1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * Copyright (C) 2008 Oracle.  All rights reserved.
4 */
5
6#ifndef BTRFS_LOCKING_H
7#define BTRFS_LOCKING_H
8
9#include <linux/atomic.h>
10#include <linux/wait.h>
11#include <linux/lockdep.h>
12#include <linux/percpu_counter.h>
13#include "extent_io.h"
14#include "locking.h"
15
16struct extent_buffer;
17struct btrfs_path;
18struct btrfs_root;
19
20#define BTRFS_WRITE_LOCK 1
21#define BTRFS_READ_LOCK 2
22
23/*
24 * We are limited in number of subclasses by MAX_LOCKDEP_SUBCLASSES, which at
25 * the time of this patch is 8, which is how many we use.  Keep this in mind if
26 * you decide you want to add another subclass.
27 */
28enum btrfs_lock_nesting {
29	BTRFS_NESTING_NORMAL,
30
31	/*
32	 * When we COW a block we are holding the lock on the original block,
33	 * and since our lockdep maps are rootid+level, this confuses lockdep
34	 * when we lock the newly allocated COW'd block.  Handle this by having
35	 * a subclass for COW'ed blocks so that lockdep doesn't complain.
36	 */
37	BTRFS_NESTING_COW,
38
39	/*
40	 * Oftentimes we need to lock adjacent nodes on the same level while
41	 * still holding the lock on the original node we searched to, such as
42	 * for searching forward or for split/balance.
43	 *
44	 * Because of this we need to indicate to lockdep that this is
45	 * acceptable by having a different subclass for each of these
46	 * operations.
47	 */
48	BTRFS_NESTING_LEFT,
49	BTRFS_NESTING_RIGHT,
50
51	/*
52	 * When splitting we will be holding a lock on the left/right node when
53	 * we need to cow that node, thus we need a new set of subclasses for
54	 * these two operations.
55	 */
56	BTRFS_NESTING_LEFT_COW,
57	BTRFS_NESTING_RIGHT_COW,
58
59	/*
60	 * When splitting we may push nodes to the left or right, but still use
61	 * the subsequent nodes in our path, keeping our locks on those adjacent
62	 * blocks.  Thus when we go to allocate a new split block we've already
63	 * used up all of our available subclasses, so this subclass exists to
64	 * handle this case where we need to allocate a new split block.
65	 */
66	BTRFS_NESTING_SPLIT,
67
68	/*
69	 * When promoting a new block to a root we need to have a special
70	 * subclass so we don't confuse lockdep, as it will appear that we are
71	 * locking a higher level node before a lower level one.  Copying also
72	 * has this problem as it appears we're locking the same block again
73	 * when we make a snapshot of an existing root.
74	 */
75	BTRFS_NESTING_NEW_ROOT,
76
77	/*
78	 * We are limited to MAX_LOCKDEP_SUBLCLASSES number of subclasses, so
79	 * add this in here and add a static_assert to keep us from going over
80	 * the limit.  As of this writing we're limited to 8, and we're
81	 * definitely using 8, hence this check to keep us from messing up in
82	 * the future.
83	 */
84	BTRFS_NESTING_MAX,
85};
86
87enum btrfs_lockdep_trans_states {
88	BTRFS_LOCKDEP_TRANS_COMMIT_PREP,
89	BTRFS_LOCKDEP_TRANS_UNBLOCKED,
90	BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED,
91	BTRFS_LOCKDEP_TRANS_COMPLETED,
92};
93
94/*
95 * Lockdep annotation for wait events.
96 *
97 * @owner:  The struct where the lockdep map is defined
98 * @lock:   The lockdep map corresponding to a wait event
99 *
100 * This macro is used to annotate a wait event. In this case a thread acquires
101 * the lockdep map as writer (exclusive lock) because it has to block until all
102 * the threads that hold the lock as readers signal the condition for the wait
103 * event and release their locks.
104 */
105#define btrfs_might_wait_for_event(owner, lock)					\
106	do {									\
107		rwsem_acquire(&owner->lock##_map, 0, 0, _THIS_IP_);		\
108		rwsem_release(&owner->lock##_map, _THIS_IP_);			\
109	} while (0)
110
111/*
112 * Protection for the resource/condition of a wait event.
113 *
114 * @owner:  The struct where the lockdep map is defined
115 * @lock:   The lockdep map corresponding to a wait event
116 *
117 * Many threads can modify the condition for the wait event at the same time
118 * and signal the threads that block on the wait event. The threads that modify
119 * the condition and do the signaling acquire the lock as readers (shared
120 * lock).
121 */
122#define btrfs_lockdep_acquire(owner, lock)					\
123	rwsem_acquire_read(&owner->lock##_map, 0, 0, _THIS_IP_)
124
125/*
126 * Used after signaling the condition for a wait event to release the lockdep
127 * map held by a reader thread.
128 */
129#define btrfs_lockdep_release(owner, lock)					\
130	rwsem_release(&owner->lock##_map, _THIS_IP_)
131
132/*
133 * Macros for the transaction states wait events, similar to the generic wait
134 * event macros.
135 */
136#define btrfs_might_wait_for_state(owner, i)					\
137	do {									\
138		rwsem_acquire(&owner->btrfs_state_change_map[i], 0, 0, _THIS_IP_); \
139		rwsem_release(&owner->btrfs_state_change_map[i], _THIS_IP_);	\
140	} while (0)
141
142#define btrfs_trans_state_lockdep_acquire(owner, i)				\
143	rwsem_acquire_read(&owner->btrfs_state_change_map[i], 0, 0, _THIS_IP_)
144
145#define btrfs_trans_state_lockdep_release(owner, i)				\
146	rwsem_release(&owner->btrfs_state_change_map[i], _THIS_IP_)
147
148/* Initialization of the lockdep map */
149#define btrfs_lockdep_init_map(owner, lock)					\
150	do {									\
151		static struct lock_class_key lock##_key;			\
152		lockdep_init_map(&owner->lock##_map, #lock, &lock##_key, 0);	\
153	} while (0)
154
155/* Initialization of the transaction states lockdep maps. */
156#define btrfs_state_lockdep_init_map(owner, lock, state)			\
157	do {									\
158		static struct lock_class_key lock##_key;			\
159		lockdep_init_map(&owner->btrfs_state_change_map[state], #lock,	\
160				 &lock##_key, 0);				\
161	} while (0)
162
163static_assert(BTRFS_NESTING_MAX <= MAX_LOCKDEP_SUBCLASSES,
164	      "too many lock subclasses defined");
165
166void __btrfs_tree_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest);
167void btrfs_tree_lock(struct extent_buffer *eb);
168void btrfs_tree_unlock(struct extent_buffer *eb);
169
170void __btrfs_tree_read_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest);
171void btrfs_tree_read_lock(struct extent_buffer *eb);
172void btrfs_tree_read_unlock(struct extent_buffer *eb);
173int btrfs_try_tree_read_lock(struct extent_buffer *eb);
174int btrfs_try_tree_write_lock(struct extent_buffer *eb);
175struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root);
176struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root);
177struct extent_buffer *btrfs_try_read_lock_root_node(struct btrfs_root *root);
178
179#ifdef CONFIG_BTRFS_DEBUG
180static inline void btrfs_assert_tree_write_locked(struct extent_buffer *eb)
181{
182	lockdep_assert_held_write(&eb->lock);
183}
184#else
185static inline void btrfs_assert_tree_write_locked(struct extent_buffer *eb) { }
186#endif
187
188void btrfs_unlock_up_safe(struct btrfs_path *path, int level);
189
190static inline void btrfs_tree_unlock_rw(struct extent_buffer *eb, int rw)
191{
192	if (rw == BTRFS_WRITE_LOCK)
193		btrfs_tree_unlock(eb);
194	else if (rw == BTRFS_READ_LOCK)
195		btrfs_tree_read_unlock(eb);
196	else
197		BUG();
198}
199
200struct btrfs_drew_lock {
201	atomic_t readers;
202	atomic_t writers;
203	wait_queue_head_t pending_writers;
204	wait_queue_head_t pending_readers;
205};
206
207void btrfs_drew_lock_init(struct btrfs_drew_lock *lock);
208void btrfs_drew_write_lock(struct btrfs_drew_lock *lock);
209bool btrfs_drew_try_write_lock(struct btrfs_drew_lock *lock);
210void btrfs_drew_write_unlock(struct btrfs_drew_lock *lock);
211void btrfs_drew_read_lock(struct btrfs_drew_lock *lock);
212void btrfs_drew_read_unlock(struct btrfs_drew_lock *lock);
213
214#ifdef CONFIG_DEBUG_LOCK_ALLOC
215void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, int level);
216void btrfs_maybe_reset_lockdep_class(struct btrfs_root *root, struct extent_buffer *eb);
217#else
218static inline void btrfs_set_buffer_lockdep_class(u64 objectid,
219					struct extent_buffer *eb, int level)
220{
221}
222static inline void btrfs_maybe_reset_lockdep_class(struct btrfs_root *root,
223						   struct extent_buffer *eb)
224{
225}
226#endif
227
228#endif
229