1 =============================== 2 PM Quality Of Service Interface 3 =============================== 4 5 This interface provides a kernel and user mode interface for registering 6 performance expectations by drivers, subsystems and user space applications on 7 one of the parameters. 8 9 Two 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 14 The latency unit used in the PM QoS framework is the microsecond (usec). 15 16 17 1. PM QoS framework 18 =================== 19 20 A 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 22 to the request list or elements of the list. For CPU latency QoS, the 23 aggregated target value is simply the min of the request values held in the list 24 elements. 25 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 29 From kernel space the use of this interface is simple: 30 31 void 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 38 void 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 43 void 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 48 int cpu_latency_qos_limit(): 49 Returns the aggregated value for the CPU latency QoS. 50 51 int 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 int cpu_latency_qos_add_notifier(notifier): 56 Adds a notification callback function to the CPU latency QoS. The callback is 57 called when the aggregated value for the CPU latency QoS is changed. 58 59 int cpu_latency_qos_remove_notifier(notifier): 60 Removes the notification callback function from the CPU latency QoS. 61 62 63 From user space: 64 65 The infrastructure exposes one device node, /dev/cpu_dma_latency, for the CPU 66 latency QoS. 67 68 Only processes can register a PM QoS request. To provide for automatic 69 cleanup of a process, the interface requires the process to register its 70 parameter requests as follows. 71 72 To register the default PM QoS target for the CPU latency QoS, the process must 73 open /dev/cpu_dma_latency. 74 75 As long as the device node is held open that process has a registered 76 request on the parameter. 77 78 To change the requested target value, the process needs to write an s32 value to 79 the open device node. Alternatively, it can write a hex string for the value 80 using the 10 char long format e.g. "0x12345678". This translates to a 81 cpu_latency_qos_update_request() call. 82 83 To remove the user mode request for a target value simply close the device 84 node. 85 86 87 2. PM QoS per-device latency and flags framework 88 ================================================ 89 90 For each device, there are three lists of PM QoS requests. Two of them are 91 maintained along with the aggregated targets of resume latency and active 92 state latency tolerance (in microseconds) and the third one is for PM QoS flags. 93 Values are updated in response to changes of the request list. 94 95 The target values of resume latency and active state latency tolerance are 96 simply the minimum of the request values held in the parameter list elements. 97 The PM QoS flags aggregate value is a gather (bitwise OR) of all list elements' 98 values. One device PM QoS flag is defined currently: PM_QOS_FLAG_NO_POWER_OFF. 99 100 Note: The aggregated target values are implemented in such a way that reading 101 the aggregated value does not require any locking mechanism. 102 103 104 From kernel mode the use of this interface is the following: 105 106 int dev_pm_qos_add_request(device, handle, type, value): 107 Will insert an element into the list for that identified device with the 108 target value. Upon change to this list the new target is recomputed and any 109 registered notifiers are called only if the target value is now different. 110 Clients of dev_pm_qos need to save the handle for future use in other 111 dev_pm_qos API functions. 112 113 int dev_pm_qos_update_request(handle, new_value): 114 Will update the list element pointed to by the handle with the new target 115 value and recompute the new aggregated target, calling the notification 116 trees if the target is changed. 117 118 int dev_pm_qos_remove_request(handle): 119 Will remove the element. After removal it will update the aggregate target 120 and call the notification trees if the target was changed as a result of 121 removing the request. 122 123 s32 dev_pm_qos_read_value(device, type): 124 Returns the aggregated value for a given device's constraints list. 125 126 enum pm_qos_flags_status dev_pm_qos_flags(device, mask) 127 Check PM QoS flags of the given device against the given mask of flags. 128 The meaning of the return values is as follows: 129 130 PM_QOS_FLAGS_ALL: 131 All flags from the mask are set 132 PM_QOS_FLAGS_SOME: 133 Some flags from the mask are set 134 PM_QOS_FLAGS_NONE: 135 No flags from the mask are set 136 PM_QOS_FLAGS_UNDEFINED: 137 The device's PM QoS structure has not been initialized 138 or the list of requests is empty. 139 140 int dev_pm_qos_add_ancestor_request(dev, handle, type, value) 141 Add a PM QoS request for the first direct ancestor of the given device whose 142 power.ignore_children flag is unset (for DEV_PM_QOS_RESUME_LATENCY requests) 143 or whose power.set_latency_tolerance callback pointer is not NULL (for 144 DEV_PM_QOS_LATENCY_TOLERANCE requests). 145 146 int dev_pm_qos_expose_latency_limit(device, value) 147 Add a request to the device's PM QoS list of resume latency constraints and 148 create a sysfs attribute pm_qos_resume_latency_us under the device's power 149 directory allowing user space to manipulate that request. 150 151 void dev_pm_qos_hide_latency_limit(device) 152 Drop the request added by dev_pm_qos_expose_latency_limit() from the device's 153 PM QoS list of resume latency constraints and remove sysfs attribute 154 pm_qos_resume_latency_us from the device's power directory. 155 156 int dev_pm_qos_expose_flags(device, value) 157 Add a request to the device's PM QoS list of flags and create sysfs attribute 158 pm_qos_no_power_off under the device's power directory allowing user space to 159 change the value of the PM_QOS_FLAG_NO_POWER_OFF flag. 160 161 void dev_pm_qos_hide_flags(device) 162 Drop the request added by dev_pm_qos_expose_flags() from the device's PM QoS 163 list of flags and remove sysfs attribute pm_qos_no_power_off from the device's 164 power directory. 165 166 Notification mechanisms: 167 168 The per-device PM QoS framework has a per-device notification tree. 169 170 int dev_pm_qos_add_notifier(device, notifier, type): 171 Adds a notification callback function for the device for a particular request 172 type. 173 174 The callback is called when the aggregated value of the device constraints 175 list is changed. 176 177 int dev_pm_qos_remove_notifier(device, notifier, type): 178 Removes the notification callback function for the device. 179 180 181 Active state latency tolerance 182 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 183 184 This device PM QoS type is used to support systems in which hardware may switch 185 to energy-saving operation modes on the fly. In those systems, if the operation 186 mode chosen by the hardware attempts to save energy in an overly aggressive way, 187 it may cause excess latencies to be visible to software, causing it to miss 188 certain protocol requirements or target frame or sample rates etc. 189 190 If there is a latency tolerance control mechanism for a given device available 191 to software, the .set_latency_tolerance callback in that device's dev_pm_info 192 structure should be populated. The routine pointed to by it is should implement 193 whatever is necessary to transfer the effective requirement value to the 194 hardware. 195 196 Whenever the effective latency tolerance changes for the device, its 197 .set_latency_tolerance() callback will be executed and the effective value will 198 be passed to it. If that value is negative, which means that the list of 199 latency tolerance requirements for the device is empty, the callback is expected 200 to switch the underlying hardware latency tolerance control mechanism to an 201 autonomous mode if available. If that value is PM_QOS_LATENCY_ANY, in turn, and 202 the hardware supports a special "no requirement" setting, the callback is 203 expected to use it. That allows software to prevent the hardware from 204 automatically updating the device's latency tolerance in response to its power 205 state changes (e.g. during transitions from D3cold to D0), which generally may 206 be done in the autonomous latency tolerance control mode. 207 208 If .set_latency_tolerance() is present for the device, sysfs attribute 209 pm_qos_latency_tolerance_us will be present in the devivce's power directory. 210 Then, user space can use that attribute to specify its latency tolerance 211 requirement for the device, if any. Writing "any" to it means "no requirement, 212 but do not let the hardware control latency tolerance" and writing "auto" to it 213 allows the hardware to be switched to the autonomous mode if there are no other 214 requirements from the kernel side in the device's list. 215 216 Kernel code can use the functions described above along with the 217 DEV_PM_QOS_LATENCY_TOLERANCE device PM QoS type to add, remove and update 218 latency tolerance requirements for devices.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.