5 This document describes the temporary OMAP PM interface. Driver
6 authors use these functions to communicate minimum latency or
7 throughput constraints to the kernel power management code.
8 Over time, the intention is to merge features from the OMAP PM
9 interface into the Linux PM QoS code.
11 Drivers need to express PM parameters which:
13 - support the range of power management parameters present in the TI SRF;
15 - separate the drivers from the underlying PM parameter
16 implementation, whether it is the TI SRF or Linux PM QoS or Linux
17 latency framework or something else;
19 - specify PM parameters in terms of fundamental units, such as
20 latency and throughput, rather than units which are specific to OMAP
21 or to particular OMAP variants;
23 - allow drivers which are shared with other architectures (e.g.,
24 DaVinci) to add these constraints in a way which won't affect non-OMAP
27 - can be implemented immediately with minimal disruption of other
31 This document proposes the OMAP PM interface, including the following
32 five power management functions for driver code:
34 1. Set the maximum MPU wakeup latency:
35 (*pdata->set_max_mpu_wakeup_lat)(struct device *dev, unsigned long t)
37 2. Set the maximum device wakeup latency:
38 (*pdata->set_max_dev_wakeup_lat)(struct device *dev, unsigned long t)
40 3. Set the maximum system DMA transfer start latency (CORE pwrdm):
41 (*pdata->set_max_sdma_lat)(struct device *dev, long t)
43 4. Set the minimum bus throughput needed by a device:
44 (*pdata->set_min_bus_tput)(struct device *dev, u8 agent_id, unsigned long r)
46 5. Return the number of times the device has lost context
47 (*pdata->get_dev_context_loss_count)(struct device *dev)
50 Further documentation for all OMAP PM interface functions can be
51 found in arch/arm/plat-omap/include/mach/omap-pm.h.
54 The OMAP PM layer is intended to be temporary
55 ---------------------------------------------
57 The intention is that eventually the Linux PM QoS layer should support
58 the range of power management features present in OMAP3. As this
59 happens, existing drivers using the OMAP PM interface can be modified
60 to use the Linux PM QoS code; and the OMAP PM interface can disappear.
63 Driver usage of the OMAP PM functions
64 -------------------------------------
66 As the 'pdata' in the above examples indicates, these functions are
67 exposed to drivers through function pointers in driver .platform_data
68 structures. The function pointers are initialized by the board-*.c
69 files to point to the corresponding OMAP PM functions:
70 .set_max_dev_wakeup_lat will point to
71 omap_pm_set_max_dev_wakeup_lat(), etc. Other architectures which do
72 not support these functions should leave these function pointers set
73 to NULL. Drivers should use the following idiom:
75 if (pdata->set_max_dev_wakeup_lat)
76 (*pdata->set_max_dev_wakeup_lat)(dev, t);
78 The most common usage of these functions will probably be to specify
79 the maximum time from when an interrupt occurs, to when the device
80 becomes accessible. To accomplish this, driver writers should use the
81 set_max_mpu_wakeup_lat() function to constrain the MPU wakeup
82 latency, and the set_max_dev_wakeup_lat() function to constrain the
83 device wakeup latency (from clk_enable() to accessibility). For
86 /* Limit MPU wakeup latency */
87 if (pdata->set_max_mpu_wakeup_lat)
88 (*pdata->set_max_mpu_wakeup_lat)(dev, tc);
90 /* Limit device powerdomain wakeup latency */
91 if (pdata->set_max_dev_wakeup_lat)
92 (*pdata->set_max_dev_wakeup_lat)(dev, td);
94 /* total wakeup latency in this example: (tc + td) */
96 The PM parameters can be overwritten by calling the function again
97 with the new value. The settings can be removed by calling the
98 function with a t argument of -1 (except in the case of
99 set_max_bus_tput(), which should be called with an r argument of 0).
101 The fifth function above, omap_pm_get_dev_context_loss_count(),
102 is intended as an optimization to allow drivers to determine whether the
103 device has lost its internal context. If context has been lost, the
104 driver must restore its internal context before proceeding.
107 Other specialized interface functions
108 -------------------------------------
110 The five functions listed above are intended to be usable by any
111 device driver. DSPBridge and CPUFreq have a few special requirements.
112 DSPBridge expresses target DSP performance levels in terms of OPP IDs.
113 CPUFreq expresses target MPU performance levels in terms of MPU
114 frequency. The OMAP PM interface contains functions for these
115 specialized cases to convert that input information (OPPs/MPU
116 frequency) into the form that the underlying power management
117 implementation needs:
119 6. (*pdata->dsp_get_opp_table)(void)
121 7. (*pdata->dsp_set_min_opp)(u8 opp_id)
123 8. (*pdata->dsp_get_opp)(void)
125 9. (*pdata->cpu_get_freq_table)(void)
127 10. (*pdata->cpu_set_freq)(unsigned long f)
129 11. (*pdata->cpu_get_freq)(void)
131 Customizing OPP for platform
132 ============================
133 Defining CONFIG_PM should enable OPP layer for the silicon
134 and the registration of OPP table should take place automatically.
135 However, in special cases, the default OPP table may need to be
137 * enable default OPPs which are disabled by default, but which
138 could be enabled on a platform
139 * Disable an unsupported OPP on the platform
140 * Define and add a custom opp table entry
141 in these cases, the board file needs to do additional steps as follows:
142 arch/arm/mach-omapx/board-xyz.c
145 static void __init omap_xyz_init_irq(void)
148 /* Initialize the default table */
150 /* Do customization to the defaults */
153 NOTE: omapx_opp_init will be omap3_opp_init or as required
154 based on the omap family.