blob: 129f7c0e14839837e1ff4c1adc53544548a7d9a9 [file] [log] [blame]
Kyle Swenson8d8f6542021-03-15 11:02:55 -06001PM Quality Of Service Interface.
2
3This interface provides a kernel and user mode interface for registering
4performance expectations by drivers, subsystems and user space applications on
5one of the parameters.
6
7Two different PM QoS frameworks are available:
81. PM QoS classes for cpu_dma_latency, network_latency, network_throughput,
9memory_bandwidth.
102. the per-device PM QoS framework provides the API to manage the per-device latency
11constraints and PM QoS flags.
12
13Each parameters have defined units:
14 * latency: usec
15 * timeout: usec
16 * throughput: kbs (kilo bit / sec)
17 * memory bandwidth: mbs (mega bit / sec)
18
19
201. PM QoS framework
21
22The infrastructure exposes multiple misc device nodes one per implemented
23parameter. The set of parameters implement is defined by pm_qos_power_init()
24and pm_qos_params.h. This is done because having the available parameters
25being runtime configurable or changeable from a driver was seen as too easy to
26abuse.
27
28For each parameter a list of performance requests is maintained along with
29an aggregated target value. The aggregated target value is updated with
30changes to the request list or elements of the list. Typically the
31aggregated target value is simply the max or min of the request values held
32in the parameter list elements.
33Note: the aggregated target value is implemented as an atomic variable so that
34reading the aggregated value does not require any locking mechanism.
35
36
37From kernel mode the use of this interface is simple:
38
39void pm_qos_add_request(handle, param_class, target_value):
40Will insert an element into the list for that identified PM QoS class with the
41target value. Upon change to this list the new target is recomputed and any
42registered notifiers are called only if the target value is now different.
43Clients of pm_qos need to save the returned handle for future use in other
44pm_qos API functions.
45
46void pm_qos_update_request(handle, new_target_value):
47Will update the list element pointed to by the handle with the new target value
48and recompute the new aggregated target, calling the notification tree if the
49target is changed.
50
51void pm_qos_remove_request(handle):
52Will remove the element. After removal it will update the aggregate target and
53call the notification tree if the target was changed as a result of removing
54the request.
55
56int pm_qos_request(param_class):
57Returns the aggregated value for a given PM QoS class.
58
59int pm_qos_request_active(handle):
60Returns if the request is still active, i.e. it has not been removed from a
61PM QoS class constraints list.
62
63int pm_qos_add_notifier(param_class, notifier):
64Adds a notification callback function to the PM QoS class. The callback is
65called when the aggregated value for the PM QoS class is changed.
66
67int pm_qos_remove_notifier(int param_class, notifier):
68Removes the notification callback function for the PM QoS class.
69
70
71From user mode:
72Only processes can register a pm_qos request. To provide for automatic
73cleanup of a process, the interface requires the process to register its
74parameter requests in the following way:
75
76To register the default pm_qos target for the specific parameter, the process
77must open one of /dev/[cpu_dma_latency, network_latency, network_throughput]
78
79As long as the device node is held open that process has a registered
80request on the parameter.
81
82To change the requested target value the process needs to write an s32 value to
83the open device node. Alternatively the user mode program could write a hex
84string for the value using 10 char long format e.g. "0x12345678". This
85translates to a pm_qos_update_request call.
86
87To remove the user mode request for a target value simply close the device
88node.
89
90
912. PM QoS per-device latency and flags framework
92
93For each device, there are three lists of PM QoS requests. Two of them are
94maintained along with the aggregated targets of resume latency and active
95state latency tolerance (in microseconds) and the third one is for PM QoS flags.
96Values are updated in response to changes of the request list.
97
98The target values of resume latency and active state latency tolerance are
99simply the minimum of the request values held in the parameter list elements.
100The PM QoS flags aggregate value is a gather (bitwise OR) of all list elements'
101values. Two device PM QoS flags are defined currently: PM_QOS_FLAG_NO_POWER_OFF
102and PM_QOS_FLAG_REMOTE_WAKEUP.
103
104Note: The aggregated target values are implemented in such a way that reading
105the aggregated value does not require any locking mechanism.
106
107
108From kernel mode the use of this interface is the following:
109
110int dev_pm_qos_add_request(device, handle, type, value):
111Will insert an element into the list for that identified device with the
112target value. Upon change to this list the new target is recomputed and any
113registered notifiers are called only if the target value is now different.
114Clients of dev_pm_qos need to save the handle for future use in other
115dev_pm_qos API functions.
116
117int dev_pm_qos_update_request(handle, new_value):
118Will update the list element pointed to by the handle with the new target value
119and recompute the new aggregated target, calling the notification trees if the
120target is changed.
121
122int dev_pm_qos_remove_request(handle):
123Will remove the element. After removal it will update the aggregate target and
124call the notification trees if the target was changed as a result of removing
125the request.
126
127s32 dev_pm_qos_read_value(device):
128Returns the aggregated value for a given device's constraints list.
129
130enum pm_qos_flags_status dev_pm_qos_flags(device, mask)
131Check PM QoS flags of the given device against the given mask of flags.
132The meaning of the return values is as follows:
133 PM_QOS_FLAGS_ALL: All flags from the mask are set
134 PM_QOS_FLAGS_SOME: Some flags from the mask are set
135 PM_QOS_FLAGS_NONE: No flags from the mask are set
136 PM_QOS_FLAGS_UNDEFINED: The device's PM QoS structure has not been
137 initialized or the list of requests is empty.
138
139int dev_pm_qos_add_ancestor_request(dev, handle, type, value)
140Add a PM QoS request for the first direct ancestor of the given device whose
141power.ignore_children flag is unset (for DEV_PM_QOS_RESUME_LATENCY requests)
142or whose power.set_latency_tolerance callback pointer is not NULL (for
143DEV_PM_QOS_LATENCY_TOLERANCE requests).
144
145int dev_pm_qos_expose_latency_limit(device, value)
146Add a request to the device's PM QoS list of resume latency constraints and
147create a sysfs attribute pm_qos_resume_latency_us under the device's power
148directory allowing user space to manipulate that request.
149
150void dev_pm_qos_hide_latency_limit(device)
151Drop the request added by dev_pm_qos_expose_latency_limit() from the device's
152PM QoS list of resume latency constraints and remove sysfs attribute
153pm_qos_resume_latency_us from the device's power directory.
154
155int dev_pm_qos_expose_flags(device, value)
156Add a request to the device's PM QoS list of flags and create sysfs attributes
157pm_qos_no_power_off and pm_qos_remote_wakeup under the device's power directory
158allowing user space to change these flags' value.
159
160void dev_pm_qos_hide_flags(device)
161Drop the request added by dev_pm_qos_expose_flags() from the device's PM QoS list
162of flags and remove sysfs attributes pm_qos_no_power_off and pm_qos_remote_wakeup
163under the device's power directory.
164
165Notification mechanisms:
166The per-device PM QoS framework has 2 different and distinct notification trees:
167a per-device notification tree and a global notification tree.
168
169int dev_pm_qos_add_notifier(device, notifier):
170Adds a notification callback function for the device.
171The callback is called when the aggregated value of the device constraints list
172is changed (for resume latency device PM QoS only).
173
174int dev_pm_qos_remove_notifier(device, notifier):
175Removes the notification callback function for the device.
176
177int dev_pm_qos_add_global_notifier(notifier):
178Adds a notification callback function in the global notification tree of the
179framework.
180The callback is called when the aggregated value for any device is changed
181(for resume latency device PM QoS only).
182
183int dev_pm_qos_remove_global_notifier(notifier):
184Removes the notification callback function from the global notification tree
185of the framework.
186
187
188Active state latency tolerance
189
190This device PM QoS type is used to support systems in which hardware may switch
191to energy-saving operation modes on the fly. In those systems, if the operation
192mode chosen by the hardware attempts to save energy in an overly aggressive way,
193it may cause excess latencies to be visible to software, causing it to miss
194certain protocol requirements or target frame or sample rates etc.
195
196If there is a latency tolerance control mechanism for a given device available
197to software, the .set_latency_tolerance callback in that device's dev_pm_info
198structure should be populated. The routine pointed to by it is should implement
199whatever is necessary to transfer the effective requirement value to the
200hardware.
201
202Whenever the effective latency tolerance changes for the device, its
203.set_latency_tolerance() callback will be executed and the effective value will
204be passed to it. If that value is negative, which means that the list of
205latency tolerance requirements for the device is empty, the callback is expected
206to switch the underlying hardware latency tolerance control mechanism to an
207autonomous mode if available. If that value is PM_QOS_LATENCY_ANY, in turn, and
208the hardware supports a special "no requirement" setting, the callback is
209expected to use it. That allows software to prevent the hardware from
210automatically updating the device's latency tolerance in response to its power
211state changes (e.g. during transitions from D3cold to D0), which generally may
212be done in the autonomous latency tolerance control mode.
213
214If .set_latency_tolerance() is present for the device, sysfs attribute
215pm_qos_latency_tolerance_us will be present in the devivce's power directory.
216Then, user space can use that attribute to specify its latency tolerance
217requirement for the device, if any. Writing "any" to it means "no requirement,
218but do not let the hardware control latency tolerance" and writing "auto" to it
219allows the hardware to be switched to the autonomous mode if there are no other
220requirements from the kernel side in the device's list.
221
222Kernel code can use the functions described above along with the
223DEV_PM_QOS_LATENCY_TOLERANCE device PM QoS type to add, remove and update
224latency tolerance requirements for devices.