fs/btrfs/locking.c

   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
  16 /*
  17  * Extent buffer locking
  18  * =====================
  19  *
  20  * We use a rw_semaphore for tree locking, and the semantics are exactly the
  21  * same:
  22  *
  23  * - reader/writer exclusion
  24  * - writer/writer exclusion
  25  * - reader/reader sharing
  26  * - try-lock semantics for readers and writers
  27  *
  28  * The rwsem implementation does opportunistic spinning which reduces number of
  29  * times the locking task needs to sleep.
  30  */
  31
  32 /*
  33  * __btrfs_tree_read_lock - lock extent buffer for read
  34  * @eb:         the eb to be locked
  35  * @nest:       the nesting level to be used for lockdep
  36  *
  37  * This takes the read lock on the extent buffer, using the specified nesting
  38  * level for lockdep purposes.
  39  */
  40 void __btrfs_tree_read_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest)
  41 {
  42         u64 start_ns = 0;
  43
  44         if (trace_btrfs_tree_read_lock_enabled())
  45                 start_ns = ktime_get_ns();
  46
  47         down_read_nested(&eb->lock, nest);
  48         eb->lock_owner = current->pid;
  49         trace_btrfs_tree_read_lock(eb, start_ns);
  50 }
  51
  52 void btrfs_tree_read_lock(struct extent_buffer *eb)
  53 {
  54         __btrfs_tree_read_lock(eb, BTRFS_NESTING_NORMAL);
  55 }
  56
  57 /*
  58  * Try-lock for read.
  59  *
  60  * Retrun 1 if the rwlock has been taken, 0 otherwise
  61  */
  62 int btrfs_try_tree_read_lock(struct extent_buffer *eb)
  63 {
  64         if (down_read_trylock(&eb->lock)) {
  65                 eb->lock_owner = current->pid;
  66                 trace_btrfs_try_tree_read_lock(eb);
  67                 return 1;
  68         }
  69         return 0;
  70 }
  71
  72 /*
  73  * Try-lock for write.
  74  *
  75  * Retrun 1 if the rwlock has been taken, 0 otherwise
  76  */
  77 int btrfs_try_tree_write_lock(struct extent_buffer *eb)
  78 {
  79         if (down_write_trylock(&eb->lock)) {
  80                 eb->lock_owner = current->pid;
  81                 trace_btrfs_try_tree_write_lock(eb);
  82                 return 1;
  83         }
  84         return 0;
  85 }
  86
  87 /*
  88  * Release read lock.
  89  */
  90 void btrfs_tree_read_unlock(struct extent_buffer *eb)
  91 {
  92         trace_btrfs_tree_read_unlock(eb);
  93         eb->lock_owner = 0;
  94         up_read(&eb->lock);
  95 }
  96
  97 /*
  98  * __btrfs_tree_lock - lock eb for write
  99  * @eb:         the eb to lock
 100  * @nest:       the nesting to use for the lock
 101  *
 102  * Returns with the eb->lock write locked.
 103  */
 104 void __btrfs_tree_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest)
 105         __acquires(&eb->lock)
 106 {
 107         u64 start_ns = 0;
 108
 109         if (trace_btrfs_tree_lock_enabled())
 110                 start_ns = ktime_get_ns();
 111
 112         down_write_nested(&eb->lock, nest);
 113         eb->lock_owner = current->pid;
 114         trace_btrfs_tree_lock(eb, start_ns);
 115 }
 116
 117 void btrfs_tree_lock(struct extent_buffer *eb)
 118 {
 119         __btrfs_tree_lock(eb, BTRFS_NESTING_NORMAL);
 120 }
 121
 122 /*
 123  * Release the write lock.
 124  */
 125 void btrfs_tree_unlock(struct extent_buffer *eb)
 126 {
 127         trace_btrfs_tree_unlock(eb);
 128         eb->lock_owner = 0;
 129         up_write(&eb->lock);
 130 }
 131
 132 /*
 133  * This releases any locks held in the path starting at level and going all the
 134  * way up to the root.
 135  *
 136  * btrfs_search_slot will keep the lock held on higher nodes in a few corner
 137  * cases, such as COW of the block at slot zero in the node.  This ignores
 138  * those rules, and it should only be called when there are no more updates to
 139  * be done higher up in the tree.
 140  */
 141 void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
 142 {
 143         int i;
 144
 145         if (path->keep_locks)
 146                 return;
 147
 148         for (i = level; i < BTRFS_MAX_LEVEL; i++) {
 149                 if (!path->nodes[i])
 150                         continue;
 151                 if (!path->locks[i])
 152                         continue;
 153                 btrfs_tree_unlock_rw(path->nodes[i], path->locks[i]);
 154                 path->locks[i] = 0;
 155         }
 156 }
 157
 158 /*
 159  * Loop around taking references on and locking the root node of the tree until
 160  * we end up with a lock on the root node.
 161  *
 162  * Return: root extent buffer with write lock held
 163  */
 164 struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
 165 {
 166         struct extent_buffer *eb;
 167
 168         while (1) {
 169                 eb = btrfs_root_node(root);
 170                 btrfs_tree_lock(eb);
 171                 if (eb == root->node)
 172                         break;
 173                 btrfs_tree_unlock(eb);
 174                 free_extent_buffer(eb);
 175         }
 176         return eb;
 177 }
 178
 179 /*
 180  * Loop around taking references on and locking the root node of the tree until
 181  * we end up with a lock on the root node.
 182  *
 183  * Return: root extent buffer with read lock held
 184  */
 185 struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root)
 186 {
 187         struct extent_buffer *eb;
 188
 189         while (1) {
 190                 eb = btrfs_root_node(root);
 191                 btrfs_tree_read_lock(eb);
 192                 if (eb == root->node)
 193                         break;
 194                 btrfs_tree_read_unlock(eb);
 195                 free_extent_buffer(eb);
 196         }
 197         return eb;
 198 }
 199
 200 /*
 201  * DREW locks
 202  * ==========
 203  *
 204  * DREW stands for double-reader-writer-exclusion lock. It's used in situation
 205  * where you want to provide A-B exclusion but not AA or BB.
 206  *
 207  * Currently implementation gives more priority to reader. If a reader and a
 208  * writer both race to acquire their respective sides of the lock the writer
 209  * would yield its lock as soon as it detects a concurrent reader. Additionally
 210  * if there are pending readers no new writers would be allowed to come in and
 211  * acquire the lock.
 212  */
 213
 214 int btrfs_drew_lock_init(struct btrfs_drew_lock *lock)
 215 {
 216         int ret;
 217
 218         ret = percpu_counter_init(&lock->writers, 0, GFP_KERNEL);
 219         if (ret)
 220                 return ret;
 221
 222         atomic_set(&lock->readers, 0);
 223         init_waitqueue_head(&lock->pending_readers);
 224         init_waitqueue_head(&lock->pending_writers);
 225
 226         return 0;
 227 }
 228
 229 void btrfs_drew_lock_destroy(struct btrfs_drew_lock *lock)
 230 {
 231         percpu_counter_destroy(&lock->writers);
 232 }
 233
 234 /* Return true if acquisition is successful, false otherwise */
 235 bool btrfs_drew_try_write_lock(struct btrfs_drew_lock *lock)
 236 {
 237         if (atomic_read(&lock->readers))
 238                 return false;
 239
 240         percpu_counter_inc(&lock->writers);
 241
 242         /* Ensure writers count is updated before we check for pending readers */
 243         smp_mb();
 244         if (atomic_read(&lock->readers)) {
 245                 btrfs_drew_write_unlock(lock);
 246                 return false;
 247         }
 248
 249         return true;
 250 }
 251
 252 void btrfs_drew_write_lock(struct btrfs_drew_lock *lock)
 253 {
 254         while (true) {
 255                 if (btrfs_drew_try_write_lock(lock))
 256                         return;
 257                 wait_event(lock->pending_writers, !atomic_read(&lock->readers));
 258         }
 259 }
 260
 261 void btrfs_drew_write_unlock(struct btrfs_drew_lock *lock)
 262 {
 263         percpu_counter_dec(&lock->writers);
 264         cond_wake_up(&lock->pending_readers);
 265 }
 266
 267 void btrfs_drew_read_lock(struct btrfs_drew_lock *lock)
 268 {
 269         atomic_inc(&lock->readers);
 270
 271         /*
 272          * Ensure the pending reader count is perceieved BEFORE this reader
 273          * goes to sleep in case of active writers. This guarantees new writers
 274          * won't be allowed and that the current reader will be woken up when
 275          * the last active writer finishes its jobs.
 276          */
 277         smp_mb__after_atomic();
 278
 279         wait_event(lock->pending_readers,
 280                    percpu_counter_sum(&lock->writers) == 0);
 281 }
 282
 283 void btrfs_drew_read_unlock(struct btrfs_drew_lock *lock)
 284 {
 285         /*
 286          * atomic_dec_and_test implies a full barrier, so woken up writers
 287          * are guaranteed to see the decrement
 288          */
 289         if (atomic_dec_and_test(&lock->readers))
 290                 wake_up(&lock->pending_writers);
 291 }