1 ==================================
2 RT-mutex subsystem with PI support
3 ==================================
4
5 RT-mutexes with priority inheritance are used
6 which enable pthread_mutex_t priority inherita
7 (PTHREAD_PRIO_INHERIT). [See Documentation/loc
8 about PI-futexes.]
9
10 This technology was developed in the -rt tree
11 pthread_mutex support.
12
13 Basic principles:
14 -----------------
15
16 RT-mutexes extend the semantics of simple mute
17 inheritance protocol.
18
19 A low priority owner of a rt-mutex inherits th
20 priority waiter until the rt-mutex is released
21 boosted owner blocks on a rt-mutex itself it p
22 boosting to the owner of the other rt_mutex it
23 priority boosting is immediately removed once
24 unlocked.
25
26 This approach allows us to shorten the block o
27 mutexes which protect shared resources. Priori
28 magic bullet for poorly designed applications,
29 well-designed applications to use userspace lo
30 an high priority thread, without losing determ
31
32 The enqueueing of the waiters into the rtmutex
33 priority order. For same priorities FIFO order
34 rtmutex, only the top priority waiter is enque
35 priority waiters tree. This tree too queues in
36 the top priority waiter of a task changes (for
37 got a signal), the priority of the owner task
38 priority enqueueing is handled by "pi_waiters"
39
40 RT-mutexes are optimized for fastpath operatio
41 locking overhead when locking an uncontended m
42 without waiters. The optimized fastpath operat
43 support. [If that is not available then the rt
44 is used]
45
46 The state of the rt-mutex is tracked via the o
47 structure:
48
49 lock->owner holds the task_struct pointer of t
50 keep track of the "lock has waiters" state:
51
52 ============ ======= ========================
53 owner bit0 Notes
54 ============ ======= ========================
55 NULL 0 lock is free (fast acqui
56 NULL 1 lock is free and has wai
57 is going to take the loc
58 taskpointer 0 lock is held (fast relea
59 taskpointer 1 lock is held and has wai
60 ============ ======= ========================
61
62 The fast atomic compare exchange based acquire
63 possible when bit 0 of lock->owner is 0.
64
65 .. [1] It also can be a transitional state whe
66 with ->wait_lock is held. To prevent an
67 we need to set the bit0 before looking
68 be NULL in this small time, hence this
69
70 .. [2] There is a small time when bit 0 is set
71 waiters. This can happen when grabbing
72 To prevent a cmpxchg of the owner relea
73 set this bit before looking at the lock
74
75 BTW, there is still technically a "Pending Own
76 that anymore. The pending owner happens to be
77 that has no owner and has been woken up to gra
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.