xref: /linux/fs/btrfs/locking.c (revision f85f5ae45ad945270a8884261de8249431e8b5a6)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2008 Oracle.  All rights reserved.
4  */
5 
6 #include <linux/sched.h>
7 #include <linux/pagemap.h>
8 #include <linux/spinlock.h>
9 #include <linux/page-flags.h>
10 #include <asm/bug.h>
11 #include "misc.h"
12 #include "ctree.h"
13 #include "extent_io.h"
14 #include "locking.h"
15 #include "accessors.h"
16 
17 /*
18  * Lockdep class keys for extent_buffer->lock's in this root.  For a given
19  * eb, the lockdep key is determined by the btrfs_root it belongs to and
20  * the level the eb occupies in the tree.
21  *
22  * Different roots are used for different purposes and may nest inside each
23  * other and they require separate keysets.  As lockdep keys should be
24  * static, assign keysets according to the purpose of the root as indicated
25  * by btrfs_root->root_key.objectid.  This ensures that all special purpose
26  * roots have separate keysets.
27  *
28  * Lock-nesting across peer nodes is always done with the immediate parent
29  * node locked thus preventing deadlock.  As lockdep doesn't know this, use
30  * subclass to avoid triggering lockdep warning in such cases.
31  *
32  * The key is set by the readpage_end_io_hook after the buffer has passed
33  * csum validation but before the pages are unlocked.  It is also set by
34  * btrfs_init_new_buffer on freshly allocated blocks.
35  *
36  * We also add a check to make sure the highest level of the tree is the
37  * same as our lockdep setup here.  If BTRFS_MAX_LEVEL changes, this code
38  * needs update as well.
39  */
40 #ifdef CONFIG_DEBUG_LOCK_ALLOC
41 #if BTRFS_MAX_LEVEL != 8
42 #error
43 #endif
44 
45 #define DEFINE_LEVEL(stem, level)					\
46 	.names[level] = "btrfs-" stem "-0" #level,
47 
48 #define DEFINE_NAME(stem)						\
49 	DEFINE_LEVEL(stem, 0)						\
50 	DEFINE_LEVEL(stem, 1)						\
51 	DEFINE_LEVEL(stem, 2)						\
52 	DEFINE_LEVEL(stem, 3)						\
53 	DEFINE_LEVEL(stem, 4)						\
54 	DEFINE_LEVEL(stem, 5)						\
55 	DEFINE_LEVEL(stem, 6)						\
56 	DEFINE_LEVEL(stem, 7)
57 
58 static struct btrfs_lockdep_keyset {
59 	u64			id;		/* root objectid */
60 	/* Longest entry: btrfs-block-group-00 */
61 	char			names[BTRFS_MAX_LEVEL][24];
62 	struct lock_class_key	keys[BTRFS_MAX_LEVEL];
63 } btrfs_lockdep_keysets[] = {
64 	{ .id = BTRFS_ROOT_TREE_OBJECTID,	DEFINE_NAME("root")	},
65 	{ .id = BTRFS_EXTENT_TREE_OBJECTID,	DEFINE_NAME("extent")	},
66 	{ .id = BTRFS_CHUNK_TREE_OBJECTID,	DEFINE_NAME("chunk")	},
67 	{ .id = BTRFS_DEV_TREE_OBJECTID,	DEFINE_NAME("dev")	},
68 	{ .id = BTRFS_CSUM_TREE_OBJECTID,	DEFINE_NAME("csum")	},
69 	{ .id = BTRFS_QUOTA_TREE_OBJECTID,	DEFINE_NAME("quota")	},
70 	{ .id = BTRFS_TREE_LOG_OBJECTID,	DEFINE_NAME("log")	},
71 	{ .id = BTRFS_TREE_RELOC_OBJECTID,	DEFINE_NAME("treloc")	},
72 	{ .id = BTRFS_DATA_RELOC_TREE_OBJECTID,	DEFINE_NAME("dreloc")	},
73 	{ .id = BTRFS_UUID_TREE_OBJECTID,	DEFINE_NAME("uuid")	},
74 	{ .id = BTRFS_FREE_SPACE_TREE_OBJECTID,	DEFINE_NAME("free-space") },
75 	{ .id = BTRFS_BLOCK_GROUP_TREE_OBJECTID, DEFINE_NAME("block-group") },
76 	{ .id = 0,				DEFINE_NAME("tree")	},
77 };
78 
79 #undef DEFINE_LEVEL
80 #undef DEFINE_NAME
81 
82 void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, int level)
83 {
84 	struct btrfs_lockdep_keyset *ks;
85 
86 	BUG_ON(level >= ARRAY_SIZE(ks->keys));
87 
88 	/* Find the matching keyset, id 0 is the default entry */
89 	for (ks = btrfs_lockdep_keysets; ks->id; ks++)
90 		if (ks->id == objectid)
91 			break;
92 
93 	lockdep_set_class_and_name(&eb->lock, &ks->keys[level], ks->names[level]);
94 }
95 
96 void btrfs_maybe_reset_lockdep_class(struct btrfs_root *root, struct extent_buffer *eb)
97 {
98 	if (test_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &root->state))
99 		btrfs_set_buffer_lockdep_class(root->root_key.objectid,
100 					       eb, btrfs_header_level(eb));
101 }
102 
103 #endif
104 
105 /*
106  * Extent buffer locking
107  * =====================
108  *
109  * We use a rw_semaphore for tree locking, and the semantics are exactly the
110  * same:
111  *
112  * - reader/writer exclusion
113  * - writer/writer exclusion
114  * - reader/reader sharing
115  * - try-lock semantics for readers and writers
116  *
117  * The rwsem implementation does opportunistic spinning which reduces number of
118  * times the locking task needs to sleep.
119  */
120 
121 /*
122  * __btrfs_tree_read_lock - lock extent buffer for read
123  * @eb:		the eb to be locked
124  * @nest:	the nesting level to be used for lockdep
125  *
126  * This takes the read lock on the extent buffer, using the specified nesting
127  * level for lockdep purposes.
128  */
129 void __btrfs_tree_read_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest)
130 {
131 	u64 start_ns = 0;
132 
133 	if (trace_btrfs_tree_read_lock_enabled())
134 		start_ns = ktime_get_ns();
135 
136 	down_read_nested(&eb->lock, nest);
137 	trace_btrfs_tree_read_lock(eb, start_ns);
138 }
139 
140 void btrfs_tree_read_lock(struct extent_buffer *eb)
141 {
142 	__btrfs_tree_read_lock(eb, BTRFS_NESTING_NORMAL);
143 }
144 
145 /*
146  * Try-lock for read.
147  *
148  * Return 1 if the rwlock has been taken, 0 otherwise
149  */
150 int btrfs_try_tree_read_lock(struct extent_buffer *eb)
151 {
152 	if (down_read_trylock(&eb->lock)) {
153 		trace_btrfs_try_tree_read_lock(eb);
154 		return 1;
155 	}
156 	return 0;
157 }
158 
159 /*
160  * Try-lock for write.
161  *
162  * Return 1 if the rwlock has been taken, 0 otherwise
163  */
164 int btrfs_try_tree_write_lock(struct extent_buffer *eb)
165 {
166 	if (down_write_trylock(&eb->lock)) {
167 		eb->lock_owner = current->pid;
168 		trace_btrfs_try_tree_write_lock(eb);
169 		return 1;
170 	}
171 	return 0;
172 }
173 
174 /*
175  * Release read lock.
176  */
177 void btrfs_tree_read_unlock(struct extent_buffer *eb)
178 {
179 	trace_btrfs_tree_read_unlock(eb);
180 	up_read(&eb->lock);
181 }
182 
183 /*
184  * __btrfs_tree_lock - lock eb for write
185  * @eb:		the eb to lock
186  * @nest:	the nesting to use for the lock
187  *
188  * Returns with the eb->lock write locked.
189  */
190 void __btrfs_tree_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest)
191 	__acquires(&eb->lock)
192 {
193 	u64 start_ns = 0;
194 
195 	if (trace_btrfs_tree_lock_enabled())
196 		start_ns = ktime_get_ns();
197 
198 	down_write_nested(&eb->lock, nest);
199 	eb->lock_owner = current->pid;
200 	trace_btrfs_tree_lock(eb, start_ns);
201 }
202 
203 void btrfs_tree_lock(struct extent_buffer *eb)
204 {
205 	__btrfs_tree_lock(eb, BTRFS_NESTING_NORMAL);
206 }
207 
208 /*
209  * Release the write lock.
210  */
211 void btrfs_tree_unlock(struct extent_buffer *eb)
212 {
213 	trace_btrfs_tree_unlock(eb);
214 	eb->lock_owner = 0;
215 	up_write(&eb->lock);
216 }
217 
218 /*
219  * This releases any locks held in the path starting at level and going all the
220  * way up to the root.
221  *
222  * btrfs_search_slot will keep the lock held on higher nodes in a few corner
223  * cases, such as COW of the block at slot zero in the node.  This ignores
224  * those rules, and it should only be called when there are no more updates to
225  * be done higher up in the tree.
226  */
227 void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
228 {
229 	int i;
230 
231 	if (path->keep_locks)
232 		return;
233 
234 	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
235 		if (!path->nodes[i])
236 			continue;
237 		if (!path->locks[i])
238 			continue;
239 		btrfs_tree_unlock_rw(path->nodes[i], path->locks[i]);
240 		path->locks[i] = 0;
241 	}
242 }
243 
244 /*
245  * Loop around taking references on and locking the root node of the tree until
246  * we end up with a lock on the root node.
247  *
248  * Return: root extent buffer with write lock held
249  */
250 struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
251 {
252 	struct extent_buffer *eb;
253 
254 	while (1) {
255 		eb = btrfs_root_node(root);
256 
257 		btrfs_maybe_reset_lockdep_class(root, eb);
258 		btrfs_tree_lock(eb);
259 		if (eb == root->node)
260 			break;
261 		btrfs_tree_unlock(eb);
262 		free_extent_buffer(eb);
263 	}
264 	return eb;
265 }
266 
267 /*
268  * Loop around taking references on and locking the root node of the tree until
269  * we end up with a lock on the root node.
270  *
271  * Return: root extent buffer with read lock held
272  */
273 struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root)
274 {
275 	struct extent_buffer *eb;
276 
277 	while (1) {
278 		eb = btrfs_root_node(root);
279 
280 		btrfs_maybe_reset_lockdep_class(root, eb);
281 		btrfs_tree_read_lock(eb);
282 		if (eb == root->node)
283 			break;
284 		btrfs_tree_read_unlock(eb);
285 		free_extent_buffer(eb);
286 	}
287 	return eb;
288 }
289 
290 /*
291  * Loop around taking references on and locking the root node of the tree in
292  * nowait mode until we end up with a lock on the root node or returning to
293  * avoid blocking.
294  *
295  * Return: root extent buffer with read lock held or -EAGAIN.
296  */
297 struct extent_buffer *btrfs_try_read_lock_root_node(struct btrfs_root *root)
298 {
299 	struct extent_buffer *eb;
300 
301 	while (1) {
302 		eb = btrfs_root_node(root);
303 		if (!btrfs_try_tree_read_lock(eb)) {
304 			free_extent_buffer(eb);
305 			return ERR_PTR(-EAGAIN);
306 		}
307 		if (eb == root->node)
308 			break;
309 		btrfs_tree_read_unlock(eb);
310 		free_extent_buffer(eb);
311 	}
312 	return eb;
313 }
314 
315 /*
316  * DREW locks
317  * ==========
318  *
319  * DREW stands for double-reader-writer-exclusion lock. It's used in situation
320  * where you want to provide A-B exclusion but not AA or BB.
321  *
322  * Currently implementation gives more priority to reader. If a reader and a
323  * writer both race to acquire their respective sides of the lock the writer
324  * would yield its lock as soon as it detects a concurrent reader. Additionally
325  * if there are pending readers no new writers would be allowed to come in and
326  * acquire the lock.
327  */
328 
329 void btrfs_drew_lock_init(struct btrfs_drew_lock *lock)
330 {
331 	atomic_set(&lock->readers, 0);
332 	atomic_set(&lock->writers, 0);
333 	init_waitqueue_head(&lock->pending_readers);
334 	init_waitqueue_head(&lock->pending_writers);
335 }
336 
337 /* Return true if acquisition is successful, false otherwise */
338 bool btrfs_drew_try_write_lock(struct btrfs_drew_lock *lock)
339 {
340 	if (atomic_read(&lock->readers))
341 		return false;
342 
343 	atomic_inc(&lock->writers);
344 
345 	/* Ensure writers count is updated before we check for pending readers */
346 	smp_mb__after_atomic();
347 	if (atomic_read(&lock->readers)) {
348 		btrfs_drew_write_unlock(lock);
349 		return false;
350 	}
351 
352 	return true;
353 }
354 
355 void btrfs_drew_write_lock(struct btrfs_drew_lock *lock)
356 {
357 	while (true) {
358 		if (btrfs_drew_try_write_lock(lock))
359 			return;
360 		wait_event(lock->pending_writers, !atomic_read(&lock->readers));
361 	}
362 }
363 
364 void btrfs_drew_write_unlock(struct btrfs_drew_lock *lock)
365 {
366 	atomic_dec(&lock->writers);
367 	cond_wake_up(&lock->pending_readers);
368 }
369 
370 void btrfs_drew_read_lock(struct btrfs_drew_lock *lock)
371 {
372 	atomic_inc(&lock->readers);
373 
374 	/*
375 	 * Ensure the pending reader count is perceieved BEFORE this reader
376 	 * goes to sleep in case of active writers. This guarantees new writers
377 	 * won't be allowed and that the current reader will be woken up when
378 	 * the last active writer finishes its jobs.
379 	 */
380 	smp_mb__after_atomic();
381 
382 	wait_event(lock->pending_readers, atomic_read(&lock->writers) == 0);
383 }
384 
385 void btrfs_drew_read_unlock(struct btrfs_drew_lock *lock)
386 {
387 	/*
388 	 * atomic_dec_and_test implies a full barrier, so woken up writers
389 	 * are guaranteed to see the decrement
390 	 */
391 	if (atomic_dec_and_test(&lock->readers))
392 		wake_up(&lock->pending_writers);
393 }
394