| 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 | |
| 15 | struct extent_buffer; |
| 16 | struct btrfs_path; |
| 17 | struct btrfs_root; |
| 18 | |
| 19 | #define BTRFS_WRITE_LOCK 1 |
| 20 | #define BTRFS_READ_LOCK 2 |
| 21 | |
| 22 | /* |
| 23 | * We are limited in number of subclasses by MAX_LOCKDEP_SUBCLASSES, which at |
| 24 | * the time of this patch is 8, which is how many we use. Keep this in mind if |
| 25 | * you decide you want to add another subclass. |
| 26 | */ |
| 27 | enum btrfs_lock_nesting { |
| 28 | BTRFS_NESTING_NORMAL, |
| 29 | |
| 30 | /* |
| 31 | * When we COW a block we are holding the lock on the original block, |
| 32 | * and since our lockdep maps are rootid+level, this confuses lockdep |
| 33 | * when we lock the newly allocated COW'd block. Handle this by having |
| 34 | * a subclass for COW'ed blocks so that lockdep doesn't complain. |
| 35 | */ |
| 36 | BTRFS_NESTING_COW, |
| 37 | |
| 38 | /* |
| 39 | * Oftentimes we need to lock adjacent nodes on the same level while |
| 40 | * still holding the lock on the original node we searched to, such as |
| 41 | * for searching forward or for split/balance. |
| 42 | * |
| 43 | * Because of this we need to indicate to lockdep that this is |
| 44 | * acceptable by having a different subclass for each of these |
| 45 | * operations. |
| 46 | */ |
| 47 | BTRFS_NESTING_LEFT, |
| 48 | BTRFS_NESTING_RIGHT, |
| 49 | |
| 50 | /* |
| 51 | * When splitting we will be holding a lock on the left/right node when |
| 52 | * we need to cow that node, thus we need a new set of subclasses for |
| 53 | * these two operations. |
| 54 | */ |
| 55 | BTRFS_NESTING_LEFT_COW, |
| 56 | BTRFS_NESTING_RIGHT_COW, |
| 57 | |
| 58 | /* |
| 59 | * When splitting we may push nodes to the left or right, but still use |
| 60 | * the subsequent nodes in our path, keeping our locks on those adjacent |
| 61 | * blocks. Thus when we go to allocate a new split block we've already |
| 62 | * used up all of our available subclasses, so this subclass exists to |
| 63 | * handle this case where we need to allocate a new split block. |
| 64 | */ |
| 65 | BTRFS_NESTING_SPLIT, |
| 66 | |
| 67 | /* |
| 68 | * When promoting a new block to a root we need to have a special |
| 69 | * subclass so we don't confuse lockdep, as it will appear that we are |
| 70 | * locking a higher level node before a lower level one. Copying also |
| 71 | * has this problem as it appears we're locking the same block again |
| 72 | * when we make a snapshot of an existing root. |
| 73 | */ |
| 74 | BTRFS_NESTING_NEW_ROOT, |
| 75 | |
| 76 | /* |
| 77 | * We are limited to MAX_LOCKDEP_SUBCLASSES number of subclasses, so |
| 78 | * add this in here and add a static_assert to keep us from going over |
| 79 | * the limit. As of this writing we're limited to 8, and we're |
| 80 | * definitely using 8, hence this check to keep us from messing up in |
| 81 | * the future. |
| 82 | */ |
| 83 | BTRFS_NESTING_MAX, |
| 84 | }; |
| 85 | |
| 86 | enum btrfs_lockdep_trans_states { |
| 87 | BTRFS_LOCKDEP_TRANS_COMMIT_PREP, |
| 88 | BTRFS_LOCKDEP_TRANS_UNBLOCKED, |
| 89 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED, |
| 90 | BTRFS_LOCKDEP_TRANS_COMPLETED, |
| 91 | }; |
| 92 | |
| 93 | /* |
| 94 | * Lockdep annotation for wait events. |
| 95 | * |
| 96 | * @owner: The struct where the lockdep map is defined |
| 97 | * @lock: The lockdep map corresponding to a wait event |
| 98 | * |
| 99 | * This macro is used to annotate a wait event. In this case a thread acquires |
| 100 | * the lockdep map as writer (exclusive lock) because it has to block until all |
| 101 | * the threads that hold the lock as readers signal the condition for the wait |
| 102 | * event and release their locks. |
| 103 | */ |
| 104 | #define btrfs_might_wait_for_event(owner, lock) \ |
| 105 | do { \ |
| 106 | rwsem_acquire(&owner->lock##_map, 0, 0, _THIS_IP_); \ |
| 107 | rwsem_release(&owner->lock##_map, _THIS_IP_); \ |
| 108 | } while (0) |
| 109 | |
| 110 | /* |
| 111 | * Protection for the resource/condition of a wait event. |
| 112 | * |
| 113 | * @owner: The struct where the lockdep map is defined |
| 114 | * @lock: The lockdep map corresponding to a wait event |
| 115 | * |
| 116 | * Many threads can modify the condition for the wait event at the same time |
| 117 | * and signal the threads that block on the wait event. The threads that modify |
| 118 | * the condition and do the signaling acquire the lock as readers (shared |
| 119 | * lock). |
| 120 | */ |
| 121 | #define btrfs_lockdep_acquire(owner, lock) \ |
| 122 | rwsem_acquire_read(&owner->lock##_map, 0, 0, _THIS_IP_) |
| 123 | |
| 124 | /* |
| 125 | * Used after signaling the condition for a wait event to release the lockdep |
| 126 | * map held by a reader thread. |
| 127 | */ |
| 128 | #define btrfs_lockdep_release(owner, lock) \ |
| 129 | rwsem_release(&owner->lock##_map, _THIS_IP_) |
| 130 | |
| 131 | /* |
| 132 | * Used to account for the fact that when doing io_uring encoded I/O, we can |
| 133 | * return to userspace with the inode lock still held. |
| 134 | */ |
| 135 | #define btrfs_lockdep_inode_acquire(owner, lock) \ |
| 136 | rwsem_acquire_read(&owner->vfs_inode.lock.dep_map, 0, 0, _THIS_IP_) |
| 137 | |
| 138 | #define btrfs_lockdep_inode_release(owner, lock) \ |
| 139 | rwsem_release(&owner->vfs_inode.lock.dep_map, _THIS_IP_) |
| 140 | |
| 141 | /* |
| 142 | * Macros for the transaction states wait events, similar to the generic wait |
| 143 | * event macros. |
| 144 | */ |
| 145 | #define btrfs_might_wait_for_state(owner, i) \ |
| 146 | do { \ |
| 147 | rwsem_acquire(&owner->btrfs_state_change_map[i], 0, 0, _THIS_IP_); \ |
| 148 | rwsem_release(&owner->btrfs_state_change_map[i], _THIS_IP_); \ |
| 149 | } while (0) |
| 150 | |
| 151 | #define btrfs_trans_state_lockdep_acquire(owner, i) \ |
| 152 | rwsem_acquire_read(&owner->btrfs_state_change_map[i], 0, 0, _THIS_IP_) |
| 153 | |
| 154 | #define btrfs_trans_state_lockdep_release(owner, i) \ |
| 155 | rwsem_release(&owner->btrfs_state_change_map[i], _THIS_IP_) |
| 156 | |
| 157 | /* Initialization of the lockdep map */ |
| 158 | #define btrfs_lockdep_init_map(owner, lock) \ |
| 159 | do { \ |
| 160 | static struct lock_class_key lock##_key; \ |
| 161 | lockdep_init_map(&owner->lock##_map, #lock, &lock##_key, 0); \ |
| 162 | } while (0) |
| 163 | |
| 164 | /* Initialization of the transaction states lockdep maps. */ |
| 165 | #define btrfs_state_lockdep_init_map(owner, lock, state) \ |
| 166 | do { \ |
| 167 | static struct lock_class_key lock##_key; \ |
| 168 | lockdep_init_map(&owner->btrfs_state_change_map[state], #lock, \ |
| 169 | &lock##_key, 0); \ |
| 170 | } while (0) |
| 171 | |
| 172 | static_assert(BTRFS_NESTING_MAX <= MAX_LOCKDEP_SUBCLASSES, |
| 173 | "too many lock subclasses defined" ); |
| 174 | |
| 175 | void btrfs_tree_lock_nested(struct extent_buffer *eb, enum btrfs_lock_nesting nest); |
| 176 | |
| 177 | static inline void btrfs_tree_lock(struct extent_buffer *eb) |
| 178 | { |
| 179 | btrfs_tree_lock_nested(eb, nest: BTRFS_NESTING_NORMAL); |
| 180 | } |
| 181 | |
| 182 | void btrfs_tree_unlock(struct extent_buffer *eb); |
| 183 | |
| 184 | void btrfs_tree_read_lock_nested(struct extent_buffer *eb, enum btrfs_lock_nesting nest); |
| 185 | |
| 186 | static inline void btrfs_tree_read_lock(struct extent_buffer *eb) |
| 187 | { |
| 188 | btrfs_tree_read_lock_nested(eb, nest: BTRFS_NESTING_NORMAL); |
| 189 | } |
| 190 | |
| 191 | void btrfs_tree_read_unlock(struct extent_buffer *eb); |
| 192 | bool btrfs_try_tree_read_lock(struct extent_buffer *eb); |
| 193 | struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root); |
| 194 | struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root); |
| 195 | struct extent_buffer *btrfs_try_read_lock_root_node(struct btrfs_root *root); |
| 196 | |
| 197 | #ifdef CONFIG_BTRFS_DEBUG |
| 198 | static inline void btrfs_assert_tree_write_locked(struct extent_buffer *eb) |
| 199 | { |
| 200 | lockdep_assert_held_write(&eb->lock); |
| 201 | } |
| 202 | static inline void btrfs_assert_tree_read_locked(struct extent_buffer *eb) |
| 203 | { |
| 204 | lockdep_assert_held_read(&eb->lock); |
| 205 | } |
| 206 | #else |
| 207 | static inline void btrfs_assert_tree_write_locked(struct extent_buffer *eb) { } |
| 208 | static inline void btrfs_assert_tree_read_locked(struct extent_buffer *eb) { } |
| 209 | #endif |
| 210 | |
| 211 | void btrfs_unlock_up_safe(struct btrfs_path *path, int level); |
| 212 | |
| 213 | static inline void btrfs_tree_unlock_rw(struct extent_buffer *eb, int rw) |
| 214 | { |
| 215 | if (rw == BTRFS_WRITE_LOCK) |
| 216 | btrfs_tree_unlock(eb); |
| 217 | else if (rw == BTRFS_READ_LOCK) |
| 218 | btrfs_tree_read_unlock(eb); |
| 219 | else |
| 220 | BUG(); |
| 221 | } |
| 222 | |
| 223 | struct btrfs_drew_lock { |
| 224 | atomic_t readers; |
| 225 | atomic_t writers; |
| 226 | wait_queue_head_t pending_writers; |
| 227 | wait_queue_head_t pending_readers; |
| 228 | }; |
| 229 | |
| 230 | void btrfs_drew_lock_init(struct btrfs_drew_lock *lock); |
| 231 | void btrfs_drew_write_lock(struct btrfs_drew_lock *lock); |
| 232 | bool btrfs_drew_try_write_lock(struct btrfs_drew_lock *lock); |
| 233 | void btrfs_drew_write_unlock(struct btrfs_drew_lock *lock); |
| 234 | void btrfs_drew_read_lock(struct btrfs_drew_lock *lock); |
| 235 | void btrfs_drew_read_unlock(struct btrfs_drew_lock *lock); |
| 236 | |
| 237 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 238 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, int level); |
| 239 | void btrfs_maybe_reset_lockdep_class(struct btrfs_root *root, struct extent_buffer *eb); |
| 240 | #else |
| 241 | static inline void btrfs_set_buffer_lockdep_class(u64 objectid, |
| 242 | struct extent_buffer *eb, int level) |
| 243 | { |
| 244 | } |
| 245 | static inline void btrfs_maybe_reset_lockdep_class(struct btrfs_root *root, |
| 246 | struct extent_buffer *eb) |
| 247 | { |
| 248 | } |
| 249 | #endif |
| 250 | |
| 251 | #endif |
| 252 | |