blob: 7d3c82431909dd8120322e2360ce32cbd93f87e5 [file] [log] [blame]
Mikulas Patocka62ac6652012-09-26 07:46:43 +02001Percpu rw semaphores
2--------------------
3
4Percpu rw semaphores is a new read-write semaphore design that is
5optimized for locking for reading.
6
7The problem with traditional read-write semaphores is that when multiple
8cores take the lock for reading, the cache line containing the semaphore
9is bouncing between L1 caches of the cores, causing performance
10degradation.
11
Mikulas Patockae6b5c082012-09-26 19:56:15 +020012Locking for reading is very fast, it uses RCU and it avoids any atomic
Mikulas Patocka62ac6652012-09-26 07:46:43 +020013instruction in the lock and unlock path. On the other hand, locking for
14writing is very expensive, it calls synchronize_rcu() that can take
Mikulas Patockae6b5c082012-09-26 19:56:15 +020015hundreds of milliseconds.
Mikulas Patocka62ac6652012-09-26 07:46:43 +020016
17The lock is declared with "struct percpu_rw_semaphore" type.
18The lock is initialized percpu_init_rwsem, it returns 0 on success and
19-ENOMEM on allocation failure.
20The lock must be freed with percpu_free_rwsem to avoid memory leak.
21
22The lock is locked for read with percpu_down_read, percpu_up_read and
23for write with percpu_down_write, percpu_up_write.
24
25The idea of using RCU for optimized rw-lock was introduced by
26Eric Dumazet <eric.dumazet@gmail.com>.
27The code was written by Mikulas Patocka <mpatocka@redhat.com>