1Operating Performance Points (OPP) Library
2==========================================
3
4(C) 2009-2010 Nishanth Menon <nm@ti.com>, Texas Instruments Incorporated
5
6Contents
7--------
81. Introduction
92. Initial OPP List Registration
103. OPP Search Functions
114. OPP Availability Control Functions
125. OPP Data Retrieval Functions
136. Data Structures
14
151. Introduction
16===============
171.1 What is an Operating Performance Point (OPP)?
18
19Complex SoCs of today consists of a multiple sub-modules working in conjunction.
20In an operational system executing varied use cases, not all modules in the SoC
21need to function at their highest performing frequency all the time. To
22facilitate this, sub-modules in a SoC are grouped into domains, allowing some
23domains to run at lower voltage and frequency while other domains run at
24voltage/frequency pairs that are higher.
25
26The set of discrete tuples consisting of frequency and voltage pairs that
27the device will support per domain are called Operating Performance Points or
28OPPs.
29
30As an example:
31Let us consider an MPU device which supports the following:
32{300MHz at minimum voltage of 1V}, {800MHz at minimum voltage of 1.2V},
33{1GHz at minimum voltage of 1.3V}
34
35We can represent these as three OPPs as the following {Hz, uV} tuples:
36{300000000, 1000000}
37{800000000, 1200000}
38{1000000000, 1300000}
39
401.2 Operating Performance Points Library
41
42OPP library provides a set of helper functions to organize and query the OPP
43information. The library is located in drivers/base/power/opp.c and the header
44is located in include/linux/pm_opp.h. OPP library can be enabled by enabling
45CONFIG_PM_OPP from power management menuconfig menu. OPP library depends on
46CONFIG_PM as certain SoCs such as Texas Instrument's OMAP framework allows to
47optionally boot at a certain OPP without needing cpufreq.
48
49Typical usage of the OPP library is as follows:
50(users)		-> registers a set of default OPPs		-> (library)
51SoC framework	-> modifies on required cases certain OPPs	-> OPP layer
52		-> queries to search/retrieve information	->
53
54OPP layer expects each domain to be represented by a unique device pointer. SoC
55framework registers a set of initial OPPs per device with the OPP layer. This
56list is expected to be an optimally small number typically around 5 per device.
57This initial list contains a set of OPPs that the framework expects to be safely
58enabled by default in the system.
59
60Note on OPP Availability:
61------------------------
62As the system proceeds to operate, SoC framework may choose to make certain
63OPPs available or not available on each device based on various external
64factors. Example usage: Thermal management or other exceptional situations where
65SoC framework might choose to disable a higher frequency OPP to safely continue
66operations until that OPP could be re-enabled if possible.
67
68OPP library facilitates this concept in it's implementation. The following
69operational functions operate only on available opps:
70opp_find_freq_{ceil, floor}, dev_pm_opp_get_voltage, dev_pm_opp_get_freq, dev_pm_opp_get_opp_count
71
72dev_pm_opp_find_freq_exact is meant to be used to find the opp pointer which can then
73be used for dev_pm_opp_enable/disable functions to make an opp available as required.
74
75WARNING: Users of OPP library should refresh their availability count using
76get_opp_count if dev_pm_opp_enable/disable functions are invoked for a device, the
77exact mechanism to trigger these or the notification mechanism to other
78dependent subsystems such as cpufreq are left to the discretion of the SoC
79specific framework which uses the OPP library. Similar care needs to be taken
80care to refresh the cpufreq table in cases of these operations.
81
82WARNING on OPP List locking mechanism:
83-------------------------------------------------
84OPP library uses RCU for exclusivity. RCU allows the query functions to operate
85in multiple contexts and this synchronization mechanism is optimal for a read
86intensive operations on data structure as the OPP library caters to.
87
88To ensure that the data retrieved are sane, the users such as SoC framework
89should ensure that the section of code operating on OPP queries are locked
90using RCU read locks. The opp_find_freq_{exact,ceil,floor},
91opp_get_{voltage, freq, opp_count} fall into this category.
92
93opp_{add,enable,disable} are updaters which use mutex and implement it's own
94RCU locking mechanisms. These functions should *NOT* be called under RCU locks
95and other contexts that prevent blocking functions in RCU or mutex operations
96from working.
97
982. Initial OPP List Registration
99================================
100The SoC implementation calls dev_pm_opp_add function iteratively to add OPPs per
101device. It is expected that the SoC framework will register the OPP entries
102optimally- typical numbers range to be less than 5. The list generated by
103registering the OPPs is maintained by OPP library throughout the device
104operation. The SoC framework can subsequently control the availability of the
105OPPs dynamically using the dev_pm_opp_enable / disable functions.
106
107dev_pm_opp_add - Add a new OPP for a specific domain represented by the device pointer.
108	The OPP is defined using the frequency and voltage. Once added, the OPP
109	is assumed to be available and control of it's availability can be done
110	with the dev_pm_opp_enable/disable functions. OPP library internally stores
111	and manages this information in the opp struct. This function may be
112	used by SoC framework to define a optimal list as per the demands of
113	SoC usage environment.
114
115	WARNING: Do not use this function in interrupt context.
116
117	Example:
118	 soc_pm_init()
119	 {
120		/* Do things */
121		r = dev_pm_opp_add(mpu_dev, 1000000, 900000);
122		if (!r) {
123			pr_err("%s: unable to register mpu opp(%d)\n", r);
124			goto no_cpufreq;
125		}
126		/* Do cpufreq things */
127	 no_cpufreq:
128		/* Do remaining things */
129	 }
130
1313. OPP Search Functions
132=======================
133High level framework such as cpufreq operates on frequencies. To map the
134frequency back to the corresponding OPP, OPP library provides handy functions
135to search the OPP list that OPP library internally manages. These search
136functions return the matching pointer representing the opp if a match is
137found, else returns error. These errors are expected to be handled by standard
138error checks such as IS_ERR() and appropriate actions taken by the caller.
139
140dev_pm_opp_find_freq_exact - Search for an OPP based on an *exact* frequency and
141	availability. This function is especially useful to enable an OPP which
142	is not available by default.
143	Example: In a case when SoC framework detects a situation where a
144	higher frequency could be made available, it can use this function to
145	find the OPP prior to call the dev_pm_opp_enable to actually make it available.
146	 rcu_read_lock();
147	 opp = dev_pm_opp_find_freq_exact(dev, 1000000000, false);
148	 rcu_read_unlock();
149	 /* dont operate on the pointer.. just do a sanity check.. */
150	 if (IS_ERR(opp)) {
151		pr_err("frequency not disabled!\n");
152		/* trigger appropriate actions.. */
153	 } else {
154		dev_pm_opp_enable(dev,1000000000);
155	 }
156
157	NOTE: This is the only search function that operates on OPPs which are
158	not available.
159
160dev_pm_opp_find_freq_floor - Search for an available OPP which is *at most* the
161	provided frequency. This function is useful while searching for a lesser
162	match OR operating on OPP information in the order of decreasing
163	frequency.
164	Example: To find the highest opp for a device:
165	 freq = ULONG_MAX;
166	 rcu_read_lock();
167	 dev_pm_opp_find_freq_floor(dev, &freq);
168	 rcu_read_unlock();
169
170dev_pm_opp_find_freq_ceil - Search for an available OPP which is *at least* the
171	provided frequency. This function is useful while searching for a
172	higher match OR operating on OPP information in the order of increasing
173	frequency.
174	Example 1: To find the lowest opp for a device:
175	 freq = 0;
176	 rcu_read_lock();
177	 dev_pm_opp_find_freq_ceil(dev, &freq);
178	 rcu_read_unlock();
179	Example 2: A simplified implementation of a SoC cpufreq_driver->target:
180	 soc_cpufreq_target(..)
181	 {
182		/* Do stuff like policy checks etc. */
183		/* Find the best frequency match for the req */
184		rcu_read_lock();
185		opp = dev_pm_opp_find_freq_ceil(dev, &freq);
186		rcu_read_unlock();
187		if (!IS_ERR(opp))
188			soc_switch_to_freq_voltage(freq);
189		else
190			/* do something when we can't satisfy the req */
191		/* do other stuff */
192	 }
193
1944. OPP Availability Control Functions
195=====================================
196A default OPP list registered with the OPP library may not cater to all possible
197situation. The OPP library provides a set of functions to modify the
198availability of a OPP within the OPP list. This allows SoC frameworks to have
199fine grained dynamic control of which sets of OPPs are operationally available.
200These functions are intended to *temporarily* remove an OPP in conditions such
201as thermal considerations (e.g. don't use OPPx until the temperature drops).
202
203WARNING: Do not use these functions in interrupt context.
204
205dev_pm_opp_enable - Make a OPP available for operation.
206	Example: Lets say that 1GHz OPP is to be made available only if the
207	SoC temperature is lower than a certain threshold. The SoC framework
208	implementation might choose to do something as follows:
209	 if (cur_temp < temp_low_thresh) {
210		/* Enable 1GHz if it was disabled */
211		rcu_read_lock();
212		opp = dev_pm_opp_find_freq_exact(dev, 1000000000, false);
213		rcu_read_unlock();
214		/* just error check */
215		if (!IS_ERR(opp))
216			ret = dev_pm_opp_enable(dev, 1000000000);
217		else
218			goto try_something_else;
219	 }
220
221dev_pm_opp_disable - Make an OPP to be not available for operation
222	Example: Lets say that 1GHz OPP is to be disabled if the temperature
223	exceeds a threshold value. The SoC framework implementation might
224	choose to do something as follows:
225	 if (cur_temp > temp_high_thresh) {
226		/* Disable 1GHz if it was enabled */
227		rcu_read_lock();
228		opp = dev_pm_opp_find_freq_exact(dev, 1000000000, true);
229		rcu_read_unlock();
230		/* just error check */
231		if (!IS_ERR(opp))
232			ret = dev_pm_opp_disable(dev, 1000000000);
233		else
234			goto try_something_else;
235	 }
236
2375. OPP Data Retrieval Functions
238===============================
239Since OPP library abstracts away the OPP information, a set of functions to pull
240information from the OPP structure is necessary. Once an OPP pointer is
241retrieved using the search functions, the following functions can be used by SoC
242framework to retrieve the information represented inside the OPP layer.
243
244dev_pm_opp_get_voltage - Retrieve the voltage represented by the opp pointer.
245	Example: At a cpufreq transition to a different frequency, SoC
246	framework requires to set the voltage represented by the OPP using
247	the regulator framework to the Power Management chip providing the
248	voltage.
249	 soc_switch_to_freq_voltage(freq)
250	 {
251		/* do things */
252		rcu_read_lock();
253		opp = dev_pm_opp_find_freq_ceil(dev, &freq);
254		v = dev_pm_opp_get_voltage(opp);
255		rcu_read_unlock();
256		if (v)
257			regulator_set_voltage(.., v);
258		/* do other things */
259	 }
260
261dev_pm_opp_get_freq - Retrieve the freq represented by the opp pointer.
262	Example: Lets say the SoC framework uses a couple of helper functions
263	we could pass opp pointers instead of doing additional parameters to
264	handle quiet a bit of data parameters.
265	 soc_cpufreq_target(..)
266	 {
267		/* do things.. */
268		 max_freq = ULONG_MAX;
269		 rcu_read_lock();
270		 max_opp = dev_pm_opp_find_freq_floor(dev,&max_freq);
271		 requested_opp = dev_pm_opp_find_freq_ceil(dev,&freq);
272		 if (!IS_ERR(max_opp) && !IS_ERR(requested_opp))
273			r = soc_test_validity(max_opp, requested_opp);
274		 rcu_read_unlock();
275		/* do other things */
276	 }
277	 soc_test_validity(..)
278	 {
279		 if(dev_pm_opp_get_voltage(max_opp) < dev_pm_opp_get_voltage(requested_opp))
280			 return -EINVAL;
281		 if(dev_pm_opp_get_freq(max_opp) < dev_pm_opp_get_freq(requested_opp))
282			 return -EINVAL;
283		/* do things.. */
284	 }
285
286dev_pm_opp_get_opp_count - Retrieve the number of available opps for a device
287	Example: Lets say a co-processor in the SoC needs to know the available
288	frequencies in a table, the main processor can notify as following:
289	 soc_notify_coproc_available_frequencies()
290	 {
291		/* Do things */
292		rcu_read_lock();
293		num_available = dev_pm_opp_get_opp_count(dev);
294		speeds = kzalloc(sizeof(u32) * num_available, GFP_KERNEL);
295		/* populate the table in increasing order */
296		freq = 0;
297		while (!IS_ERR(opp = dev_pm_opp_find_freq_ceil(dev, &freq))) {
298			speeds[i] = freq;
299			freq++;
300			i++;
301		}
302		rcu_read_unlock();
303
304		soc_notify_coproc(AVAILABLE_FREQs, speeds, num_available);
305		/* Do other things */
306	 }
307
3086. Data Structures
309==================
310Typically an SoC contains multiple voltage domains which are variable. Each
311domain is represented by a device pointer. The relationship to OPP can be
312represented as follows:
313SoC
314 |- device 1
315 |	|- opp 1 (availability, freq, voltage)
316 |	|- opp 2 ..
317 ...	...
318 |	`- opp n ..
319 |- device 2
320 ...
321 `- device m
322
323OPP library maintains a internal list that the SoC framework populates and
324accessed by various functions as described above. However, the structures
325representing the actual OPPs and domains are internal to the OPP library itself
326to allow for suitable abstraction reusable across systems.
327
328struct dev_pm_opp - The internal data structure of OPP library which is used to
329	represent an OPP. In addition to the freq, voltage, availability
330	information, it also contains internal book keeping information required
331	for the OPP library to operate on.  Pointer to this structure is
332	provided back to the users such as SoC framework to be used as a
333	identifier for OPP in the interactions with OPP layer.
334
335	WARNING: The struct dev_pm_opp pointer should not be parsed or modified by the
336	users. The defaults of for an instance is populated by dev_pm_opp_add, but the
337	availability of the OPP can be modified by dev_pm_opp_enable/disable functions.
338
339struct device - This is used to identify a domain to the OPP layer. The
340	nature of the device and it's implementation is left to the user of
341	OPP library such as the SoC framework.
342
343Overall, in a simplistic view, the data structure operations is represented as
344following:
345
346Initialization / modification:
347            +-----+        /- dev_pm_opp_enable
348dev_pm_opp_add --> | opp | <-------
349  |         +-----+        \- dev_pm_opp_disable
350  \-------> domain_info(device)
351
352Search functions:
353             /-- dev_pm_opp_find_freq_ceil  ---\   +-----+
354domain_info<---- dev_pm_opp_find_freq_exact -----> | opp |
355             \-- dev_pm_opp_find_freq_floor ---/   +-----+
356
357Retrieval functions:
358+-----+     /- dev_pm_opp_get_voltage
359| opp | <---
360+-----+     \- dev_pm_opp_get_freq
361
362domain_info <- dev_pm_opp_get_opp_count
363