xref: /linux/Documentation/power/pm_qos_interface.rst (revision 7fc2cd2e4b398c57c9cf961cfea05eadbf34c05c)
1===============================
2PM Quality Of Service Interface
3===============================
4
5This interface provides a kernel and user mode interface for registering
6performance expectations by drivers, subsystems and user space applications on
7one of the parameters.
8
9Two different PM QoS frameworks are available:
10 * CPU latency QoS.
11 * The per-device PM QoS framework provides the API to manage the
12   per-device latency constraints and PM QoS flags.
13
14The latency unit used in the PM QoS framework is the microsecond (usec).
15
16
171. PM QoS framework
18===================
19
20A global list of CPU latency QoS requests is maintained along with an aggregated
21(effective) target value.  The aggregated target value is updated with changes
22to the request list or elements of the list.  For CPU latency QoS, the
23aggregated target value is simply the min of the request values held in the list
24elements.
25
26Note: the aggregated target value is implemented as an atomic variable so that
27reading the aggregated value does not require any locking mechanism.
28
29From kernel space the use of this interface is simple:
30
31void cpu_latency_qos_add_request(handle, target_value):
32  Will insert an element into the CPU latency QoS list with the target value.
33  Upon change to this list the new target is recomputed and any registered
34  notifiers are called only if the target value is now different.
35  Clients of PM QoS need to save the returned handle for future use in other
36  PM QoS API functions.
37
38void cpu_latency_qos_update_request(handle, new_target_value):
39  Will update the list element pointed to by the handle with the new target
40  value and recompute the new aggregated target, calling the notification tree
41  if the target is changed.
42
43void cpu_latency_qos_remove_request(handle):
44  Will remove the element.  After removal it will update the aggregate target
45  and call the notification tree if the target was changed as a result of
46  removing the request.
47
48int cpu_latency_qos_limit():
49  Returns the aggregated value for the CPU latency QoS.
50
51int cpu_latency_qos_request_active(handle):
52  Returns if the request is still active, i.e. it has not been removed from the
53  CPU latency QoS list.
54
55
56From user space:
57
58The infrastructure exposes two separate device nodes, /dev/cpu_dma_latency for
59the CPU latency QoS and /dev/cpu_wakeup_latency for the CPU system wakeup
60latency QoS.
61
62Only processes can register a PM QoS request.  To provide for automatic
63cleanup of a process, the interface requires the process to register its
64parameter requests as follows.
65
66To register the default PM QoS target for the CPU latency QoS, the process must
67open /dev/cpu_dma_latency.  To register a CPU system wakeup QoS limit, the
68process must open /dev/cpu_wakeup_latency.
69
70As long as the device node is held open that process has a registered
71request on the parameter.
72
73To change the requested target value, the process needs to write an s32 value to
74the open device node.  Alternatively, it can write a hex string for the value
75using the 10 char long format e.g. "0x12345678".
76
77To remove the user mode request for a target value simply close the device
78node.
79
80
812. PM QoS per-device latency and flags framework
82================================================
83
84For each device, there are three lists of PM QoS requests. Two of them are
85maintained along with the aggregated targets of resume latency and active
86state latency tolerance (in microseconds) and the third one is for PM QoS flags.
87Values are updated in response to changes of the request list.
88
89The target values of resume latency and active state latency tolerance are
90simply the minimum of the request values held in the parameter list elements.
91The PM QoS flags aggregate value is a gather (bitwise OR) of all list elements'
92values.  One device PM QoS flag is defined currently: PM_QOS_FLAG_NO_POWER_OFF.
93
94Note: The aggregated target values are implemented in such a way that reading
95the aggregated value does not require any locking mechanism.
96
97
98From kernel mode the use of this interface is the following:
99
100int dev_pm_qos_add_request(device, handle, type, value):
101  Will insert an element into the list for that identified device with the
102  target value.  Upon change to this list the new target is recomputed and any
103  registered notifiers are called only if the target value is now different.
104  Clients of dev_pm_qos need to save the handle for future use in other
105  dev_pm_qos API functions.
106
107int dev_pm_qos_update_request(handle, new_value):
108  Will update the list element pointed to by the handle with the new target
109  value and recompute the new aggregated target, calling the notification
110  trees if the target is changed.
111
112int dev_pm_qos_remove_request(handle):
113  Will remove the element.  After removal it will update the aggregate target
114  and call the notification trees if the target was changed as a result of
115  removing the request.
116
117s32 dev_pm_qos_read_value(device, type):
118  Returns the aggregated value for a given device's constraints list.
119
120enum pm_qos_flags_status dev_pm_qos_flags(device, mask)
121  Check PM QoS flags of the given device against the given mask of flags.
122  The meaning of the return values is as follows:
123
124	PM_QOS_FLAGS_ALL:
125		All flags from the mask are set
126	PM_QOS_FLAGS_SOME:
127		Some flags from the mask are set
128	PM_QOS_FLAGS_NONE:
129		No flags from the mask are set
130	PM_QOS_FLAGS_UNDEFINED:
131		The device's PM QoS structure has not been initialized
132		or the list of requests is empty.
133
134int dev_pm_qos_add_ancestor_request(dev, handle, type, value)
135  Add a PM QoS request for the first direct ancestor of the given device whose
136  power.ignore_children flag is unset (for DEV_PM_QOS_RESUME_LATENCY requests)
137  or whose power.set_latency_tolerance callback pointer is not NULL (for
138  DEV_PM_QOS_LATENCY_TOLERANCE requests).
139
140int dev_pm_qos_expose_latency_limit(device, value)
141  Add a request to the device's PM QoS list of resume latency constraints and
142  create a sysfs attribute pm_qos_resume_latency_us under the device's power
143  directory allowing user space to manipulate that request.
144
145void dev_pm_qos_hide_latency_limit(device)
146  Drop the request added by dev_pm_qos_expose_latency_limit() from the device's
147  PM QoS list of resume latency constraints and remove sysfs attribute
148  pm_qos_resume_latency_us from the device's power directory.
149
150int dev_pm_qos_expose_flags(device, value)
151  Add a request to the device's PM QoS list of flags and create sysfs attribute
152  pm_qos_no_power_off under the device's power directory allowing user space to
153  change the value of the PM_QOS_FLAG_NO_POWER_OFF flag.
154
155void dev_pm_qos_hide_flags(device)
156  Drop the request added by dev_pm_qos_expose_flags() from the device's PM QoS
157  list of flags and remove sysfs attribute pm_qos_no_power_off from the device's
158  power directory.
159
160Notification mechanisms:
161
162The per-device PM QoS framework has a per-device notification tree.
163
164int dev_pm_qos_add_notifier(device, notifier, type):
165  Adds a notification callback function for the device for a particular request
166  type.
167
168  The callback is called when the aggregated value of the device constraints
169  list is changed.
170
171int dev_pm_qos_remove_notifier(device, notifier, type):
172  Removes the notification callback function for the device.
173
174
175Active state latency tolerance
176^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
177
178This device PM QoS type is used to support systems in which hardware may switch
179to energy-saving operation modes on the fly.  In those systems, if the operation
180mode chosen by the hardware attempts to save energy in an overly aggressive way,
181it may cause excess latencies to be visible to software, causing it to miss
182certain protocol requirements or target frame or sample rates etc.
183
184If there is a latency tolerance control mechanism for a given device available
185to software, the .set_latency_tolerance callback in that device's dev_pm_info
186structure should be populated.  The routine pointed to by it is should implement
187whatever is necessary to transfer the effective requirement value to the
188hardware.
189
190Whenever the effective latency tolerance changes for the device, its
191.set_latency_tolerance() callback will be executed and the effective value will
192be passed to it.  If that value is negative, which means that the list of
193latency tolerance requirements for the device is empty, the callback is expected
194to switch the underlying hardware latency tolerance control mechanism to an
195autonomous mode if available.  If that value is PM_QOS_LATENCY_ANY, in turn, and
196the hardware supports a special "no requirement" setting, the callback is
197expected to use it.  That allows software to prevent the hardware from
198automatically updating the device's latency tolerance in response to its power
199state changes (e.g. during transitions from D3cold to D0), which generally may
200be done in the autonomous latency tolerance control mode.
201
202If .set_latency_tolerance() is present for the device, sysfs attribute
203pm_qos_latency_tolerance_us will be present in the devivce's power directory.
204Then, user space can use that attribute to specify its latency tolerance
205requirement for the device, if any.  Writing "any" to it means "no requirement,
206but do not let the hardware control latency tolerance" and writing "auto" to it
207allows the hardware to be switched to the autonomous mode if there are no other
208requirements from the kernel side in the device's list.
209
210Kernel code can use the functions described above along with the
211DEV_PM_QOS_LATENCY_TOLERANCE device PM QoS type to add, remove and update
212latency tolerance requirements for devices.
213