Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * Common time routines among all ppc machines. |
3 | * | |
4 | * Written by Cort Dougan (cort@cs.nmt.edu) to merge | |
5 | * Paul Mackerras' version and mine for PReP and Pmac. | |
6 | * MPC8xx/MBX changes by Dan Malek (dmalek@jlc.net). | |
7 | * Converted for 64-bit by Mike Corrigan (mikejc@us.ibm.com) | |
8 | * | |
9 | * First round of bugfixes by Gabriel Paubert (paubert@iram.es) | |
10 | * to make clock more stable (2.4.0-test5). The only thing | |
11 | * that this code assumes is that the timebases have been synchronized | |
12 | * by firmware on SMP and are never stopped (never do sleep | |
13 | * on SMP then, nap and doze are OK). | |
14 | * | |
15 | * Speeded up do_gettimeofday by getting rid of references to | |
16 | * xtime (which required locks for consistency). (mikejc@us.ibm.com) | |
17 | * | |
18 | * TODO (not necessarily in this file): | |
19 | * - improve precision and reproducibility of timebase frequency | |
f5339277 | 20 | * measurement at boot time. |
1da177e4 LT |
21 | * - for astronomical applications: add a new function to get |
22 | * non ambiguous timestamps even around leap seconds. This needs | |
23 | * a new timestamp format and a good name. | |
24 | * | |
25 | * 1997-09-10 Updated NTP code according to technical memorandum Jan '96 | |
26 | * "A Kernel Model for Precision Timekeeping" by Dave Mills | |
27 | * | |
28 | * This program is free software; you can redistribute it and/or | |
29 | * modify it under the terms of the GNU General Public License | |
30 | * as published by the Free Software Foundation; either version | |
31 | * 2 of the License, or (at your option) any later version. | |
32 | */ | |
33 | ||
1da177e4 | 34 | #include <linux/errno.h> |
4b16f8e2 | 35 | #include <linux/export.h> |
1da177e4 | 36 | #include <linux/sched.h> |
e6017571 | 37 | #include <linux/sched/clock.h> |
1da177e4 LT |
38 | #include <linux/kernel.h> |
39 | #include <linux/param.h> | |
40 | #include <linux/string.h> | |
41 | #include <linux/mm.h> | |
42 | #include <linux/interrupt.h> | |
43 | #include <linux/timex.h> | |
44 | #include <linux/kernel_stat.h> | |
1da177e4 | 45 | #include <linux/time.h> |
0d948730 | 46 | #include <linux/clockchips.h> |
1da177e4 LT |
47 | #include <linux/init.h> |
48 | #include <linux/profile.h> | |
49 | #include <linux/cpu.h> | |
50 | #include <linux/security.h> | |
f2783c15 PM |
51 | #include <linux/percpu.h> |
52 | #include <linux/rtc.h> | |
092b8f34 | 53 | #include <linux/jiffies.h> |
c6622f63 | 54 | #include <linux/posix-timers.h> |
7d12e780 | 55 | #include <linux/irq.h> |
177996e6 | 56 | #include <linux/delay.h> |
e360adbe | 57 | #include <linux/irq_work.h> |
f0d37300 | 58 | #include <linux/clk-provider.h> |
7f92bc56 | 59 | #include <linux/suspend.h> |
169047f4 | 60 | #include <linux/rtc.h> |
32ef5517 | 61 | #include <linux/sched/cputime.h> |
4e287e65 | 62 | #include <linux/processor.h> |
6795b85c | 63 | #include <asm/trace.h> |
1da177e4 | 64 | |
1da177e4 | 65 | #include <asm/io.h> |
1da177e4 LT |
66 | #include <asm/nvram.h> |
67 | #include <asm/cache.h> | |
68 | #include <asm/machdep.h> | |
7c0f6ba6 | 69 | #include <linux/uaccess.h> |
1da177e4 | 70 | #include <asm/time.h> |
1da177e4 | 71 | #include <asm/prom.h> |
f2783c15 PM |
72 | #include <asm/irq.h> |
73 | #include <asm/div64.h> | |
2249ca9d | 74 | #include <asm/smp.h> |
a7f290da | 75 | #include <asm/vdso_datapage.h> |
1ababe11 | 76 | #include <asm/firmware.h> |
0545d543 | 77 | #include <asm/asm-prototypes.h> |
1da177e4 | 78 | |
4a4cfe38 TB |
79 | /* powerpc clocksource/clockevent code */ |
80 | ||
d831d0b8 | 81 | #include <linux/clockchips.h> |
189374ae | 82 | #include <linux/timekeeper_internal.h> |
4a4cfe38 | 83 | |
a5a1d1c2 | 84 | static u64 rtc_read(struct clocksource *); |
4a4cfe38 TB |
85 | static struct clocksource clocksource_rtc = { |
86 | .name = "rtc", | |
87 | .rating = 400, | |
88 | .flags = CLOCK_SOURCE_IS_CONTINUOUS, | |
89 | .mask = CLOCKSOURCE_MASK(64), | |
4a4cfe38 TB |
90 | .read = rtc_read, |
91 | }; | |
92 | ||
a5a1d1c2 | 93 | static u64 timebase_read(struct clocksource *); |
4a4cfe38 TB |
94 | static struct clocksource clocksource_timebase = { |
95 | .name = "timebase", | |
96 | .rating = 400, | |
97 | .flags = CLOCK_SOURCE_IS_CONTINUOUS, | |
98 | .mask = CLOCKSOURCE_MASK(64), | |
4a4cfe38 TB |
99 | .read = timebase_read, |
100 | }; | |
101 | ||
79901024 OH |
102 | #define DECREMENTER_DEFAULT_MAX 0x7FFFFFFF |
103 | u64 decrementer_max = DECREMENTER_DEFAULT_MAX; | |
d831d0b8 TB |
104 | |
105 | static int decrementer_set_next_event(unsigned long evt, | |
106 | struct clock_event_device *dev); | |
37a13e78 | 107 | static int decrementer_shutdown(struct clock_event_device *evt); |
d831d0b8 | 108 | |
6e35994d | 109 | struct clock_event_device decrementer_clockevent = { |
37a13e78 VK |
110 | .name = "decrementer", |
111 | .rating = 200, | |
112 | .irq = 0, | |
113 | .set_next_event = decrementer_set_next_event, | |
81759360 | 114 | .set_state_oneshot_stopped = decrementer_shutdown, |
37a13e78 VK |
115 | .set_state_shutdown = decrementer_shutdown, |
116 | .tick_resume = decrementer_shutdown, | |
117 | .features = CLOCK_EVT_FEAT_ONESHOT | | |
118 | CLOCK_EVT_FEAT_C3STOP, | |
d831d0b8 | 119 | }; |
6e35994d | 120 | EXPORT_SYMBOL(decrementer_clockevent); |
d831d0b8 | 121 | |
7df10275 AB |
122 | DEFINE_PER_CPU(u64, decrementers_next_tb); |
123 | static DEFINE_PER_CPU(struct clock_event_device, decrementers); | |
d831d0b8 | 124 | |
1da177e4 LT |
125 | #define XSEC_PER_SEC (1024*1024) |
126 | ||
f2783c15 PM |
127 | #ifdef CONFIG_PPC64 |
128 | #define SCALE_XSEC(xsec, max) (((xsec) * max) / XSEC_PER_SEC) | |
129 | #else | |
130 | /* compute ((xsec << 12) * max) >> 32 */ | |
131 | #define SCALE_XSEC(xsec, max) mulhwu((xsec) << 12, max) | |
132 | #endif | |
133 | ||
1da177e4 LT |
134 | unsigned long tb_ticks_per_jiffy; |
135 | unsigned long tb_ticks_per_usec = 100; /* sane default */ | |
136 | EXPORT_SYMBOL(tb_ticks_per_usec); | |
137 | unsigned long tb_ticks_per_sec; | |
2cf82c02 | 138 | EXPORT_SYMBOL(tb_ticks_per_sec); /* for cputime_t conversions */ |
092b8f34 | 139 | |
1da177e4 | 140 | DEFINE_SPINLOCK(rtc_lock); |
6ae3db11 | 141 | EXPORT_SYMBOL_GPL(rtc_lock); |
1da177e4 | 142 | |
fc9069fe TB |
143 | static u64 tb_to_ns_scale __read_mostly; |
144 | static unsigned tb_to_ns_shift __read_mostly; | |
364a1246 | 145 | static u64 boot_tb __read_mostly; |
1da177e4 | 146 | |
1da177e4 | 147 | extern struct timezone sys_tz; |
f2783c15 | 148 | static long timezone_offset; |
1da177e4 | 149 | |
10f7e7c1 | 150 | unsigned long ppc_proc_freq; |
55ec2fca | 151 | EXPORT_SYMBOL_GPL(ppc_proc_freq); |
10f7e7c1 | 152 | unsigned long ppc_tb_freq; |
55ec2fca | 153 | EXPORT_SYMBOL_GPL(ppc_tb_freq); |
96c44507 | 154 | |
abf917cd | 155 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
c6622f63 | 156 | /* |
e7f340ca FW |
157 | * Factor for converting from cputime_t (timebase ticks) to |
158 | * microseconds. This is stored as 0.64 fixed-point binary fraction. | |
c6622f63 | 159 | */ |
9f5072d4 AS |
160 | u64 __cputime_usec_factor; |
161 | EXPORT_SYMBOL(__cputime_usec_factor); | |
a42548a1 | 162 | |
c223c903 | 163 | #ifdef CONFIG_PPC_SPLPAR |
872e439a | 164 | void (*dtl_consumer)(struct dtl_entry *, u64); |
c223c903 CL |
165 | #endif |
166 | ||
c6622f63 PM |
167 | static void calc_cputime_factors(void) |
168 | { | |
169 | struct div_result res; | |
170 | ||
9f5072d4 AS |
171 | div128_by_32(1000000, 0, tb_ticks_per_sec, &res); |
172 | __cputime_usec_factor = res.result_low; | |
c6622f63 PM |
173 | } |
174 | ||
175 | /* | |
cf9efce0 PM |
176 | * Read the SPURR on systems that have it, otherwise the PURR, |
177 | * or if that doesn't exist return the timebase value passed in. | |
c6622f63 | 178 | */ |
abcff86d | 179 | static inline unsigned long read_spurr(unsigned long tb) |
c6622f63 | 180 | { |
cf9efce0 PM |
181 | if (cpu_has_feature(CPU_FTR_SPURR)) |
182 | return mfspr(SPRN_SPURR); | |
c6622f63 PM |
183 | if (cpu_has_feature(CPU_FTR_PURR)) |
184 | return mfspr(SPRN_PURR); | |
cf9efce0 | 185 | return tb; |
c6622f63 PM |
186 | } |
187 | ||
cf9efce0 PM |
188 | #ifdef CONFIG_PPC_SPLPAR |
189 | ||
4603ac18 | 190 | /* |
cf9efce0 PM |
191 | * Scan the dispatch trace log and count up the stolen time. |
192 | * Should be called with interrupts disabled. | |
4603ac18 | 193 | */ |
cf9efce0 | 194 | static u64 scan_dispatch_log(u64 stop_tb) |
4603ac18 | 195 | { |
872e439a | 196 | u64 i = local_paca->dtl_ridx; |
cf9efce0 PM |
197 | struct dtl_entry *dtl = local_paca->dtl_curr; |
198 | struct dtl_entry *dtl_end = local_paca->dispatch_log_end; | |
199 | struct lppaca *vpa = local_paca->lppaca_ptr; | |
200 | u64 tb_delta; | |
201 | u64 stolen = 0; | |
202 | u64 dtb; | |
203 | ||
84ffae55 AB |
204 | if (!dtl) |
205 | return 0; | |
206 | ||
7ffcf8ec | 207 | if (i == be64_to_cpu(vpa->dtl_idx)) |
cf9efce0 | 208 | return 0; |
7ffcf8ec | 209 | while (i < be64_to_cpu(vpa->dtl_idx)) { |
7ffcf8ec AB |
210 | dtb = be64_to_cpu(dtl->timebase); |
211 | tb_delta = be32_to_cpu(dtl->enqueue_to_dispatch_time) + | |
212 | be32_to_cpu(dtl->ready_to_enqueue_time); | |
cf9efce0 | 213 | barrier(); |
7ffcf8ec | 214 | if (i + N_DISPATCH_LOG < be64_to_cpu(vpa->dtl_idx)) { |
cf9efce0 | 215 | /* buffer has overflowed */ |
7ffcf8ec | 216 | i = be64_to_cpu(vpa->dtl_idx) - N_DISPATCH_LOG; |
cf9efce0 PM |
217 | dtl = local_paca->dispatch_log + (i % N_DISPATCH_LOG); |
218 | continue; | |
219 | } | |
220 | if (dtb > stop_tb) | |
221 | break; | |
84b07386 AB |
222 | if (dtl_consumer) |
223 | dtl_consumer(dtl, i); | |
cf9efce0 PM |
224 | stolen += tb_delta; |
225 | ++i; | |
226 | ++dtl; | |
227 | if (dtl == dtl_end) | |
228 | dtl = local_paca->dispatch_log; | |
229 | } | |
230 | local_paca->dtl_ridx = i; | |
231 | local_paca->dtl_curr = dtl; | |
232 | return stolen; | |
4603ac18 MN |
233 | } |
234 | ||
cf9efce0 PM |
235 | /* |
236 | * Accumulate stolen time by scanning the dispatch trace log. | |
237 | * Called on entry from user mode. | |
238 | */ | |
239 | void accumulate_stolen_time(void) | |
240 | { | |
241 | u64 sst, ust; | |
4e26bc4a | 242 | unsigned long save_irq_soft_mask = irq_soft_mask_return(); |
c223c903 | 243 | struct cpu_accounting_data *acct = &local_paca->accounting; |
b18ae08d TH |
244 | |
245 | /* We are called early in the exception entry, before | |
246 | * soft/hard_enabled are sync'ed to the expected state | |
247 | * for the exception. We are hard disabled but the PACA | |
248 | * needs to reflect that so various debug stuff doesn't | |
249 | * complain | |
250 | */ | |
4e26bc4a | 251 | irq_soft_mask_set(IRQS_DISABLED); |
b18ae08d | 252 | |
c223c903 CL |
253 | sst = scan_dispatch_log(acct->starttime_user); |
254 | ust = scan_dispatch_log(acct->starttime); | |
8c8b73c4 FW |
255 | acct->stime -= sst; |
256 | acct->utime -= ust; | |
f828c3d0 | 257 | acct->steal_time += ust + sst; |
b18ae08d | 258 | |
4e26bc4a | 259 | irq_soft_mask_set(save_irq_soft_mask); |
cf9efce0 PM |
260 | } |
261 | ||
262 | static inline u64 calculate_stolen_time(u64 stop_tb) | |
263 | { | |
a6201da3 AK |
264 | if (!firmware_has_feature(FW_FEATURE_SPLPAR)) |
265 | return 0; | |
266 | ||
a19ff1a2 FW |
267 | if (get_paca()->dtl_ridx != be64_to_cpu(get_lppaca()->dtl_idx)) |
268 | return scan_dispatch_log(stop_tb); | |
cf9efce0 | 269 | |
a19ff1a2 | 270 | return 0; |
4603ac18 MN |
271 | } |
272 | ||
cf9efce0 PM |
273 | #else /* CONFIG_PPC_SPLPAR */ |
274 | static inline u64 calculate_stolen_time(u64 stop_tb) | |
275 | { | |
276 | return 0; | |
277 | } | |
278 | ||
279 | #endif /* CONFIG_PPC_SPLPAR */ | |
280 | ||
c6622f63 PM |
281 | /* |
282 | * Account time for a transition between system, hard irq | |
283 | * or soft irq state. | |
284 | */ | |
b38a181c CL |
285 | static unsigned long vtime_delta_scaled(struct cpu_accounting_data *acct, |
286 | unsigned long now, unsigned long stime) | |
c6622f63 | 287 | { |
abcff86d CL |
288 | unsigned long stime_scaled = 0; |
289 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME | |
b38a181c | 290 | unsigned long nowscaled, deltascaled; |
a19ff1a2 | 291 | unsigned long utime, utime_scaled; |
c6622f63 | 292 | |
4603ac18 | 293 | nowscaled = read_spurr(now); |
c223c903 CL |
294 | deltascaled = nowscaled - acct->startspurr; |
295 | acct->startspurr = nowscaled; | |
a19ff1a2 | 296 | utime = acct->utime - acct->utime_sspurr; |
8c8b73c4 | 297 | acct->utime_sspurr = acct->utime; |
cf9efce0 PM |
298 | |
299 | /* | |
300 | * Because we don't read the SPURR on every kernel entry/exit, | |
301 | * deltascaled includes both user and system SPURR ticks. | |
302 | * Apportion these ticks to system SPURR ticks and user | |
303 | * SPURR ticks in the same ratio as the system time (delta) | |
304 | * and user time (udelta) values obtained from the timebase | |
305 | * over the same interval. The system ticks get accounted here; | |
306 | * the user ticks get saved up in paca->user_time_scaled to be | |
307 | * used by account_process_tick. | |
308 | */ | |
b38a181c | 309 | stime_scaled = stime; |
a19ff1a2 FW |
310 | utime_scaled = utime; |
311 | if (deltascaled != stime + utime) { | |
312 | if (utime) { | |
b38a181c CL |
313 | stime_scaled = deltascaled * stime / (stime + utime); |
314 | utime_scaled = deltascaled - stime_scaled; | |
cf9efce0 | 315 | } else { |
b38a181c | 316 | stime_scaled = deltascaled; |
cf9efce0 PM |
317 | } |
318 | } | |
a19ff1a2 | 319 | acct->utime_scaled += utime_scaled; |
abcff86d | 320 | #endif |
cf9efce0 | 321 | |
b38a181c CL |
322 | return stime_scaled; |
323 | } | |
324 | ||
325 | static unsigned long vtime_delta(struct task_struct *tsk, | |
326 | unsigned long *stime_scaled, | |
327 | unsigned long *steal_time) | |
328 | { | |
329 | unsigned long now, stime; | |
330 | struct cpu_accounting_data *acct = get_accounting(tsk); | |
331 | ||
332 | WARN_ON_ONCE(!irqs_disabled()); | |
333 | ||
334 | now = mftb(); | |
335 | stime = now - acct->starttime; | |
336 | acct->starttime = now; | |
337 | ||
338 | *stime_scaled = vtime_delta_scaled(acct, now, stime); | |
339 | ||
340 | *steal_time = calculate_stolen_time(now); | |
341 | ||
a19ff1a2 | 342 | return stime; |
a7e1a9e3 FW |
343 | } |
344 | ||
fd25b4c2 | 345 | void vtime_account_system(struct task_struct *tsk) |
a7e1a9e3 | 346 | { |
a19ff1a2 FW |
347 | unsigned long stime, stime_scaled, steal_time; |
348 | struct cpu_accounting_data *acct = get_accounting(tsk); | |
349 | ||
350 | stime = vtime_delta(tsk, &stime_scaled, &steal_time); | |
351 | ||
352 | stime -= min(stime, steal_time); | |
353 | acct->steal_time += steal_time; | |
a7e1a9e3 | 354 | |
a19ff1a2 FW |
355 | if ((tsk->flags & PF_VCPU) && !irq_count()) { |
356 | acct->gtime += stime; | |
abcff86d | 357 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
a19ff1a2 | 358 | acct->utime_scaled += stime_scaled; |
abcff86d | 359 | #endif |
a19ff1a2 FW |
360 | } else { |
361 | if (hardirq_count()) | |
362 | acct->hardirq_time += stime; | |
363 | else if (in_serving_softirq()) | |
364 | acct->softirq_time += stime; | |
365 | else | |
366 | acct->stime += stime; | |
367 | ||
abcff86d | 368 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
a19ff1a2 | 369 | acct->stime_scaled += stime_scaled; |
abcff86d | 370 | #endif |
a19ff1a2 | 371 | } |
a7e1a9e3 | 372 | } |
c11f11fc | 373 | EXPORT_SYMBOL_GPL(vtime_account_system); |
a7e1a9e3 | 374 | |
fd25b4c2 | 375 | void vtime_account_idle(struct task_struct *tsk) |
a7e1a9e3 | 376 | { |
a19ff1a2 FW |
377 | unsigned long stime, stime_scaled, steal_time; |
378 | struct cpu_accounting_data *acct = get_accounting(tsk); | |
a7e1a9e3 | 379 | |
a19ff1a2 FW |
380 | stime = vtime_delta(tsk, &stime_scaled, &steal_time); |
381 | acct->idle_time += stime + steal_time; | |
c6622f63 PM |
382 | } |
383 | ||
b38a181c CL |
384 | static void vtime_flush_scaled(struct task_struct *tsk, |
385 | struct cpu_accounting_data *acct) | |
386 | { | |
abcff86d | 387 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
b38a181c CL |
388 | if (acct->utime_scaled) |
389 | tsk->utimescaled += cputime_to_nsecs(acct->utime_scaled); | |
390 | if (acct->stime_scaled) | |
391 | tsk->stimescaled += cputime_to_nsecs(acct->stime_scaled); | |
392 | ||
393 | acct->utime_scaled = 0; | |
394 | acct->utime_sspurr = 0; | |
395 | acct->stime_scaled = 0; | |
abcff86d | 396 | #endif |
b38a181c CL |
397 | } |
398 | ||
c6622f63 | 399 | /* |
c8d7dabf | 400 | * Account the whole cputime accumulated in the paca |
c6622f63 | 401 | * Must be called with interrupts disabled. |
bcebdf84 FW |
402 | * Assumes that vtime_account_system/idle() has been called |
403 | * recently (i.e. since the last entry from usermode) so that | |
cf9efce0 | 404 | * get_paca()->user_time_scaled is up to date. |
c6622f63 | 405 | */ |
c8d7dabf | 406 | void vtime_flush(struct task_struct *tsk) |
c6622f63 | 407 | { |
c223c903 | 408 | struct cpu_accounting_data *acct = get_accounting(tsk); |
c6622f63 | 409 | |
a19ff1a2 | 410 | if (acct->utime) |
23244a5c | 411 | account_user_time(tsk, cputime_to_nsecs(acct->utime)); |
a19ff1a2 | 412 | |
a19ff1a2 | 413 | if (acct->gtime) |
fb8b049c | 414 | account_guest_time(tsk, cputime_to_nsecs(acct->gtime)); |
a19ff1a2 | 415 | |
51eeef9e | 416 | if (IS_ENABLED(CONFIG_PPC_SPLPAR) && acct->steal_time) { |
be9095ed | 417 | account_steal_time(cputime_to_nsecs(acct->steal_time)); |
51eeef9e CL |
418 | acct->steal_time = 0; |
419 | } | |
a19ff1a2 FW |
420 | |
421 | if (acct->idle_time) | |
18b43a9b | 422 | account_idle_time(cputime_to_nsecs(acct->idle_time)); |
a19ff1a2 FW |
423 | |
424 | if (acct->stime) | |
fb8b049c FW |
425 | account_system_index_time(tsk, cputime_to_nsecs(acct->stime), |
426 | CPUTIME_SYSTEM); | |
a19ff1a2 FW |
427 | |
428 | if (acct->hardirq_time) | |
fb8b049c FW |
429 | account_system_index_time(tsk, cputime_to_nsecs(acct->hardirq_time), |
430 | CPUTIME_IRQ); | |
a19ff1a2 | 431 | if (acct->softirq_time) |
fb8b049c FW |
432 | account_system_index_time(tsk, cputime_to_nsecs(acct->softirq_time), |
433 | CPUTIME_SOFTIRQ); | |
a19ff1a2 | 434 | |
b38a181c CL |
435 | vtime_flush_scaled(tsk, acct); |
436 | ||
8c8b73c4 | 437 | acct->utime = 0; |
a19ff1a2 | 438 | acct->gtime = 0; |
a19ff1a2 FW |
439 | acct->idle_time = 0; |
440 | acct->stime = 0; | |
a19ff1a2 FW |
441 | acct->hardirq_time = 0; |
442 | acct->softirq_time = 0; | |
c6622f63 PM |
443 | } |
444 | ||
abf917cd | 445 | #else /* ! CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */ |
c6622f63 | 446 | #define calc_cputime_factors() |
c6622f63 PM |
447 | #endif |
448 | ||
6defa38b PM |
449 | void __delay(unsigned long loops) |
450 | { | |
451 | unsigned long start; | |
452 | int diff; | |
453 | ||
4e287e65 | 454 | spin_begin(); |
6defa38b PM |
455 | if (__USE_RTC()) { |
456 | start = get_rtcl(); | |
457 | do { | |
458 | /* the RTCL register wraps at 1000000000 */ | |
459 | diff = get_rtcl() - start; | |
460 | if (diff < 0) | |
461 | diff += 1000000000; | |
4e287e65 | 462 | spin_cpu_relax(); |
6defa38b PM |
463 | } while (diff < loops); |
464 | } else { | |
465 | start = get_tbl(); | |
466 | while (get_tbl() - start < loops) | |
4e287e65 | 467 | spin_cpu_relax(); |
6defa38b | 468 | } |
4e287e65 | 469 | spin_end(); |
6defa38b PM |
470 | } |
471 | EXPORT_SYMBOL(__delay); | |
472 | ||
473 | void udelay(unsigned long usecs) | |
474 | { | |
475 | __delay(tb_ticks_per_usec * usecs); | |
476 | } | |
477 | EXPORT_SYMBOL(udelay); | |
478 | ||
1da177e4 LT |
479 | #ifdef CONFIG_SMP |
480 | unsigned long profile_pc(struct pt_regs *regs) | |
481 | { | |
482 | unsigned long pc = instruction_pointer(regs); | |
483 | ||
484 | if (in_lock_functions(pc)) | |
485 | return regs->link; | |
486 | ||
487 | return pc; | |
488 | } | |
489 | EXPORT_SYMBOL(profile_pc); | |
490 | #endif | |
491 | ||
e360adbe | 492 | #ifdef CONFIG_IRQ_WORK |
105988c0 | 493 | |
0fe1ac48 PM |
494 | /* |
495 | * 64-bit uses a byte in the PACA, 32-bit uses a per-cpu variable... | |
496 | */ | |
497 | #ifdef CONFIG_PPC64 | |
e360adbe | 498 | static inline unsigned long test_irq_work_pending(void) |
105988c0 | 499 | { |
0fe1ac48 PM |
500 | unsigned long x; |
501 | ||
502 | asm volatile("lbz %0,%1(13)" | |
503 | : "=r" (x) | |
e360adbe | 504 | : "i" (offsetof(struct paca_struct, irq_work_pending))); |
0fe1ac48 PM |
505 | return x; |
506 | } | |
507 | ||
e360adbe | 508 | static inline void set_irq_work_pending_flag(void) |
0fe1ac48 PM |
509 | { |
510 | asm volatile("stb %0,%1(13)" : : | |
511 | "r" (1), | |
e360adbe | 512 | "i" (offsetof(struct paca_struct, irq_work_pending))); |
0fe1ac48 PM |
513 | } |
514 | ||
e360adbe | 515 | static inline void clear_irq_work_pending(void) |
0fe1ac48 PM |
516 | { |
517 | asm volatile("stb %0,%1(13)" : : | |
518 | "r" (0), | |
e360adbe | 519 | "i" (offsetof(struct paca_struct, irq_work_pending))); |
105988c0 PM |
520 | } |
521 | ||
ebb37cf3 NP |
522 | void arch_irq_work_raise(void) |
523 | { | |
524 | preempt_disable(); | |
525 | set_irq_work_pending_flag(); | |
526 | /* | |
527 | * Non-nmi code running with interrupts disabled will replay | |
528 | * irq_happened before it re-enables interrupts, so setthe | |
529 | * decrementer there instead of causing a hardware exception | |
530 | * which would immediately hit the masked interrupt handler | |
531 | * and have the net effect of setting the decrementer in | |
532 | * irq_happened. | |
533 | * | |
534 | * NMI interrupts can not check this when they return, so the | |
535 | * decrementer hardware exception is raised, which will fire | |
536 | * when interrupts are next enabled. | |
537 | * | |
538 | * BookE does not support this yet, it must audit all NMI | |
539 | * interrupt handlers to ensure they call nmi_enter() so this | |
540 | * check would be correct. | |
541 | */ | |
542 | if (IS_ENABLED(CONFIG_BOOKE) || !irqs_disabled() || in_nmi()) { | |
543 | set_dec(1); | |
544 | } else { | |
545 | hard_irq_disable(); | |
546 | local_paca->irq_happened |= PACA_IRQ_DEC; | |
547 | } | |
548 | preempt_enable(); | |
549 | } | |
550 | ||
0fe1ac48 PM |
551 | #else /* 32-bit */ |
552 | ||
e360adbe | 553 | DEFINE_PER_CPU(u8, irq_work_pending); |
0fe1ac48 | 554 | |
69111bac CL |
555 | #define set_irq_work_pending_flag() __this_cpu_write(irq_work_pending, 1) |
556 | #define test_irq_work_pending() __this_cpu_read(irq_work_pending) | |
557 | #define clear_irq_work_pending() __this_cpu_write(irq_work_pending, 0) | |
105988c0 | 558 | |
4f8b50bb | 559 | void arch_irq_work_raise(void) |
0fe1ac48 PM |
560 | { |
561 | preempt_disable(); | |
e360adbe | 562 | set_irq_work_pending_flag(); |
0fe1ac48 PM |
563 | set_dec(1); |
564 | preempt_enable(); | |
565 | } | |
566 | ||
ebb37cf3 NP |
567 | #endif /* 32 vs 64 bit */ |
568 | ||
e360adbe | 569 | #else /* CONFIG_IRQ_WORK */ |
105988c0 | 570 | |
e360adbe PZ |
571 | #define test_irq_work_pending() 0 |
572 | #define clear_irq_work_pending() | |
105988c0 | 573 | |
e360adbe | 574 | #endif /* CONFIG_IRQ_WORK */ |
105988c0 | 575 | |
1da177e4 LT |
576 | /* |
577 | * timer_interrupt - gets called when the decrementer overflows, | |
578 | * with interrupts disabled. | |
579 | */ | |
3f984620 | 580 | void timer_interrupt(struct pt_regs *regs) |
1da177e4 | 581 | { |
3f984620 | 582 | struct clock_event_device *evt = this_cpu_ptr(&decrementers); |
69111bac | 583 | u64 *next_tb = this_cpu_ptr(&decrementers_next_tb); |
3f984620 NP |
584 | struct pt_regs *old_regs; |
585 | u64 now; | |
d831d0b8 | 586 | |
963e5d3b | 587 | /* Some implementations of hotplug will get timer interrupts while |
689dfa89 TC |
588 | * offline, just ignore these and we also need to set |
589 | * decrementers_next_tb as MAX to make sure __check_irq_replay | |
590 | * don't replay timer interrupt when return, otherwise we'll trap | |
591 | * here infinitely :( | |
963e5d3b | 592 | */ |
a7cba02d | 593 | if (unlikely(!cpu_online(smp_processor_id()))) { |
689dfa89 | 594 | *next_tb = ~(u64)0; |
a7cba02d | 595 | set_dec(decrementer_max); |
963e5d3b | 596 | return; |
689dfa89 | 597 | } |
963e5d3b | 598 | |
a7cba02d NP |
599 | /* Ensure a positive value is written to the decrementer, or else |
600 | * some CPUs will continue to take decrementer exceptions. When the | |
601 | * PPC_WATCHDOG (decrementer based) is configured, keep this at most | |
602 | * 31 bits, which is about 4 seconds on most systems, which gives | |
603 | * the watchdog a chance of catching timer interrupt hard lockups. | |
604 | */ | |
605 | if (IS_ENABLED(CONFIG_PPC_WATCHDOG)) | |
606 | set_dec(0x7fffffff); | |
607 | else | |
608 | set_dec(decrementer_max); | |
609 | ||
7230c564 BH |
610 | /* Conditionally hard-enable interrupts now that the DEC has been |
611 | * bumped to its maximum value | |
612 | */ | |
613 | may_hard_irq_enable(); | |
614 | ||
89713ed1 | 615 | |
6e0fdf9a | 616 | #if defined(CONFIG_PPC32) && defined(CONFIG_PPC_PMAC) |
f2783c15 PM |
617 | if (atomic_read(&ppc_n_lost_interrupts) != 0) |
618 | do_IRQ(regs); | |
619 | #endif | |
1da177e4 | 620 | |
7d12e780 | 621 | old_regs = set_irq_regs(regs); |
1da177e4 | 622 | irq_enter(); |
3f984620 NP |
623 | trace_timer_interrupt_entry(regs); |
624 | ||
625 | if (test_irq_work_pending()) { | |
626 | clear_irq_work_pending(); | |
627 | irq_work_run(); | |
628 | } | |
629 | ||
630 | now = get_tb_or_rtc(); | |
631 | if (now >= *next_tb) { | |
632 | *next_tb = ~(u64)0; | |
633 | if (evt->event_handler) | |
634 | evt->event_handler(evt); | |
635 | __this_cpu_inc(irq_stat.timer_irqs_event); | |
636 | } else { | |
637 | now = *next_tb - now; | |
638 | if (now <= decrementer_max) | |
639 | set_dec(now); | |
640 | /* We may have raced with new irq work */ | |
641 | if (test_irq_work_pending()) | |
642 | set_dec(1); | |
643 | __this_cpu_inc(irq_stat.timer_irqs_others); | |
644 | } | |
1da177e4 | 645 | |
3f984620 | 646 | trace_timer_interrupt_exit(regs); |
1da177e4 | 647 | irq_exit(); |
7d12e780 | 648 | set_irq_regs(old_regs); |
1da177e4 | 649 | } |
9445aa1a | 650 | EXPORT_SYMBOL(timer_interrupt); |
1da177e4 | 651 | |
bc907113 | 652 | #ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST |
3f984620 NP |
653 | void timer_broadcast_interrupt(void) |
654 | { | |
655 | u64 *next_tb = this_cpu_ptr(&decrementers_next_tb); | |
3f984620 | 656 | |
3f984620 NP |
657 | *next_tb = ~(u64)0; |
658 | tick_receive_broadcast(); | |
e360cd37 | 659 | __this_cpu_inc(irq_stat.broadcast_irqs_event); |
3f984620 | 660 | } |
bc907113 | 661 | #endif |
3f984620 | 662 | |
dabe859e PM |
663 | /* |
664 | * Hypervisor decrementer interrupts shouldn't occur but are sometimes | |
665 | * left pending on exit from a KVM guest. We don't need to do anything | |
666 | * to clear them, as they are edge-triggered. | |
667 | */ | |
668 | void hdec_interrupt(struct pt_regs *regs) | |
669 | { | |
670 | } | |
671 | ||
7ac5dde9 | 672 | #ifdef CONFIG_SUSPEND |
d75d68cf | 673 | static void generic_suspend_disable_irqs(void) |
7ac5dde9 | 674 | { |
7ac5dde9 SW |
675 | /* Disable the decrementer, so that it doesn't interfere |
676 | * with suspending. | |
677 | */ | |
678 | ||
79901024 | 679 | set_dec(decrementer_max); |
7ac5dde9 | 680 | local_irq_disable(); |
79901024 | 681 | set_dec(decrementer_max); |
7ac5dde9 SW |
682 | } |
683 | ||
d75d68cf | 684 | static void generic_suspend_enable_irqs(void) |
7ac5dde9 | 685 | { |
7ac5dde9 | 686 | local_irq_enable(); |
7ac5dde9 SW |
687 | } |
688 | ||
689 | /* Overrides the weak version in kernel/power/main.c */ | |
690 | void arch_suspend_disable_irqs(void) | |
691 | { | |
692 | if (ppc_md.suspend_disable_irqs) | |
693 | ppc_md.suspend_disable_irqs(); | |
694 | generic_suspend_disable_irqs(); | |
695 | } | |
696 | ||
697 | /* Overrides the weak version in kernel/power/main.c */ | |
698 | void arch_suspend_enable_irqs(void) | |
699 | { | |
700 | generic_suspend_enable_irqs(); | |
701 | if (ppc_md.suspend_enable_irqs) | |
702 | ppc_md.suspend_enable_irqs(); | |
703 | } | |
704 | #endif | |
705 | ||
b6c295df PM |
706 | unsigned long long tb_to_ns(unsigned long long ticks) |
707 | { | |
708 | return mulhdu(ticks, tb_to_ns_scale) << tb_to_ns_shift; | |
709 | } | |
710 | EXPORT_SYMBOL_GPL(tb_to_ns); | |
711 | ||
1da177e4 LT |
712 | /* |
713 | * Scheduler clock - returns current time in nanosec units. | |
714 | * | |
715 | * Note: mulhdu(a, b) (multiply high double unsigned) returns | |
716 | * the high 64 bits of a * b, i.e. (a * b) >> 64, where a and b | |
717 | * are 64-bit unsigned numbers. | |
718 | */ | |
6b847d79 | 719 | notrace unsigned long long sched_clock(void) |
1da177e4 | 720 | { |
96c44507 PM |
721 | if (__USE_RTC()) |
722 | return get_rtc(); | |
fc9069fe | 723 | return mulhdu(get_tb() - boot_tb, tb_to_ns_scale) << tb_to_ns_shift; |
1da177e4 LT |
724 | } |
725 | ||
4be1b297 CB |
726 | |
727 | #ifdef CONFIG_PPC_PSERIES | |
728 | ||
729 | /* | |
730 | * Running clock - attempts to give a view of time passing for a virtualised | |
731 | * kernels. | |
732 | * Uses the VTB register if available otherwise a next best guess. | |
733 | */ | |
734 | unsigned long long running_clock(void) | |
735 | { | |
736 | /* | |
737 | * Don't read the VTB as a host since KVM does not switch in host | |
738 | * timebase into the VTB when it takes a guest off the CPU, reading the | |
739 | * VTB would result in reading 'last switched out' guest VTB. | |
740 | * | |
741 | * Host kernels are often compiled with CONFIG_PPC_PSERIES checked, it | |
742 | * would be unsafe to rely only on the #ifdef above. | |
743 | */ | |
744 | if (firmware_has_feature(FW_FEATURE_LPAR) && | |
745 | cpu_has_feature(CPU_FTR_ARCH_207S)) | |
746 | return mulhdu(get_vtb() - boot_tb, tb_to_ns_scale) << tb_to_ns_shift; | |
747 | ||
748 | /* | |
749 | * This is a next best approximation without a VTB. | |
750 | * On a host which is running bare metal there should never be any stolen | |
751 | * time and on a host which doesn't do any virtualisation TB *should* equal | |
752 | * VTB so it makes no difference anyway. | |
753 | */ | |
9f3768e0 | 754 | return local_clock() - kcpustat_this_cpu->cpustat[CPUTIME_STEAL]; |
4be1b297 CB |
755 | } |
756 | #endif | |
757 | ||
0bb474a4 | 758 | static int __init get_freq(char *name, int cells, unsigned long *val) |
10f7e7c1 AB |
759 | { |
760 | struct device_node *cpu; | |
6f7aba7b | 761 | const __be32 *fp; |
0bb474a4 | 762 | int found = 0; |
10f7e7c1 | 763 | |
0bb474a4 | 764 | /* The cpu node should have timebase and clock frequency properties */ |
10f7e7c1 AB |
765 | cpu = of_find_node_by_type(NULL, "cpu"); |
766 | ||
d8a8188d | 767 | if (cpu) { |
e2eb6392 | 768 | fp = of_get_property(cpu, name, NULL); |
d8a8188d | 769 | if (fp) { |
0bb474a4 | 770 | found = 1; |
a4dc7ff0 | 771 | *val = of_read_ulong(fp, cells); |
10f7e7c1 | 772 | } |
0bb474a4 AB |
773 | |
774 | of_node_put(cpu); | |
10f7e7c1 | 775 | } |
0bb474a4 AB |
776 | |
777 | return found; | |
778 | } | |
779 | ||
e51df2c1 | 780 | static void start_cpu_decrementer(void) |
77c0a700 BH |
781 | { |
782 | #if defined(CONFIG_BOOKE) || defined(CONFIG_40x) | |
6e2f03e2 IM |
783 | unsigned int tcr; |
784 | ||
77c0a700 BH |
785 | /* Clear any pending timer interrupts */ |
786 | mtspr(SPRN_TSR, TSR_ENW | TSR_WIS | TSR_DIS | TSR_FIS); | |
787 | ||
6e2f03e2 IM |
788 | tcr = mfspr(SPRN_TCR); |
789 | /* | |
790 | * The watchdog may have already been enabled by u-boot. So leave | |
791 | * TRC[WP] (Watchdog Period) alone. | |
792 | */ | |
793 | tcr &= TCR_WP_MASK; /* Clear all bits except for TCR[WP] */ | |
794 | tcr |= TCR_DIE; /* Enable decrementer */ | |
795 | mtspr(SPRN_TCR, tcr); | |
796 | #endif | |
77c0a700 BH |
797 | } |
798 | ||
0bb474a4 AB |
799 | void __init generic_calibrate_decr(void) |
800 | { | |
801 | ppc_tb_freq = DEFAULT_TB_FREQ; /* hardcoded default */ | |
802 | ||
803 | if (!get_freq("ibm,extended-timebase-frequency", 2, &ppc_tb_freq) && | |
804 | !get_freq("timebase-frequency", 1, &ppc_tb_freq)) { | |
805 | ||
10f7e7c1 AB |
806 | printk(KERN_ERR "WARNING: Estimating decrementer frequency " |
807 | "(not found)\n"); | |
0bb474a4 | 808 | } |
10f7e7c1 | 809 | |
0bb474a4 AB |
810 | ppc_proc_freq = DEFAULT_PROC_FREQ; /* hardcoded default */ |
811 | ||
812 | if (!get_freq("ibm,extended-clock-frequency", 2, &ppc_proc_freq) && | |
813 | !get_freq("clock-frequency", 1, &ppc_proc_freq)) { | |
814 | ||
815 | printk(KERN_ERR "WARNING: Estimating processor frequency " | |
816 | "(not found)\n"); | |
10f7e7c1 | 817 | } |
10f7e7c1 | 818 | } |
10f7e7c1 | 819 | |
5235afa8 | 820 | int update_persistent_clock64(struct timespec64 now) |
f2783c15 PM |
821 | { |
822 | struct rtc_time tm; | |
823 | ||
aa3be5f3 | 824 | if (!ppc_md.set_rtc_time) |
023f333a | 825 | return -ENODEV; |
aa3be5f3 | 826 | |
5235afa8 | 827 | rtc_time64_to_tm(now.tv_sec + 1 + timezone_offset, &tm); |
aa3be5f3 TB |
828 | |
829 | return ppc_md.set_rtc_time(&tm); | |
830 | } | |
831 | ||
5bfd6435 | 832 | static void __read_persistent_clock(struct timespec64 *ts) |
aa3be5f3 TB |
833 | { |
834 | struct rtc_time tm; | |
835 | static int first = 1; | |
836 | ||
d90246cd | 837 | ts->tv_nsec = 0; |
aa3be5f3 TB |
838 | /* XXX this is a litle fragile but will work okay in the short term */ |
839 | if (first) { | |
840 | first = 0; | |
841 | if (ppc_md.time_init) | |
842 | timezone_offset = ppc_md.time_init(); | |
843 | ||
844 | /* get_boot_time() isn't guaranteed to be safe to call late */ | |
d90246cd MS |
845 | if (ppc_md.get_boot_time) { |
846 | ts->tv_sec = ppc_md.get_boot_time() - timezone_offset; | |
847 | return; | |
848 | } | |
849 | } | |
850 | if (!ppc_md.get_rtc_time) { | |
851 | ts->tv_sec = 0; | |
852 | return; | |
aa3be5f3 | 853 | } |
f2783c15 | 854 | ppc_md.get_rtc_time(&tm); |
978d7eb3 | 855 | |
5bfd6435 | 856 | ts->tv_sec = rtc_tm_to_time64(&tm); |
f2783c15 PM |
857 | } |
858 | ||
5bfd6435 | 859 | void read_persistent_clock64(struct timespec64 *ts) |
978d7eb3 BH |
860 | { |
861 | __read_persistent_clock(ts); | |
862 | ||
863 | /* Sanitize it in case real time clock is set below EPOCH */ | |
864 | if (ts->tv_sec < 0) { | |
865 | ts->tv_sec = 0; | |
866 | ts->tv_nsec = 0; | |
867 | } | |
868 | ||
869 | } | |
870 | ||
4a4cfe38 | 871 | /* clocksource code */ |
6b847d79 | 872 | static notrace u64 rtc_read(struct clocksource *cs) |
4a4cfe38 | 873 | { |
a5a1d1c2 | 874 | return (u64)get_rtc(); |
4a4cfe38 TB |
875 | } |
876 | ||
6b847d79 | 877 | static notrace u64 timebase_read(struct clocksource *cs) |
4a4cfe38 | 878 | { |
a5a1d1c2 | 879 | return (u64)get_tb(); |
4a4cfe38 TB |
880 | } |
881 | ||
d4cfb113 PM |
882 | |
883 | void update_vsyscall(struct timekeeper *tk) | |
4a4cfe38 | 884 | { |
d4cfb113 PM |
885 | struct timespec xt; |
886 | struct clocksource *clock = tk->tkr_mono.clock; | |
887 | u32 mult = tk->tkr_mono.mult; | |
888 | u32 shift = tk->tkr_mono.shift; | |
889 | u64 cycle_last = tk->tkr_mono.cycle_last; | |
b0797b60 | 890 | u64 new_tb_to_xs, new_stamp_xsec; |
d4cfb113 | 891 | u64 frac_sec; |
4a4cfe38 TB |
892 | |
893 | if (clock != &clocksource_timebase) | |
894 | return; | |
895 | ||
d4cfb113 PM |
896 | xt.tv_sec = tk->xtime_sec; |
897 | xt.tv_nsec = (long)(tk->tkr_mono.xtime_nsec >> tk->tkr_mono.shift); | |
898 | ||
4a4cfe38 TB |
899 | /* Make userspace gettimeofday spin until we're done. */ |
900 | ++vdso_data->tb_update_count; | |
901 | smp_mb(); | |
902 | ||
d4cfb113 PM |
903 | /* |
904 | * This computes ((2^20 / 1e9) * mult) >> shift as a | |
905 | * 0.64 fixed-point fraction. | |
906 | * The computation in the else clause below won't overflow | |
907 | * (as long as the timebase frequency is >= 1.049 MHz) | |
908 | * but loses precision because we lose the low bits of the constant | |
909 | * in the shift. Note that 19342813113834067 ~= 2^(20+64) / 1e9. | |
910 | * For a shift of 24 the error is about 0.5e-9, or about 0.5ns | |
911 | * over a second. (Shift values are usually 22, 23 or 24.) | |
912 | * For high frequency clocks such as the 512MHz timebase clock | |
913 | * on POWER[6789], the mult value is small (e.g. 32768000) | |
914 | * and so we can shift the constant by 16 initially | |
915 | * (295147905179 ~= 2^(20+64-16) / 1e9) and then do the | |
916 | * remaining shifts after the multiplication, which gives a | |
917 | * more accurate result (e.g. with mult = 32768000, shift = 24, | |
918 | * the error is only about 1.2e-12, or 0.7ns over 10 minutes). | |
919 | */ | |
920 | if (mult <= 62500000 && clock->shift >= 16) | |
921 | new_tb_to_xs = ((u64) mult * 295147905179ULL) >> (clock->shift - 16); | |
922 | else | |
923 | new_tb_to_xs = (u64) mult * (19342813113834067ULL >> clock->shift); | |
924 | ||
925 | /* | |
926 | * Compute the fractional second in units of 2^-32 seconds. | |
927 | * The fractional second is tk->tkr_mono.xtime_nsec >> tk->tkr_mono.shift | |
928 | * in nanoseconds, so multiplying that by 2^32 / 1e9 gives | |
929 | * it in units of 2^-32 seconds. | |
930 | * We assume shift <= 32 because clocks_calc_mult_shift() | |
931 | * generates shift values in the range 0 - 32. | |
932 | */ | |
933 | frac_sec = tk->tkr_mono.xtime_nsec << (32 - shift); | |
934 | do_div(frac_sec, NSEC_PER_SEC); | |
b0797b60 | 935 | |
d4cfb113 PM |
936 | /* |
937 | * Work out new stamp_xsec value for any legacy users of systemcfg. | |
938 | * stamp_xsec is in units of 2^-20 seconds. | |
939 | */ | |
940 | new_stamp_xsec = frac_sec >> 12; | |
941 | new_stamp_xsec += tk->xtime_sec * XSEC_PER_SEC; | |
47916be4 | 942 | |
b0797b60 JS |
943 | /* |
944 | * tb_update_count is used to allow the userspace gettimeofday code | |
945 | * to assure itself that it sees a consistent view of the tb_to_xs and | |
946 | * stamp_xsec variables. It reads the tb_update_count, then reads | |
947 | * tb_to_xs and stamp_xsec and then reads tb_update_count again. If | |
948 | * the two values of tb_update_count match and are even then the | |
949 | * tb_to_xs and stamp_xsec values are consistent. If not, then it | |
950 | * loops back and reads them again until this criteria is met. | |
b0797b60 | 951 | */ |
4a0e6377 | 952 | vdso_data->tb_orig_stamp = cycle_last; |
b0797b60 JS |
953 | vdso_data->stamp_xsec = new_stamp_xsec; |
954 | vdso_data->tb_to_xs = new_tb_to_xs; | |
d4cfb113 PM |
955 | vdso_data->wtom_clock_sec = tk->wall_to_monotonic.tv_sec; |
956 | vdso_data->wtom_clock_nsec = tk->wall_to_monotonic.tv_nsec; | |
957 | vdso_data->stamp_xtime = xt; | |
0e469db8 | 958 | vdso_data->stamp_sec_fraction = frac_sec; |
b0797b60 JS |
959 | smp_wmb(); |
960 | ++(vdso_data->tb_update_count); | |
4a4cfe38 TB |
961 | } |
962 | ||
963 | void update_vsyscall_tz(void) | |
964 | { | |
4a4cfe38 TB |
965 | vdso_data->tz_minuteswest = sys_tz.tz_minuteswest; |
966 | vdso_data->tz_dsttime = sys_tz.tz_dsttime; | |
4a4cfe38 TB |
967 | } |
968 | ||
1c21a293 | 969 | static void __init clocksource_init(void) |
4a4cfe38 TB |
970 | { |
971 | struct clocksource *clock; | |
972 | ||
973 | if (__USE_RTC()) | |
974 | clock = &clocksource_rtc; | |
975 | else | |
976 | clock = &clocksource_timebase; | |
977 | ||
11b8633a | 978 | if (clocksource_register_hz(clock, tb_ticks_per_sec)) { |
4a4cfe38 TB |
979 | printk(KERN_ERR "clocksource: %s is already registered\n", |
980 | clock->name); | |
981 | return; | |
982 | } | |
983 | ||
984 | printk(KERN_INFO "clocksource: %s mult[%x] shift[%d] registered\n", | |
985 | clock->name, clock->mult, clock->shift); | |
986 | } | |
987 | ||
d831d0b8 TB |
988 | static int decrementer_set_next_event(unsigned long evt, |
989 | struct clock_event_device *dev) | |
990 | { | |
69111bac | 991 | __this_cpu_write(decrementers_next_tb, get_tb_or_rtc() + evt); |
d831d0b8 | 992 | set_dec(evt); |
0215f7d8 BH |
993 | |
994 | /* We may have raced with new irq work */ | |
995 | if (test_irq_work_pending()) | |
996 | set_dec(1); | |
997 | ||
d831d0b8 TB |
998 | return 0; |
999 | } | |
1000 | ||
37a13e78 | 1001 | static int decrementer_shutdown(struct clock_event_device *dev) |
d831d0b8 | 1002 | { |
79901024 | 1003 | decrementer_set_next_event(decrementer_max, dev); |
37a13e78 | 1004 | return 0; |
d831d0b8 TB |
1005 | } |
1006 | ||
1007 | static void register_decrementer_clockevent(int cpu) | |
1008 | { | |
7df10275 | 1009 | struct clock_event_device *dec = &per_cpu(decrementers, cpu); |
d831d0b8 TB |
1010 | |
1011 | *dec = decrementer_clockevent; | |
320ab2b0 | 1012 | dec->cpumask = cpumask_of(cpu); |
d831d0b8 | 1013 | |
8b78fdb0 AB |
1014 | clockevents_config_and_register(dec, ppc_tb_freq, 2, decrementer_max); |
1015 | ||
b919ee82 AB |
1016 | printk_once(KERN_DEBUG "clockevent: %s mult[%x] shift[%d] cpu[%d]\n", |
1017 | dec->name, dec->mult, dec->shift, cpu); | |
b4d16ab5 ME |
1018 | |
1019 | /* Set values for KVM, see kvm_emulate_dec() */ | |
1020 | decrementer_clockevent.mult = dec->mult; | |
1021 | decrementer_clockevent.shift = dec->shift; | |
d831d0b8 TB |
1022 | } |
1023 | ||
79901024 OH |
1024 | static void enable_large_decrementer(void) |
1025 | { | |
1026 | if (!cpu_has_feature(CPU_FTR_ARCH_300)) | |
1027 | return; | |
1028 | ||
1029 | if (decrementer_max <= DECREMENTER_DEFAULT_MAX) | |
1030 | return; | |
1031 | ||
1032 | /* | |
1033 | * If we're running as the hypervisor we need to enable the LD manually | |
1034 | * otherwise firmware should have done it for us. | |
1035 | */ | |
1036 | if (cpu_has_feature(CPU_FTR_HVMODE)) | |
1037 | mtspr(SPRN_LPCR, mfspr(SPRN_LPCR) | LPCR_LD); | |
1038 | } | |
1039 | ||
1040 | static void __init set_decrementer_max(void) | |
1041 | { | |
1042 | struct device_node *cpu; | |
1043 | u32 bits = 32; | |
1044 | ||
1045 | /* Prior to ISAv3 the decrementer is always 32 bit */ | |
1046 | if (!cpu_has_feature(CPU_FTR_ARCH_300)) | |
1047 | return; | |
1048 | ||
1049 | cpu = of_find_node_by_type(NULL, "cpu"); | |
1050 | ||
1051 | if (of_property_read_u32(cpu, "ibm,dec-bits", &bits) == 0) { | |
1052 | if (bits > 64 || bits < 32) { | |
1053 | pr_warn("time_init: firmware supplied invalid ibm,dec-bits"); | |
1054 | bits = 32; | |
1055 | } | |
1056 | ||
1057 | /* calculate the signed maximum given this many bits */ | |
1058 | decrementer_max = (1ul << (bits - 1)) - 1; | |
1059 | } | |
1060 | ||
1061 | of_node_put(cpu); | |
1062 | ||
1063 | pr_info("time_init: %u bit decrementer (max: %llx)\n", | |
1064 | bits, decrementer_max); | |
1065 | } | |
1066 | ||
c481887f | 1067 | static void __init init_decrementer_clockevent(void) |
d831d0b8 | 1068 | { |
8b78fdb0 | 1069 | register_decrementer_clockevent(smp_processor_id()); |
d831d0b8 TB |
1070 | } |
1071 | ||
1072 | void secondary_cpu_time_init(void) | |
1073 | { | |
79901024 OH |
1074 | /* Enable and test the large decrementer for this cpu */ |
1075 | enable_large_decrementer(); | |
1076 | ||
77c0a700 BH |
1077 | /* Start the decrementer on CPUs that have manual control |
1078 | * such as BookE | |
1079 | */ | |
1080 | start_cpu_decrementer(); | |
1081 | ||
d831d0b8 TB |
1082 | /* FIME: Should make unrelatred change to move snapshot_timebase |
1083 | * call here ! */ | |
1084 | register_decrementer_clockevent(smp_processor_id()); | |
1085 | } | |
1086 | ||
f2783c15 | 1087 | /* This function is only called on the boot processor */ |
1da177e4 LT |
1088 | void __init time_init(void) |
1089 | { | |
1da177e4 | 1090 | struct div_result res; |
d75d68cf | 1091 | u64 scale; |
f2783c15 PM |
1092 | unsigned shift; |
1093 | ||
96c44507 PM |
1094 | if (__USE_RTC()) { |
1095 | /* 601 processor: dec counts down by 128 every 128ns */ | |
1096 | ppc_tb_freq = 1000000000; | |
96c44507 PM |
1097 | } else { |
1098 | /* Normal PowerPC with timebase register */ | |
1099 | ppc_md.calibrate_decr(); | |
224ad80a | 1100 | printk(KERN_DEBUG "time_init: decrementer frequency = %lu.%.6lu MHz\n", |
96c44507 | 1101 | ppc_tb_freq / 1000000, ppc_tb_freq % 1000000); |
224ad80a | 1102 | printk(KERN_DEBUG "time_init: processor frequency = %lu.%.6lu MHz\n", |
96c44507 | 1103 | ppc_proc_freq / 1000000, ppc_proc_freq % 1000000); |
96c44507 | 1104 | } |
374e99d4 PM |
1105 | |
1106 | tb_ticks_per_jiffy = ppc_tb_freq / HZ; | |
092b8f34 | 1107 | tb_ticks_per_sec = ppc_tb_freq; |
374e99d4 | 1108 | tb_ticks_per_usec = ppc_tb_freq / 1000000; |
c6622f63 | 1109 | calc_cputime_factors(); |
092b8f34 | 1110 | |
1da177e4 LT |
1111 | /* |
1112 | * Compute scale factor for sched_clock. | |
1113 | * The calibrate_decr() function has set tb_ticks_per_sec, | |
1114 | * which is the timebase frequency. | |
1115 | * We compute 1e9 * 2^64 / tb_ticks_per_sec and interpret | |
1116 | * the 128-bit result as a 64.64 fixed-point number. | |
1117 | * We then shift that number right until it is less than 1.0, | |
1118 | * giving us the scale factor and shift count to use in | |
1119 | * sched_clock(). | |
1120 | */ | |
1121 | div128_by_32(1000000000, 0, tb_ticks_per_sec, &res); | |
1122 | scale = res.result_low; | |
1123 | for (shift = 0; res.result_high != 0; ++shift) { | |
1124 | scale = (scale >> 1) | (res.result_high << 63); | |
1125 | res.result_high >>= 1; | |
1126 | } | |
1127 | tb_to_ns_scale = scale; | |
1128 | tb_to_ns_shift = shift; | |
fc9069fe | 1129 | /* Save the current timebase to pretty up CONFIG_PRINTK_TIME */ |
c27da339 | 1130 | boot_tb = get_tb_or_rtc(); |
1da177e4 | 1131 | |
092b8f34 | 1132 | /* If platform provided a timezone (pmac), we correct the time */ |
621692cb | 1133 | if (timezone_offset) { |
092b8f34 PM |
1134 | sys_tz.tz_minuteswest = -timezone_offset / 60; |
1135 | sys_tz.tz_dsttime = 0; | |
621692cb | 1136 | } |
092b8f34 | 1137 | |
a7f290da BH |
1138 | vdso_data->tb_update_count = 0; |
1139 | vdso_data->tb_ticks_per_sec = tb_ticks_per_sec; | |
1da177e4 | 1140 | |
79901024 OH |
1141 | /* initialise and enable the large decrementer (if we have one) */ |
1142 | set_decrementer_max(); | |
1143 | enable_large_decrementer(); | |
1144 | ||
77c0a700 BH |
1145 | /* Start the decrementer on CPUs that have manual control |
1146 | * such as BookE | |
1147 | */ | |
1148 | start_cpu_decrementer(); | |
1149 | ||
f5339277 SR |
1150 | /* Register the clocksource */ |
1151 | clocksource_init(); | |
4a4cfe38 | 1152 | |
d831d0b8 | 1153 | init_decrementer_clockevent(); |
0d948730 | 1154 | tick_setup_hrtimer_broadcast(); |
f0d37300 KH |
1155 | |
1156 | #ifdef CONFIG_COMMON_CLK | |
1157 | of_clk_init(NULL); | |
1158 | #endif | |
1da177e4 LT |
1159 | } |
1160 | ||
1da177e4 LT |
1161 | /* |
1162 | * Divide a 128-bit dividend by a 32-bit divisor, leaving a 128 bit | |
1163 | * result. | |
1164 | */ | |
f2783c15 PM |
1165 | void div128_by_32(u64 dividend_high, u64 dividend_low, |
1166 | unsigned divisor, struct div_result *dr) | |
1da177e4 | 1167 | { |
f2783c15 PM |
1168 | unsigned long a, b, c, d; |
1169 | unsigned long w, x, y, z; | |
1170 | u64 ra, rb, rc; | |
1da177e4 LT |
1171 | |
1172 | a = dividend_high >> 32; | |
1173 | b = dividend_high & 0xffffffff; | |
1174 | c = dividend_low >> 32; | |
1175 | d = dividend_low & 0xffffffff; | |
1176 | ||
f2783c15 PM |
1177 | w = a / divisor; |
1178 | ra = ((u64)(a - (w * divisor)) << 32) + b; | |
1179 | ||
f2783c15 PM |
1180 | rb = ((u64) do_div(ra, divisor) << 32) + c; |
1181 | x = ra; | |
1da177e4 | 1182 | |
f2783c15 PM |
1183 | rc = ((u64) do_div(rb, divisor) << 32) + d; |
1184 | y = rb; | |
1185 | ||
1186 | do_div(rc, divisor); | |
1187 | z = rc; | |
1da177e4 | 1188 | |
f2783c15 PM |
1189 | dr->result_high = ((u64)w << 32) + x; |
1190 | dr->result_low = ((u64)y << 32) + z; | |
1da177e4 LT |
1191 | |
1192 | } | |
bcd68a70 | 1193 | |
177996e6 BH |
1194 | /* We don't need to calibrate delay, we use the CPU timebase for that */ |
1195 | void calibrate_delay(void) | |
1196 | { | |
1197 | /* Some generic code (such as spinlock debug) use loops_per_jiffy | |
1198 | * as the number of __delay(1) in a jiffy, so make it so | |
1199 | */ | |
1200 | loops_per_jiffy = tb_ticks_per_jiffy; | |
1201 | } | |
1202 | ||
169047f4 AB |
1203 | #if IS_ENABLED(CONFIG_RTC_DRV_GENERIC) |
1204 | static int rtc_generic_get_time(struct device *dev, struct rtc_time *tm) | |
1205 | { | |
1206 | ppc_md.get_rtc_time(tm); | |
890ae797 | 1207 | return 0; |
169047f4 AB |
1208 | } |
1209 | ||
1210 | static int rtc_generic_set_time(struct device *dev, struct rtc_time *tm) | |
1211 | { | |
1212 | if (!ppc_md.set_rtc_time) | |
1213 | return -EOPNOTSUPP; | |
1214 | ||
1215 | if (ppc_md.set_rtc_time(tm) < 0) | |
1216 | return -EOPNOTSUPP; | |
1217 | ||
1218 | return 0; | |
1219 | } | |
1220 | ||
1221 | static const struct rtc_class_ops rtc_generic_ops = { | |
1222 | .read_time = rtc_generic_get_time, | |
1223 | .set_time = rtc_generic_set_time, | |
1224 | }; | |
1225 | ||
bcd68a70 GU |
1226 | static int __init rtc_init(void) |
1227 | { | |
1228 | struct platform_device *pdev; | |
1229 | ||
1230 | if (!ppc_md.get_rtc_time) | |
1231 | return -ENODEV; | |
1232 | ||
169047f4 AB |
1233 | pdev = platform_device_register_data(NULL, "rtc-generic", -1, |
1234 | &rtc_generic_ops, | |
1235 | sizeof(rtc_generic_ops)); | |
bcd68a70 | 1236 | |
8c6ffba0 | 1237 | return PTR_ERR_OR_ZERO(pdev); |
bcd68a70 GU |
1238 | } |
1239 | ||
8f6b9512 | 1240 | device_initcall(rtc_init); |
169047f4 | 1241 | #endif |