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: 101. PM QoS classes for cpu_dma_latency 112. The per-device PM QoS framework provides the API to manage the 12 per-device latency constraints and PM QoS flags. 13 14Each parameters have defined units: 15 16 * latency: usec 17 * timeout: usec 18 * throughput: kbs (kilo bit / sec) 19 * memory bandwidth: mbs (mega bit / sec) 20 21 221. PM QoS framework 23=================== 24 25The infrastructure exposes multiple misc device nodes one per implemented 26parameter. The set of parameters implement is defined by pm_qos_power_init() 27and pm_qos_params.h. This is done because having the available parameters 28being runtime configurable or changeable from a driver was seen as too easy to 29abuse. 30 31For each parameter a list of performance requests is maintained along with 32an aggregated target value. The aggregated target value is updated with 33changes to the request list or elements of the list. Typically the 34aggregated target value is simply the max or min of the request values held 35in the parameter list elements. 36Note: the aggregated target value is implemented as an atomic variable so that 37reading the aggregated value does not require any locking mechanism. 38 39 40From kernel mode the use of this interface is simple: 41 42void pm_qos_add_request(handle, param_class, target_value): 43 Will insert an element into the list for that identified PM QoS class with the 44 target value. Upon change to this list the new target is recomputed and any 45 registered notifiers are called only if the target value is now different. 46 Clients of pm_qos need to save the returned handle for future use in other 47 pm_qos API functions. 48 49void pm_qos_update_request(handle, new_target_value): 50 Will update the list element pointed to by the handle with the new target 51 value and recompute the new aggregated target, calling the notification tree 52 if the target is changed. 53 54void pm_qos_remove_request(handle): 55 Will remove the element. After removal it will update the aggregate target 56 and call the notification tree if the target was changed as a result of 57 removing the request. 58 59int pm_qos_request(param_class): 60 Returns the aggregated value for a given PM QoS class. 61 62int pm_qos_request_active(handle): 63 Returns if the request is still active, i.e. it has not been removed from a 64 PM QoS class constraints list. 65 66int pm_qos_add_notifier(param_class, notifier): 67 Adds a notification callback function to the PM QoS class. The callback is 68 called when the aggregated value for the PM QoS class is changed. 69 70int pm_qos_remove_notifier(int param_class, notifier): 71 Removes the notification callback function for the PM QoS class. 72 73 74From user mode: 75 76Only processes can register a pm_qos request. To provide for automatic 77cleanup of a process, the interface requires the process to register its 78parameter requests in the following way: 79 80To register the default pm_qos target for the specific parameter, the process 81must open /dev/cpu_dma_latency 82 83As long as the device node is held open that process has a registered 84request on the parameter. 85 86To change the requested target value the process needs to write an s32 value to 87the open device node. Alternatively the user mode program could write a hex 88string for the value using 10 char long format e.g. "0x12345678". This 89translates to a pm_qos_update_request call. 90 91To remove the user mode request for a target value simply close the device 92node. 93 94 952. PM QoS per-device latency and flags framework 96================================================ 97 98For each device, there are three lists of PM QoS requests. Two of them are 99maintained along with the aggregated targets of resume latency and active 100state latency tolerance (in microseconds) and the third one is for PM QoS flags. 101Values are updated in response to changes of the request list. 102 103The target values of resume latency and active state latency tolerance are 104simply the minimum of the request values held in the parameter list elements. 105The PM QoS flags aggregate value is a gather (bitwise OR) of all list elements' 106values. One device PM QoS flag is defined currently: PM_QOS_FLAG_NO_POWER_OFF. 107 108Note: The aggregated target values are implemented in such a way that reading 109the aggregated value does not require any locking mechanism. 110 111 112From kernel mode the use of this interface is the following: 113 114int dev_pm_qos_add_request(device, handle, type, value): 115 Will insert an element into the list for that identified device with the 116 target value. Upon change to this list the new target is recomputed and any 117 registered notifiers are called only if the target value is now different. 118 Clients of dev_pm_qos need to save the handle for future use in other 119 dev_pm_qos API functions. 120 121int dev_pm_qos_update_request(handle, new_value): 122 Will update the list element pointed to by the handle with the new target 123 value and recompute the new aggregated target, calling the notification 124 trees if the target is changed. 125 126int dev_pm_qos_remove_request(handle): 127 Will remove the element. After removal it will update the aggregate target 128 and call the notification trees if the target was changed as a result of 129 removing the request. 130 131s32 dev_pm_qos_read_value(device, type): 132 Returns the aggregated value for a given device's constraints list. 133 134enum pm_qos_flags_status dev_pm_qos_flags(device, mask) 135 Check PM QoS flags of the given device against the given mask of flags. 136 The meaning of the return values is as follows: 137 138 PM_QOS_FLAGS_ALL: 139 All flags from the mask are set 140 PM_QOS_FLAGS_SOME: 141 Some flags from the mask are set 142 PM_QOS_FLAGS_NONE: 143 No flags from the mask are set 144 PM_QOS_FLAGS_UNDEFINED: 145 The device's PM QoS structure has not been initialized 146 or the list of requests is empty. 147 148int dev_pm_qos_add_ancestor_request(dev, handle, type, value) 149 Add a PM QoS request for the first direct ancestor of the given device whose 150 power.ignore_children flag is unset (for DEV_PM_QOS_RESUME_LATENCY requests) 151 or whose power.set_latency_tolerance callback pointer is not NULL (for 152 DEV_PM_QOS_LATENCY_TOLERANCE requests). 153 154int dev_pm_qos_expose_latency_limit(device, value) 155 Add a request to the device's PM QoS list of resume latency constraints and 156 create a sysfs attribute pm_qos_resume_latency_us under the device's power 157 directory allowing user space to manipulate that request. 158 159void dev_pm_qos_hide_latency_limit(device) 160 Drop the request added by dev_pm_qos_expose_latency_limit() from the device's 161 PM QoS list of resume latency constraints and remove sysfs attribute 162 pm_qos_resume_latency_us from the device's power directory. 163 164int dev_pm_qos_expose_flags(device, value) 165 Add a request to the device's PM QoS list of flags and create sysfs attribute 166 pm_qos_no_power_off under the device's power directory allowing user space to 167 change the value of the PM_QOS_FLAG_NO_POWER_OFF flag. 168 169void dev_pm_qos_hide_flags(device) 170 Drop the request added by dev_pm_qos_expose_flags() from the device's PM QoS 171 list of flags and remove sysfs attribute pm_qos_no_power_off from the device's 172 power directory. 173 174Notification mechanisms: 175 176The per-device PM QoS framework has a per-device notification tree. 177 178int dev_pm_qos_add_notifier(device, notifier, type): 179 Adds a notification callback function for the device for a particular request 180 type. 181 182 The callback is called when the aggregated value of the device constraints 183 list is changed. 184 185int dev_pm_qos_remove_notifier(device, notifier, type): 186 Removes the notification callback function for the device. 187 188 189Active state latency tolerance 190^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 191 192This device PM QoS type is used to support systems in which hardware may switch 193to energy-saving operation modes on the fly. In those systems, if the operation 194mode chosen by the hardware attempts to save energy in an overly aggressive way, 195it may cause excess latencies to be visible to software, causing it to miss 196certain protocol requirements or target frame or sample rates etc. 197 198If there is a latency tolerance control mechanism for a given device available 199to software, the .set_latency_tolerance callback in that device's dev_pm_info 200structure should be populated. The routine pointed to by it is should implement 201whatever is necessary to transfer the effective requirement value to the 202hardware. 203 204Whenever the effective latency tolerance changes for the device, its 205.set_latency_tolerance() callback will be executed and the effective value will 206be passed to it. If that value is negative, which means that the list of 207latency tolerance requirements for the device is empty, the callback is expected 208to switch the underlying hardware latency tolerance control mechanism to an 209autonomous mode if available. If that value is PM_QOS_LATENCY_ANY, in turn, and 210the hardware supports a special "no requirement" setting, the callback is 211expected to use it. That allows software to prevent the hardware from 212automatically updating the device's latency tolerance in response to its power 213state changes (e.g. during transitions from D3cold to D0), which generally may 214be done in the autonomous latency tolerance control mode. 215 216If .set_latency_tolerance() is present for the device, sysfs attribute 217pm_qos_latency_tolerance_us will be present in the devivce's power directory. 218Then, user space can use that attribute to specify its latency tolerance 219requirement for the device, if any. Writing "any" to it means "no requirement, 220but do not let the hardware control latency tolerance" and writing "auto" to it 221allows the hardware to be switched to the autonomous mode if there are no other 222requirements from the kernel side in the device's list. 223 224Kernel code can use the functions described above along with the 225DEV_PM_QOS_LATENCY_TOLERANCE device PM QoS type to add, remove and update 226latency tolerance requirements for devices. 227