tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
PM / QoS: Update Documentation for the pm_qos and dev_pm_qos frameworks
Update the documentation for the recently updated pm_qos API, kernel
and user space. Add documentation for the per-device PM QoS
(dev_pm_qos) framework API.
Signed-off-by: Jean Pihet <j-pihet@ti.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
authored by
Jean Pihet
and committed by
Rafael J. Wysocki
e3cba324
1a9a9152
+87
-5
+87
-5
Documentation/power/pm_qos_interface.txt
+87
-5
Documentation/power/pm_qos_interface.txt
···
4
4
performance expectations by drivers, subsystems and user space applications on
5
5
one of the parameters.
6
6
7
-
Currently we have {cpu_dma_latency, network_latency, network_throughput} as the
8
-
initial set of pm_qos parameters.
7
+
Two different PM QoS frameworks are available:
8
+
1. PM QoS classes for cpu_dma_latency, network_latency, network_throughput.
9
+
2. the per-device PM QoS framework provides the API to manage the per-device latency
10
+
constraints.
9
11
10
12
Each parameters have defined units:
11
13
* latency: usec
12
14
* timeout: usec
13
15
* throughput: kbs (kilo bit / sec)
16
+
17
+
18
+
1. PM QoS framework
14
19
15
20
The infrastructure exposes multiple misc device nodes one per implemented
16
21
parameter. The set of parameters implement is defined by pm_qos_power_init()
···
28
23
changes to the request list or elements of the list. Typically the
29
24
aggregated target value is simply the max or min of the request values held
30
25
in the parameter list elements.
26
+
Note: the aggregated target value is implemented as an atomic variable so that
27
+
reading the aggregated value does not require any locking mechanism.
28
+
31
29
32
30
From kernel mode the use of this interface is simple:
33
31
34
-
handle = pm_qos_add_request(param_class, target_value):
35
-
Will insert an element into the list for that identified PM_QOS class with the
32
+
void pm_qos_add_request(handle, param_class, target_value):
33
+
Will insert an element into the list for that identified PM QoS class with the
36
34
target value. Upon change to this list the new target is recomputed and any
37
35
registered notifiers are called only if the target value is now different.
38
-
Clients of pm_qos need to save the returned handle.
36
+
Clients of pm_qos need to save the returned handle for future use in other
37
+
pm_qos API functions.
39
38
40
39
void pm_qos_update_request(handle, new_target_value):
41
40
Will update the list element pointed to by the handle with the new target value
···
50
41
Will remove the element. After removal it will update the aggregate target and
51
42
call the notification tree if the target was changed as a result of removing
52
43
the request.
44
+
45
+
int pm_qos_request(param_class):
46
+
Returns the aggregated value for a given PM QoS class.
47
+
48
+
int pm_qos_request_active(handle):
49
+
Returns if the request is still active, i.e. it has not been removed from a
50
+
PM QoS class constraints list.
51
+
52
+
int pm_qos_add_notifier(param_class, notifier):
53
+
Adds a notification callback function to the PM QoS class. The callback is
54
+
called when the aggregated value for the PM QoS class is changed.
55
+
56
+
int pm_qos_remove_notifier(int param_class, notifier):
57
+
Removes the notification callback function for the PM QoS class.
53
58
54
59
55
60
From user mode:
···
86
63
node.
87
64
88
65
66
+
2. PM QoS per-device latency framework
67
+
68
+
For each device a list of performance requests is maintained along with
69
+
an aggregated target value. The aggregated target value is updated with
70
+
changes to the request list or elements of the list. Typically the
71
+
aggregated target value is simply the max or min of the request values held
72
+
in the parameter list elements.
73
+
Note: the aggregated target value is implemented as an atomic variable so that
74
+
reading the aggregated value does not require any locking mechanism.
75
+
76
+
77
+
From kernel mode the use of this interface is the following:
78
+
79
+
int dev_pm_qos_add_request(device, handle, value):
80
+
Will insert an element into the list for that identified device with the
81
+
target value. Upon change to this list the new target is recomputed and any
82
+
registered notifiers are called only if the target value is now different.
83
+
Clients of dev_pm_qos need to save the handle for future use in other
84
+
dev_pm_qos API functions.
85
+
86
+
int dev_pm_qos_update_request(handle, new_value):
87
+
Will update the list element pointed to by the handle with the new target value
88
+
and recompute the new aggregated target, calling the notification trees if the
89
+
target is changed.
90
+
91
+
int dev_pm_qos_remove_request(handle):
92
+
Will remove the element. After removal it will update the aggregate target and
93
+
call the notification trees if the target was changed as a result of removing
94
+
the request.
95
+
96
+
s32 dev_pm_qos_read_value(device):
97
+
Returns the aggregated value for a given device's constraints list.
98
+
99
+
100
+
Notification mechanisms:
101
+
The per-device PM QoS framework has 2 different and distinct notification trees:
102
+
a per-device notification tree and a global notification tree.
103
+
104
+
int dev_pm_qos_add_notifier(device, notifier):
105
+
Adds a notification callback function for the device.
106
+
The callback is called when the aggregated value of the device constraints list
107
+
is changed.
108
+
109
+
int dev_pm_qos_remove_notifier(device, notifier):
110
+
Removes the notification callback function for the device.
111
+
112
+
int dev_pm_qos_add_global_notifier(notifier):
113
+
Adds a notification callback function in the global notification tree of the
114
+
framework.
115
+
The callback is called when the aggregated value for any device is changed.
116
+
117
+
int dev_pm_qos_remove_global_notifier(notifier):
118
+
Removes the notification callback function from the global notification tree
119
+
of the framework.
120
+
121
+
122
+
From user mode:
123
+
No API for user space access to the per-device latency constraints is provided
124
+
yet - still under discussion.
89
125