_SIZE))
287 			ret = 1;
288 		break;
289 	}
290 	btrfs_put_root(subvol_root);
291 
292 	return ret;
293 }
294 
295 int btrfs_uuid_tree_iterate(struct btrfs_fs_info *fs_info)
296 {
297 	struct btrfs_root *root = fs_info->uuid_root;
298 	struct btrfs_key key;
299 	BTRFS_PATH_AUTO_FREE(path);
300 	int ret = 0;
301 	struct extent_buffer *leaf;
302 	int slot;
303 	u32 item_size;
304 	unsigned long offset;
305 
306 	path = btrfs_alloc_path();
307 	if (!path)
308 		return -ENOMEM;
309 
310 	key.objectid = 0;
311 	key.type = 0;
312 	key.offset = 0;
313 
314 again_search_slot:
315 	ret = btrfs_search_forward(root, &key, path, BTRFS_OLDEST_GENERATION);
316 	if (ret < 0)
317 		return ret;
318 	if (ret > 0)
319 		return 0;
320 
321 	while (1) {
322 		if (btrfs_fs_closing(fs_info))
323 			return -EINTR;
324 
325 		cond_resched();
326 		leaf = path->nodes[0];
327 		slot = path->slots[0];
328 		btrfs_item_key_to_cpu(leaf, &key, slot);
329 
330 		if (key.type != BTRFS_UUID_KEY_SUBVOL &&
331 		    key.type != BTRFS_UUID_KEY_RECEIVED_SUBVOL)
332 			goto skip;
333 
334 		offset = btrfs_item_ptr_offset(leaf, slot);
335 		item_size = btrfs_item_size(leaf, slot);
336 		if (!IS_ALIGNED(item_size, sizeof(u64))) {
337 			btrfs_warn(fs_info,
338 				   "uuid item with illegal size %lu!",
339 				   (unsigned long)item_size);
340 			goto skip;
341 		}
342 		while (item_size) {
343 			u8 uuid[BTRFS_UUID_SIZE];
344 			__le64 subid_le;
345 			u64 subid_cpu;
346 
347 			put_unaligned_le64(key.objectid, uuid);
348 			put_unaligned_le64(key.offset, uuid + sizeof(u64));
349 			read_extent_buffer(leaf, &subid_le, offset,
350 					   sizeof(subid_le));
351 			subid_cpu = le64_to_cpu(subid_le);
352 			ret = btrfs_check_uuid_tree_entry(fs_info, uuid,
353 							  key.type, subid_cpu);
354 			if (ret < 0)
355 				return ret;
356 			if (ret > 0) {
357 				btrfs_release_path(path);
358 				ret = btrfs_uuid_iter_rem(root, uuid, key.type,
359 							  subid_cpu);
360 				if (ret == 0) {
361 					/*
362 					 * this might look inefficient, but the
363 					 * justification is that it is an
364 					 * exception that check_func returns 1,
365 					 * and that in the regular case only one
366 					 * entry per UUID exists.
367 					 */
368 					goto again_search_slot;
369 				}
370 				if (ret < 0 && ret != -ENOENT)
371 					return ret;
372 				key.offset++;
373 				goto again_search_slot;
374 			}
375 			item_size -= sizeof(subid_le);
376 			offset += sizeof(subid_le);
377 		}
378 
379 skip:
380 		ret = btrfs_next_item(root, path);
381 		if (ret == 0)
382 			continue;
383 		else if (ret > 0)
384 			ret = 0;
385 		break;
386 	}
387 
388 	return ret;
389 }
390 
391 int btrfs_uuid_scan_kthread(void *data)
392 {
393 	struct btrfs_fs_info *fs_info = data;
394 	struct btrfs_root *root = fs_info->tree_root;
395 	struct btrfs_key key;
396 	struct btrfs_path *path = NULL;
397 	int ret = 0;
398 	struct extent_buffer *eb;
399 	int slot;
400 	struct btrfs_root_item root_item;
401 	u32 item_size;
402 	struct btrfs_trans_handle *trans = NULL;
403 	bool closing = false;
404 
405 	path = btrfs_alloc_path();
406 	if (!path) {
407 		ret = -ENOMEM;
408 		goto out;
409 	}
410 
411 	key.objectid = 0;
412 	key.type = BTRFS_ROOT_ITEM_KEY;
413 	key.offset = 0;
414 
415 	while (1) {
416 		if (btrfs_fs_closing(fs_info)) {
417 			closing = true;
418 			break;
419 		}
420 		ret = btrfs_search_forward(root, &key, path,
421 				BTRFS_OLDEST_GENERATION);
422 		if (ret) {
423 			if (ret > 0)
424 				ret = 0;
425 			break;
426 		}
427 
428 		if (key.type != BTRFS_ROOT_ITEM_KEY ||
429 		    (key.objectid < BTRFS_FIRST_FREE_OBJECTID &&
430 		     key.objectid != BTRFS_FS_TREE_OBJECTID) ||
431 		    key.objectid > BTRFS_LAST_FREE_OBJECTID)
432 			goto skip;
433 
434 		eb = path->nodes[0];
435 		slot = path->slots[0];
436 		item_size = btrfs_item_size(eb, slot);
437 		if (item_size < sizeof(root_item))
438 			goto skip;
439 
440 		read_extent_buffer(eb, &root_item,
441 				   btrfs_item_ptr_offset(eb, slot),
442 				   (int)sizeof(root_item));
443 		if (btrfs_root_refs(&root_item) == 0)
444 			goto skip;
445 
446 		if (!btrfs_is_empty_uuid(root_item.uuid) ||
447 		    !btrfs_is_empty_uuid(root_item.received_uuid)) {
448 			if (trans)
449 				goto update_tree;
450 
451 			btrfs_release_path(path);
452 			/*
453 			 * 1 - subvol uuid item
454 			 * 1 - received_subvol uuid item
455 			 */
456 			trans = btrfs_start_transaction(fs_info->uuid_root, 2);
457 			if (IS_ERR(trans)) {
458 				ret = PTR_ERR(trans);
459 				break;
460 			}
461 			continue;
462 		} else {
463 			goto skip;
464 		}
465 update_tree:
466 		btrfs_release_path(path);
467 		if (!btrfs_is_empty_uuid(root_item.uuid)) {
468 			ret = btrfs_uuid_tree_add(trans, root_item.uuid,
469 						  BTRFS_UUID_KEY_SUBVOL,
470 						  key.objectid);
471 			if (ret < 0) {
472 				btrfs_warn(fs_info, "uuid_tree_add failed %d",
473 					ret);
474 				break;
475 			}
476 		}
477 
478 		if (!btrfs_is_empty_uuid(root_item.received_uuid)) {
479 			ret = btrfs_uuid_tree_add(trans,
480 						  root_item.received_uuid,
481 						 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
482 						  key.objectid);
483 			if (ret < 0) {
484 				btrfs_warn(fs_info, "uuid_tree_add failed %d",
485 					ret);
486 				break;
487 			}
488 		}
489 
490 skip:
491 		btrfs_release_path(path);
492 		if (trans) {
493 			ret = btrfs_end_transaction(trans);
494 			trans = NULL;
495 			if (ret)
496 				break;
497 		}
498 
499 		if (key.offset < (u64)-1) {
500 			key.offset++;
501 		} else if (key.type < BTRFS_ROOT_ITEM_KEY) {
502 			key.offset = 0;
503 			key.type = BTRFS_ROOT_ITEM_KEY;
504 		} else if (key.objectid < (u64)-1) {
505 			key.offset = 0;
506 			key.type = BTRFS_ROOT_ITEM_KEY;
507 			key.objectid++;
508 		} else {
509 			break;
510 		}
511 		cond_resched();
512 	}
513 
514 out:
515 	btrfs_free_path(path);
516 	if (!IS_ERR_OR_NULL(trans))
517 		btrfs_end_transaction(trans);
518 	if (ret)
519 		btrfs_warn(fs_info, "btrfs_uuid_scan_kthread failed %d", ret);
520 	else if (!closing)
521 		set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags);
522 	up(&fs_info->uuid_tree_rescan_sem);
523 	return 0;
524 }
525 
526 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info)
527 {
528 	struct btrfs_trans_handle *trans;
529 	struct btrfs_root *tree_root = fs_info->tree_root;
530 	struct btrfs_root *uuid_root;
531 	struct task_struct *task;
532 	int ret;
533 
534 	/*
535 	 * 1 - root node
536 	 * 1 - root item
537 	 */
538 	trans = btrfs_start_transaction(tree_root, 2);
539 	if (IS_ERR(trans))
540 		return PTR_ERR(trans);
541 
542 	uuid_root = btrfs_create_tree(trans, BTRFS_UUID_TREE_OBJECTID);
543 	if (IS_ERR(uuid_root)) {
544 		ret = PTR_ERR(uuid_root);
545 		btrfs_abort_transaction(trans, ret);
546 		btrfs_end_transaction(trans);
547 		return ret;
548 	}
549 
550 	fs_info->uuid_root = uuid_root;
551 
552 	ret = btrfs_commit_transaction(trans);
553 	if (ret)
554 		return ret;
555 
556 	down(&fs_info->uuid_tree_rescan_sem);
557 	task = kthread_run(btrfs_uuid_scan_kthread, fs_info, "btrfs-uuid");
558 	if (IS_ERR(task)) {
559 		/* fs_info->update_uuid_tree_gen remains 0 in all error case */
560 		btrfs_warn(fs_info, "failed to start uuid_scan task");
561 		up(&fs_info->uuid_tree_rescan_sem);
562 		return PTR_ERR(task);
563 	}
564 
565 	return 0;
566 }
567