Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | README for MIPS time services |
| 2 | |
| 3 | Jun Sun |
| 4 | jsun@mvista.com or jsun@junsun.net |
| 5 | |
| 6 | |
| 7 | ABOUT |
| 8 | ----- |
| 9 | This file describes the new arch/mips/kernel/time.c, related files and the |
| 10 | services they provide. |
| 11 | |
| 12 | If you are short in patience and just want to know how to use time.c for a |
| 13 | new board or convert an existing board, go to the last section. |
| 14 | |
| 15 | |
| 16 | FILES, COMPATABILITY AND CONFIGS |
| 17 | --------------------------------- |
| 18 | |
| 19 | The old arch/mips/kernel/time.c is renamed to old-time.c. |
| 20 | |
| 21 | A new time.c is put there, together with include/asm-mips/time.h. |
| 22 | |
| 23 | Two configs variables are introduced, CONFIG_OLD_TIME_C and CONFIG_NEW_TIME_C. |
| 24 | So we allow boards using |
| 25 | |
| 26 | 1) old time.c (CONFIG_OLD_TIME_C) |
| 27 | 2) new time.c (CONFIG_NEW_TIME_C) |
| 28 | 3) neither (their own private time.c) |
| 29 | |
| 30 | However, it is expected every board will move to the new time.c in the near |
| 31 | future. |
| 32 | |
| 33 | |
| 34 | WHAT THE NEW CODE PROVIDES? |
| 35 | --------------------------- |
| 36 | |
| 37 | The new time code provide the following services: |
| 38 | |
| 39 | a) Implements functions required by Linux common code: |
| 40 | time_init |
| 41 | do_gettimeofday |
| 42 | do_settimeofday |
| 43 | |
| 44 | b) provides an abstraction of RTC and null RTC implementation as default. |
| 45 | extern unsigned long (*rtc_get_time)(void); |
| 46 | extern int (*rtc_set_time)(unsigned long); |
| 47 | |
| 48 | c) a set of gettimeoffset functions for different CPUs and different |
| 49 | needs. |
| 50 | |
| 51 | d) high-level and low-level timer interrupt routines where the timer |
| 52 | interrupt source may or may not be the CPU timer. The high-level |
| 53 | routine is dispatched through do_IRQ() while the low-level is |
| 54 | dispatched in assemably code (usually int-handler.S) |
| 55 | |
| 56 | |
| 57 | WHAT THE NEW CODE REQUIRES? |
| 58 | --------------------------- |
| 59 | |
| 60 | For the new code to work properly, each board implementation needs to supply |
| 61 | the following functions or values: |
| 62 | |
| 63 | a) board_time_init - a function pointer. Invoked at the beginnig of |
| 64 | time_init(). It is optional. |
| 65 | 1. (optional) set up RTC routines |
| 66 | 2. (optional) calibrate and set the mips_counter_frequency |
| 67 | |
| 68 | b) board_timer_setup - a function pointer. Invoked at the end of time_init() |
| 69 | 1. (optional) over-ride any decisions made in time_init() |
| 70 | 2. set up the irqaction for timer interrupt. |
| 71 | 3. enable the timer interrupt |
| 72 | |
| 73 | c) (optional) board-specific RTC routines. |
| 74 | |
| 75 | d) (optional) mips_counter_frequency - It must be definied if the board |
| 76 | is using CPU counter for timer interrupt or it is using fixed rate |
| 77 | gettimeoffset(). |
| 78 | |
| 79 | |
| 80 | PORTING GUIDE |
| 81 | ------------- |
| 82 | |
| 83 | Step 1: decide how you like to implement the time services. |
| 84 | |
| 85 | a) does this board have a RTC? If yes, implement the two RTC funcs. |
| 86 | |
| 87 | b) does the CPU have counter/compare registers? |
| 88 | |
| 89 | If the answer is no, you need a timer to provide the timer interrupt |
| 90 | at 100 HZ speed. |
| 91 | |
| 92 | You cannot use the fast gettimeoffset functions, i.e., |
| 93 | |
| 94 | unsigned long fixed_rate_gettimeoffset(void); |
| 95 | unsigned long calibrate_div32_gettimeoffset(void); |
| 96 | unsigned long calibrate_div64_gettimeoffset(void); |
| 97 | |
| 98 | You can use null_gettimeoffset() will gives the same time resolution as |
| 99 | jiffy. Or you can implement your own gettimeoffset (probably based on |
| 100 | some ad hoc hardware on your machine.) |
| 101 | |
| 102 | c) The following sub steps assume your CPU has counter register. |
| 103 | Do you plan to use the CPU counter register as the timer interrupt |
| 104 | or use an exnternal timer? |
| 105 | |
| 106 | In order to use CPU counter register as the timer interrupt source, you |
| 107 | must know the counter speed (mips_counter_frequency). It is usually the |
| 108 | same as the CPU speed or an integral divisor of it. |
| 109 | |
| 110 | d) decide on whether you want to use high-level or low-level timer |
| 111 | interrupt routines. The low-level one is presumably faster, but should |
| 112 | not make too mcuh difference. |
| 113 | |
| 114 | |
| 115 | Step 2: the machine setup() function |
| 116 | |
| 117 | If you supply board_time_init(), set the function poointer. |
| 118 | |
| 119 | Set the function pointer board_timer_setup() (mandatory) |
| 120 | |
| 121 | |
| 122 | Step 3: implement rtc routines, board_time_init() and board_timer_setup() |
| 123 | if needed. |
| 124 | |
| 125 | board_time_init() - |
| 126 | a) (optional) set up RTC routines, |
| 127 | b) (optional) calibrate and set the mips_counter_frequency |
| 128 | (only needed if you intended to use fixed_rate_gettimeoffset |
| 129 | or use cpu counter as timer interrupt source) |
| 130 | |
| 131 | board_timer_setup() - |
| 132 | a) (optional) over-write any choices made above by time_init(). |
| 133 | b) machine specific code should setup the timer irqaction. |
| 134 | c) enable the timer interrupt |
| 135 | |
| 136 | |
| 137 | If the RTC chip is a common chip, I suggest the routines are put under |
| 138 | arch/mips/libs. For example, for DS1386 chip, one would create |
| 139 | rtc-ds1386.c under arch/mips/lib directory. Add the following line to |
| 140 | the arch/mips/lib/Makefile: |
| 141 | |
| 142 | obj-$(CONFIG_DDB5476) += rtc-ds1386.o |
| 143 | |
| 144 | Step 4: if you are using low-level timer interrupt, change your interrupt |
| 145 | dispathcing code to check for timer interrupt and jump to |
| 146 | ll_timer_interrupt() directly if one is detected. |
| 147 | |
| 148 | Step 5: Modify arch/mips/config.in and add CONFIG_NEW_TIME_C to your machine. |
| 149 | Modify the appropriate defconfig if applicable. |
| 150 | |
| 151 | Final notes: |
| 152 | |
| 153 | For some tricky cases, you may need to add your own wrapper functions |
| 154 | for some of the functions in time.c. |
| 155 | |
| 156 | For example, you may define your own timer interrupt routine, which does |
| 157 | some of its own processing and then calls timer_interrupt(). |
| 158 | |
| 159 | You can also over-ride any of the built-in functions (gettimeoffset, |
| 160 | RTC routines and/or timer interrupt routine). |
| 161 | |
| 162 | |
| 163 | PORTING NOTES FOR SMP |
| 164 | ---------------------- |
| 165 | |
| 166 | If you have a SMP box, things are slightly more complicated. |
| 167 | |
| 168 | The time service running every jiffy is logically divided into two parts: |
| 169 | |
| 170 | 1) the one for the whole system (defined in timer_interrupt()) |
| 171 | 2) the one that should run for each CPU (defined in local_timer_interrupt()) |
| 172 | |
| 173 | You need to decide on your timer interrupt sources. |
| 174 | |
| 175 | case 1) - whole system has only one timer interrupt delivered to one CPU |
| 176 | |
| 177 | In this case, you set up timer interrupt as in UP systems. In addtion, |
| 178 | you need to set emulate_local_timer_interrupt to 1 so that other |
| 179 | CPUs get to call local_timer_interrupt(). |
| 180 | |
| 181 | THIS IS CURRENTLY NOT IMPLEMNETED. However, it is rather easy to write |
| 182 | one should such a need arise. You simply make a IPI call. |
| 183 | |
| 184 | case 2) - each CPU has a separate timer interrupt |
| 185 | |
| 186 | In this case, you need to set up IRQ such that each of them will |
| 187 | call local_timer_interrupt(). In addition, you need to arrange |
| 188 | one and only one of them to call timer_interrupt(). |
| 189 | |
| 190 | You can also do the low-level version of those interrupt routines, |
| 191 | following similar dispatching routes described above. |
| 192 | |
| 193 | Note about do_gettimeoffset(): |
| 194 | |
| 195 | It is very likely the CPU counter registers are not sync'ed up in a SMP box. |
| 196 | Therefore you cannot really use the many of the existing routines that |
| 197 | are based on CPU counter. You should wirte your own gettimeoffset rouinte |
| 198 | if you want intra-jiffy resolution. |