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, | |
114 | .set_state_shutdown = decrementer_shutdown, | |
115 | .tick_resume = decrementer_shutdown, | |
116 | .features = CLOCK_EVT_FEAT_ONESHOT | | |
117 | CLOCK_EVT_FEAT_C3STOP, | |
d831d0b8 | 118 | }; |
6e35994d | 119 | EXPORT_SYMBOL(decrementer_clockevent); |
d831d0b8 | 120 | |
7df10275 AB |
121 | DEFINE_PER_CPU(u64, decrementers_next_tb); |
122 | static DEFINE_PER_CPU(struct clock_event_device, decrementers); | |
d831d0b8 | 123 | |
1da177e4 LT |
124 | #define XSEC_PER_SEC (1024*1024) |
125 | ||
f2783c15 PM |
126 | #ifdef CONFIG_PPC64 |
127 | #define SCALE_XSEC(xsec, max) (((xsec) * max) / XSEC_PER_SEC) | |
128 | #else | |
129 | /* compute ((xsec << 12) * max) >> 32 */ | |
130 | #define SCALE_XSEC(xsec, max) mulhwu((xsec) << 12, max) | |
131 | #endif | |
132 | ||
1da177e4 LT |
133 | unsigned long tb_ticks_per_jiffy; |
134 | unsigned long tb_ticks_per_usec = 100; /* sane default */ | |
135 | EXPORT_SYMBOL(tb_ticks_per_usec); | |
136 | unsigned long tb_ticks_per_sec; | |
2cf82c02 | 137 | EXPORT_SYMBOL(tb_ticks_per_sec); /* for cputime_t conversions */ |
092b8f34 | 138 | |
1da177e4 | 139 | DEFINE_SPINLOCK(rtc_lock); |
6ae3db11 | 140 | EXPORT_SYMBOL_GPL(rtc_lock); |
1da177e4 | 141 | |
fc9069fe TB |
142 | static u64 tb_to_ns_scale __read_mostly; |
143 | static unsigned tb_to_ns_shift __read_mostly; | |
364a1246 | 144 | static u64 boot_tb __read_mostly; |
1da177e4 | 145 | |
1da177e4 | 146 | extern struct timezone sys_tz; |
f2783c15 | 147 | static long timezone_offset; |
1da177e4 | 148 | |
10f7e7c1 | 149 | unsigned long ppc_proc_freq; |
55ec2fca | 150 | EXPORT_SYMBOL_GPL(ppc_proc_freq); |
10f7e7c1 | 151 | unsigned long ppc_tb_freq; |
55ec2fca | 152 | EXPORT_SYMBOL_GPL(ppc_tb_freq); |
96c44507 | 153 | |
abf917cd | 154 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
c6622f63 | 155 | /* |
e7f340ca FW |
156 | * Factor for converting from cputime_t (timebase ticks) to |
157 | * microseconds. This is stored as 0.64 fixed-point binary fraction. | |
c6622f63 | 158 | */ |
9f5072d4 AS |
159 | u64 __cputime_usec_factor; |
160 | EXPORT_SYMBOL(__cputime_usec_factor); | |
a42548a1 | 161 | |
c223c903 | 162 | #ifdef CONFIG_PPC_SPLPAR |
872e439a | 163 | void (*dtl_consumer)(struct dtl_entry *, u64); |
c223c903 CL |
164 | #endif |
165 | ||
166 | #ifdef CONFIG_PPC64 | |
167 | #define get_accounting(tsk) (&get_paca()->accounting) | |
168 | #else | |
169 | #define get_accounting(tsk) (&task_thread_info(tsk)->accounting) | |
170 | #endif | |
872e439a | 171 | |
c6622f63 PM |
172 | static void calc_cputime_factors(void) |
173 | { | |
174 | struct div_result res; | |
175 | ||
9f5072d4 AS |
176 | div128_by_32(1000000, 0, tb_ticks_per_sec, &res); |
177 | __cputime_usec_factor = res.result_low; | |
c6622f63 PM |
178 | } |
179 | ||
180 | /* | |
cf9efce0 PM |
181 | * Read the SPURR on systems that have it, otherwise the PURR, |
182 | * or if that doesn't exist return the timebase value passed in. | |
c6622f63 | 183 | */ |
c223c903 | 184 | static unsigned long read_spurr(unsigned long tb) |
c6622f63 | 185 | { |
cf9efce0 PM |
186 | if (cpu_has_feature(CPU_FTR_SPURR)) |
187 | return mfspr(SPRN_SPURR); | |
c6622f63 PM |
188 | if (cpu_has_feature(CPU_FTR_PURR)) |
189 | return mfspr(SPRN_PURR); | |
cf9efce0 | 190 | return tb; |
c6622f63 PM |
191 | } |
192 | ||
cf9efce0 PM |
193 | #ifdef CONFIG_PPC_SPLPAR |
194 | ||
4603ac18 | 195 | /* |
cf9efce0 PM |
196 | * Scan the dispatch trace log and count up the stolen time. |
197 | * Should be called with interrupts disabled. | |
4603ac18 | 198 | */ |
cf9efce0 | 199 | static u64 scan_dispatch_log(u64 stop_tb) |
4603ac18 | 200 | { |
872e439a | 201 | u64 i = local_paca->dtl_ridx; |
cf9efce0 PM |
202 | struct dtl_entry *dtl = local_paca->dtl_curr; |
203 | struct dtl_entry *dtl_end = local_paca->dispatch_log_end; | |
204 | struct lppaca *vpa = local_paca->lppaca_ptr; | |
205 | u64 tb_delta; | |
206 | u64 stolen = 0; | |
207 | u64 dtb; | |
208 | ||
84ffae55 AB |
209 | if (!dtl) |
210 | return 0; | |
211 | ||
7ffcf8ec | 212 | if (i == be64_to_cpu(vpa->dtl_idx)) |
cf9efce0 | 213 | return 0; |
7ffcf8ec | 214 | while (i < be64_to_cpu(vpa->dtl_idx)) { |
7ffcf8ec AB |
215 | dtb = be64_to_cpu(dtl->timebase); |
216 | tb_delta = be32_to_cpu(dtl->enqueue_to_dispatch_time) + | |
217 | be32_to_cpu(dtl->ready_to_enqueue_time); | |
cf9efce0 | 218 | barrier(); |
7ffcf8ec | 219 | if (i + N_DISPATCH_LOG < be64_to_cpu(vpa->dtl_idx)) { |
cf9efce0 | 220 | /* buffer has overflowed */ |
7ffcf8ec | 221 | i = be64_to_cpu(vpa->dtl_idx) - N_DISPATCH_LOG; |
cf9efce0 PM |
222 | dtl = local_paca->dispatch_log + (i % N_DISPATCH_LOG); |
223 | continue; | |
224 | } | |
225 | if (dtb > stop_tb) | |
226 | break; | |
84b07386 AB |
227 | if (dtl_consumer) |
228 | dtl_consumer(dtl, i); | |
cf9efce0 PM |
229 | stolen += tb_delta; |
230 | ++i; | |
231 | ++dtl; | |
232 | if (dtl == dtl_end) | |
233 | dtl = local_paca->dispatch_log; | |
234 | } | |
235 | local_paca->dtl_ridx = i; | |
236 | local_paca->dtl_curr = dtl; | |
237 | return stolen; | |
4603ac18 MN |
238 | } |
239 | ||
cf9efce0 PM |
240 | /* |
241 | * Accumulate stolen time by scanning the dispatch trace log. | |
242 | * Called on entry from user mode. | |
243 | */ | |
244 | void accumulate_stolen_time(void) | |
245 | { | |
246 | u64 sst, ust; | |
4e26bc4a | 247 | unsigned long save_irq_soft_mask = irq_soft_mask_return(); |
c223c903 | 248 | struct cpu_accounting_data *acct = &local_paca->accounting; |
b18ae08d TH |
249 | |
250 | /* We are called early in the exception entry, before | |
251 | * soft/hard_enabled are sync'ed to the expected state | |
252 | * for the exception. We are hard disabled but the PACA | |
253 | * needs to reflect that so various debug stuff doesn't | |
254 | * complain | |
255 | */ | |
4e26bc4a | 256 | irq_soft_mask_set(IRQS_DISABLED); |
b18ae08d | 257 | |
c223c903 CL |
258 | sst = scan_dispatch_log(acct->starttime_user); |
259 | ust = scan_dispatch_log(acct->starttime); | |
8c8b73c4 FW |
260 | acct->stime -= sst; |
261 | acct->utime -= ust; | |
f828c3d0 | 262 | acct->steal_time += ust + sst; |
b18ae08d | 263 | |
4e26bc4a | 264 | irq_soft_mask_set(save_irq_soft_mask); |
cf9efce0 PM |
265 | } |
266 | ||
267 | static inline u64 calculate_stolen_time(u64 stop_tb) | |
268 | { | |
a6201da3 AK |
269 | if (!firmware_has_feature(FW_FEATURE_SPLPAR)) |
270 | return 0; | |
271 | ||
a19ff1a2 FW |
272 | if (get_paca()->dtl_ridx != be64_to_cpu(get_lppaca()->dtl_idx)) |
273 | return scan_dispatch_log(stop_tb); | |
cf9efce0 | 274 | |
a19ff1a2 | 275 | return 0; |
4603ac18 MN |
276 | } |
277 | ||
cf9efce0 PM |
278 | #else /* CONFIG_PPC_SPLPAR */ |
279 | static inline u64 calculate_stolen_time(u64 stop_tb) | |
280 | { | |
281 | return 0; | |
282 | } | |
283 | ||
284 | #endif /* CONFIG_PPC_SPLPAR */ | |
285 | ||
c6622f63 PM |
286 | /* |
287 | * Account time for a transition between system, hard irq | |
288 | * or soft irq state. | |
289 | */ | |
c223c903 | 290 | static unsigned long vtime_delta(struct task_struct *tsk, |
a19ff1a2 FW |
291 | unsigned long *stime_scaled, |
292 | unsigned long *steal_time) | |
c6622f63 | 293 | { |
c223c903 | 294 | unsigned long now, nowscaled, deltascaled; |
a19ff1a2 FW |
295 | unsigned long stime; |
296 | unsigned long utime, utime_scaled; | |
c223c903 | 297 | struct cpu_accounting_data *acct = get_accounting(tsk); |
c6622f63 | 298 | |
1b2852b1 FW |
299 | WARN_ON_ONCE(!irqs_disabled()); |
300 | ||
cf9efce0 | 301 | now = mftb(); |
4603ac18 | 302 | nowscaled = read_spurr(now); |
a19ff1a2 | 303 | stime = now - acct->starttime; |
c223c903 CL |
304 | acct->starttime = now; |
305 | deltascaled = nowscaled - acct->startspurr; | |
306 | acct->startspurr = nowscaled; | |
cf9efce0 | 307 | |
a19ff1a2 | 308 | *steal_time = calculate_stolen_time(now); |
cf9efce0 | 309 | |
a19ff1a2 | 310 | utime = acct->utime - acct->utime_sspurr; |
8c8b73c4 | 311 | acct->utime_sspurr = acct->utime; |
cf9efce0 PM |
312 | |
313 | /* | |
314 | * Because we don't read the SPURR on every kernel entry/exit, | |
315 | * deltascaled includes both user and system SPURR ticks. | |
316 | * Apportion these ticks to system SPURR ticks and user | |
317 | * SPURR ticks in the same ratio as the system time (delta) | |
318 | * and user time (udelta) values obtained from the timebase | |
319 | * over the same interval. The system ticks get accounted here; | |
320 | * the user ticks get saved up in paca->user_time_scaled to be | |
321 | * used by account_process_tick. | |
322 | */ | |
a19ff1a2 FW |
323 | *stime_scaled = stime; |
324 | utime_scaled = utime; | |
325 | if (deltascaled != stime + utime) { | |
326 | if (utime) { | |
327 | *stime_scaled = deltascaled * stime / (stime + utime); | |
328 | utime_scaled = deltascaled - *stime_scaled; | |
cf9efce0 | 329 | } else { |
a19ff1a2 | 330 | *stime_scaled = deltascaled; |
cf9efce0 PM |
331 | } |
332 | } | |
a19ff1a2 | 333 | acct->utime_scaled += utime_scaled; |
cf9efce0 | 334 | |
a19ff1a2 | 335 | return stime; |
a7e1a9e3 FW |
336 | } |
337 | ||
fd25b4c2 | 338 | void vtime_account_system(struct task_struct *tsk) |
a7e1a9e3 | 339 | { |
a19ff1a2 FW |
340 | unsigned long stime, stime_scaled, steal_time; |
341 | struct cpu_accounting_data *acct = get_accounting(tsk); | |
342 | ||
343 | stime = vtime_delta(tsk, &stime_scaled, &steal_time); | |
344 | ||
345 | stime -= min(stime, steal_time); | |
346 | acct->steal_time += steal_time; | |
a7e1a9e3 | 347 | |
a19ff1a2 FW |
348 | if ((tsk->flags & PF_VCPU) && !irq_count()) { |
349 | acct->gtime += stime; | |
350 | acct->utime_scaled += stime_scaled; | |
351 | } else { | |
352 | if (hardirq_count()) | |
353 | acct->hardirq_time += stime; | |
354 | else if (in_serving_softirq()) | |
355 | acct->softirq_time += stime; | |
356 | else | |
357 | acct->stime += stime; | |
358 | ||
359 | acct->stime_scaled += stime_scaled; | |
360 | } | |
a7e1a9e3 | 361 | } |
c11f11fc | 362 | EXPORT_SYMBOL_GPL(vtime_account_system); |
a7e1a9e3 | 363 | |
fd25b4c2 | 364 | void vtime_account_idle(struct task_struct *tsk) |
a7e1a9e3 | 365 | { |
a19ff1a2 FW |
366 | unsigned long stime, stime_scaled, steal_time; |
367 | struct cpu_accounting_data *acct = get_accounting(tsk); | |
a7e1a9e3 | 368 | |
a19ff1a2 FW |
369 | stime = vtime_delta(tsk, &stime_scaled, &steal_time); |
370 | acct->idle_time += stime + steal_time; | |
c6622f63 PM |
371 | } |
372 | ||
373 | /* | |
c8d7dabf | 374 | * Account the whole cputime accumulated in the paca |
c6622f63 | 375 | * Must be called with interrupts disabled. |
bcebdf84 FW |
376 | * Assumes that vtime_account_system/idle() has been called |
377 | * recently (i.e. since the last entry from usermode) so that | |
cf9efce0 | 378 | * get_paca()->user_time_scaled is up to date. |
c6622f63 | 379 | */ |
c8d7dabf | 380 | void vtime_flush(struct task_struct *tsk) |
c6622f63 | 381 | { |
c223c903 | 382 | struct cpu_accounting_data *acct = get_accounting(tsk); |
c6622f63 | 383 | |
a19ff1a2 | 384 | if (acct->utime) |
23244a5c | 385 | account_user_time(tsk, cputime_to_nsecs(acct->utime)); |
a19ff1a2 FW |
386 | |
387 | if (acct->utime_scaled) | |
5613fda9 | 388 | tsk->utimescaled += cputime_to_nsecs(acct->utime_scaled); |
a19ff1a2 FW |
389 | |
390 | if (acct->gtime) | |
fb8b049c | 391 | account_guest_time(tsk, cputime_to_nsecs(acct->gtime)); |
a19ff1a2 FW |
392 | |
393 | if (acct->steal_time) | |
be9095ed | 394 | account_steal_time(cputime_to_nsecs(acct->steal_time)); |
a19ff1a2 FW |
395 | |
396 | if (acct->idle_time) | |
18b43a9b | 397 | account_idle_time(cputime_to_nsecs(acct->idle_time)); |
a19ff1a2 FW |
398 | |
399 | if (acct->stime) | |
fb8b049c FW |
400 | account_system_index_time(tsk, cputime_to_nsecs(acct->stime), |
401 | CPUTIME_SYSTEM); | |
a19ff1a2 | 402 | if (acct->stime_scaled) |
5613fda9 | 403 | tsk->stimescaled += cputime_to_nsecs(acct->stime_scaled); |
a19ff1a2 FW |
404 | |
405 | if (acct->hardirq_time) | |
fb8b049c FW |
406 | account_system_index_time(tsk, cputime_to_nsecs(acct->hardirq_time), |
407 | CPUTIME_IRQ); | |
a19ff1a2 | 408 | if (acct->softirq_time) |
fb8b049c FW |
409 | account_system_index_time(tsk, cputime_to_nsecs(acct->softirq_time), |
410 | CPUTIME_SOFTIRQ); | |
a19ff1a2 | 411 | |
8c8b73c4 FW |
412 | acct->utime = 0; |
413 | acct->utime_scaled = 0; | |
c223c903 | 414 | acct->utime_sspurr = 0; |
a19ff1a2 FW |
415 | acct->gtime = 0; |
416 | acct->steal_time = 0; | |
417 | acct->idle_time = 0; | |
418 | acct->stime = 0; | |
419 | acct->stime_scaled = 0; | |
420 | acct->hardirq_time = 0; | |
421 | acct->softirq_time = 0; | |
c6622f63 PM |
422 | } |
423 | ||
c223c903 CL |
424 | #ifdef CONFIG_PPC32 |
425 | /* | |
426 | * Called from the context switch with interrupts disabled, to charge all | |
427 | * accumulated times to the current process, and to prepare accounting on | |
428 | * the next process. | |
429 | */ | |
430 | void arch_vtime_task_switch(struct task_struct *prev) | |
431 | { | |
432 | struct cpu_accounting_data *acct = get_accounting(current); | |
433 | ||
434 | acct->starttime = get_accounting(prev)->starttime; | |
90d08ba2 | 435 | acct->startspurr = get_accounting(prev)->startspurr; |
c223c903 CL |
436 | } |
437 | #endif /* CONFIG_PPC32 */ | |
438 | ||
abf917cd | 439 | #else /* ! CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */ |
c6622f63 | 440 | #define calc_cputime_factors() |
c6622f63 PM |
441 | #endif |
442 | ||
6defa38b PM |
443 | void __delay(unsigned long loops) |
444 | { | |
445 | unsigned long start; | |
446 | int diff; | |
447 | ||
4e287e65 | 448 | spin_begin(); |
6defa38b PM |
449 | if (__USE_RTC()) { |
450 | start = get_rtcl(); | |
451 | do { | |
452 | /* the RTCL register wraps at 1000000000 */ | |
453 | diff = get_rtcl() - start; | |
454 | if (diff < 0) | |
455 | diff += 1000000000; | |
4e287e65 | 456 | spin_cpu_relax(); |
6defa38b PM |
457 | } while (diff < loops); |
458 | } else { | |
459 | start = get_tbl(); | |
460 | while (get_tbl() - start < loops) | |
4e287e65 | 461 | spin_cpu_relax(); |
6defa38b | 462 | } |
4e287e65 | 463 | spin_end(); |
6defa38b PM |
464 | } |
465 | EXPORT_SYMBOL(__delay); | |
466 | ||
467 | void udelay(unsigned long usecs) | |
468 | { | |
469 | __delay(tb_ticks_per_usec * usecs); | |
470 | } | |
471 | EXPORT_SYMBOL(udelay); | |
472 | ||
1da177e4 LT |
473 | #ifdef CONFIG_SMP |
474 | unsigned long profile_pc(struct pt_regs *regs) | |
475 | { | |
476 | unsigned long pc = instruction_pointer(regs); | |
477 | ||
478 | if (in_lock_functions(pc)) | |
479 | return regs->link; | |
480 | ||
481 | return pc; | |
482 | } | |
483 | EXPORT_SYMBOL(profile_pc); | |
484 | #endif | |
485 | ||
e360adbe | 486 | #ifdef CONFIG_IRQ_WORK |
105988c0 | 487 | |
0fe1ac48 PM |
488 | /* |
489 | * 64-bit uses a byte in the PACA, 32-bit uses a per-cpu variable... | |
490 | */ | |
491 | #ifdef CONFIG_PPC64 | |
e360adbe | 492 | static inline unsigned long test_irq_work_pending(void) |
105988c0 | 493 | { |
0fe1ac48 PM |
494 | unsigned long x; |
495 | ||
496 | asm volatile("lbz %0,%1(13)" | |
497 | : "=r" (x) | |
e360adbe | 498 | : "i" (offsetof(struct paca_struct, irq_work_pending))); |
0fe1ac48 PM |
499 | return x; |
500 | } | |
501 | ||
e360adbe | 502 | static inline void set_irq_work_pending_flag(void) |
0fe1ac48 PM |
503 | { |
504 | asm volatile("stb %0,%1(13)" : : | |
505 | "r" (1), | |
e360adbe | 506 | "i" (offsetof(struct paca_struct, irq_work_pending))); |
0fe1ac48 PM |
507 | } |
508 | ||
e360adbe | 509 | static inline void clear_irq_work_pending(void) |
0fe1ac48 PM |
510 | { |
511 | asm volatile("stb %0,%1(13)" : : | |
512 | "r" (0), | |
e360adbe | 513 | "i" (offsetof(struct paca_struct, irq_work_pending))); |
105988c0 PM |
514 | } |
515 | ||
ebb37cf3 NP |
516 | void arch_irq_work_raise(void) |
517 | { | |
518 | preempt_disable(); | |
519 | set_irq_work_pending_flag(); | |
520 | /* | |
521 | * Non-nmi code running with interrupts disabled will replay | |
522 | * irq_happened before it re-enables interrupts, so setthe | |
523 | * decrementer there instead of causing a hardware exception | |
524 | * which would immediately hit the masked interrupt handler | |
525 | * and have the net effect of setting the decrementer in | |
526 | * irq_happened. | |
527 | * | |
528 | * NMI interrupts can not check this when they return, so the | |
529 | * decrementer hardware exception is raised, which will fire | |
530 | * when interrupts are next enabled. | |
531 | * | |
532 | * BookE does not support this yet, it must audit all NMI | |
533 | * interrupt handlers to ensure they call nmi_enter() so this | |
534 | * check would be correct. | |
535 | */ | |
536 | if (IS_ENABLED(CONFIG_BOOKE) || !irqs_disabled() || in_nmi()) { | |
537 | set_dec(1); | |
538 | } else { | |
539 | hard_irq_disable(); | |
540 | local_paca->irq_happened |= PACA_IRQ_DEC; | |
541 | } | |
542 | preempt_enable(); | |
543 | } | |
544 | ||
0fe1ac48 PM |
545 | #else /* 32-bit */ |
546 | ||
e360adbe | 547 | DEFINE_PER_CPU(u8, irq_work_pending); |
0fe1ac48 | 548 | |
69111bac CL |
549 | #define set_irq_work_pending_flag() __this_cpu_write(irq_work_pending, 1) |
550 | #define test_irq_work_pending() __this_cpu_read(irq_work_pending) | |
551 | #define clear_irq_work_pending() __this_cpu_write(irq_work_pending, 0) | |
105988c0 | 552 | |
4f8b50bb | 553 | void arch_irq_work_raise(void) |
0fe1ac48 PM |
554 | { |
555 | preempt_disable(); | |
e360adbe | 556 | set_irq_work_pending_flag(); |
0fe1ac48 PM |
557 | set_dec(1); |
558 | preempt_enable(); | |
559 | } | |
560 | ||
ebb37cf3 NP |
561 | #endif /* 32 vs 64 bit */ |
562 | ||
e360adbe | 563 | #else /* CONFIG_IRQ_WORK */ |
105988c0 | 564 | |
e360adbe PZ |
565 | #define test_irq_work_pending() 0 |
566 | #define clear_irq_work_pending() | |
105988c0 | 567 | |
e360adbe | 568 | #endif /* CONFIG_IRQ_WORK */ |
105988c0 | 569 | |
1da177e4 LT |
570 | /* |
571 | * timer_interrupt - gets called when the decrementer overflows, | |
572 | * with interrupts disabled. | |
573 | */ | |
3f984620 | 574 | void timer_interrupt(struct pt_regs *regs) |
1da177e4 | 575 | { |
3f984620 | 576 | struct clock_event_device *evt = this_cpu_ptr(&decrementers); |
69111bac | 577 | u64 *next_tb = this_cpu_ptr(&decrementers_next_tb); |
3f984620 NP |
578 | struct pt_regs *old_regs; |
579 | u64 now; | |
d831d0b8 | 580 | |
963e5d3b | 581 | /* Some implementations of hotplug will get timer interrupts while |
689dfa89 TC |
582 | * offline, just ignore these and we also need to set |
583 | * decrementers_next_tb as MAX to make sure __check_irq_replay | |
584 | * don't replay timer interrupt when return, otherwise we'll trap | |
585 | * here infinitely :( | |
963e5d3b | 586 | */ |
a7cba02d | 587 | if (unlikely(!cpu_online(smp_processor_id()))) { |
689dfa89 | 588 | *next_tb = ~(u64)0; |
a7cba02d | 589 | set_dec(decrementer_max); |
963e5d3b | 590 | return; |
689dfa89 | 591 | } |
963e5d3b | 592 | |
a7cba02d NP |
593 | /* Ensure a positive value is written to the decrementer, or else |
594 | * some CPUs will continue to take decrementer exceptions. When the | |
595 | * PPC_WATCHDOG (decrementer based) is configured, keep this at most | |
596 | * 31 bits, which is about 4 seconds on most systems, which gives | |
597 | * the watchdog a chance of catching timer interrupt hard lockups. | |
598 | */ | |
599 | if (IS_ENABLED(CONFIG_PPC_WATCHDOG)) | |
600 | set_dec(0x7fffffff); | |
601 | else | |
602 | set_dec(decrementer_max); | |
603 | ||
7230c564 BH |
604 | /* Conditionally hard-enable interrupts now that the DEC has been |
605 | * bumped to its maximum value | |
606 | */ | |
607 | may_hard_irq_enable(); | |
608 | ||
89713ed1 | 609 | |
6e0fdf9a | 610 | #if defined(CONFIG_PPC32) && defined(CONFIG_PPC_PMAC) |
f2783c15 PM |
611 | if (atomic_read(&ppc_n_lost_interrupts) != 0) |
612 | do_IRQ(regs); | |
613 | #endif | |
1da177e4 | 614 | |
7d12e780 | 615 | old_regs = set_irq_regs(regs); |
1da177e4 | 616 | irq_enter(); |
3f984620 NP |
617 | trace_timer_interrupt_entry(regs); |
618 | ||
619 | if (test_irq_work_pending()) { | |
620 | clear_irq_work_pending(); | |
621 | irq_work_run(); | |
622 | } | |
623 | ||
624 | now = get_tb_or_rtc(); | |
625 | if (now >= *next_tb) { | |
626 | *next_tb = ~(u64)0; | |
627 | if (evt->event_handler) | |
628 | evt->event_handler(evt); | |
629 | __this_cpu_inc(irq_stat.timer_irqs_event); | |
630 | } else { | |
631 | now = *next_tb - now; | |
632 | if (now <= decrementer_max) | |
633 | set_dec(now); | |
634 | /* We may have raced with new irq work */ | |
635 | if (test_irq_work_pending()) | |
636 | set_dec(1); | |
637 | __this_cpu_inc(irq_stat.timer_irqs_others); | |
638 | } | |
1da177e4 | 639 | |
3f984620 | 640 | trace_timer_interrupt_exit(regs); |
1da177e4 | 641 | irq_exit(); |
7d12e780 | 642 | set_irq_regs(old_regs); |
1da177e4 | 643 | } |
9445aa1a | 644 | EXPORT_SYMBOL(timer_interrupt); |
1da177e4 | 645 | |
bc907113 | 646 | #ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST |
3f984620 NP |
647 | void timer_broadcast_interrupt(void) |
648 | { | |
649 | u64 *next_tb = this_cpu_ptr(&decrementers_next_tb); | |
650 | struct pt_regs *regs = get_irq_regs(); | |
651 | ||
652 | trace_timer_interrupt_entry(regs); | |
653 | *next_tb = ~(u64)0; | |
654 | tick_receive_broadcast(); | |
655 | __this_cpu_inc(irq_stat.timer_irqs_event); | |
656 | trace_timer_interrupt_exit(regs); | |
657 | } | |
bc907113 | 658 | #endif |
3f984620 | 659 | |
dabe859e PM |
660 | /* |
661 | * Hypervisor decrementer interrupts shouldn't occur but are sometimes | |
662 | * left pending on exit from a KVM guest. We don't need to do anything | |
663 | * to clear them, as they are edge-triggered. | |
664 | */ | |
665 | void hdec_interrupt(struct pt_regs *regs) | |
666 | { | |
667 | } | |
668 | ||
7ac5dde9 | 669 | #ifdef CONFIG_SUSPEND |
d75d68cf | 670 | static void generic_suspend_disable_irqs(void) |
7ac5dde9 | 671 | { |
7ac5dde9 SW |
672 | /* Disable the decrementer, so that it doesn't interfere |
673 | * with suspending. | |
674 | */ | |
675 | ||
79901024 | 676 | set_dec(decrementer_max); |
7ac5dde9 | 677 | local_irq_disable(); |
79901024 | 678 | set_dec(decrementer_max); |
7ac5dde9 SW |
679 | } |
680 | ||
d75d68cf | 681 | static void generic_suspend_enable_irqs(void) |
7ac5dde9 | 682 | { |
7ac5dde9 | 683 | local_irq_enable(); |
7ac5dde9 SW |
684 | } |
685 | ||
686 | /* Overrides the weak version in kernel/power/main.c */ | |
687 | void arch_suspend_disable_irqs(void) | |
688 | { | |
689 | if (ppc_md.suspend_disable_irqs) | |
690 | ppc_md.suspend_disable_irqs(); | |
691 | generic_suspend_disable_irqs(); | |
692 | } | |
693 | ||
694 | /* Overrides the weak version in kernel/power/main.c */ | |
695 | void arch_suspend_enable_irqs(void) | |
696 | { | |
697 | generic_suspend_enable_irqs(); | |
698 | if (ppc_md.suspend_enable_irqs) | |
699 | ppc_md.suspend_enable_irqs(); | |
700 | } | |
701 | #endif | |
702 | ||
b6c295df PM |
703 | unsigned long long tb_to_ns(unsigned long long ticks) |
704 | { | |
705 | return mulhdu(ticks, tb_to_ns_scale) << tb_to_ns_shift; | |
706 | } | |
707 | EXPORT_SYMBOL_GPL(tb_to_ns); | |
708 | ||
1da177e4 LT |
709 | /* |
710 | * Scheduler clock - returns current time in nanosec units. | |
711 | * | |
712 | * Note: mulhdu(a, b) (multiply high double unsigned) returns | |
713 | * the high 64 bits of a * b, i.e. (a * b) >> 64, where a and b | |
714 | * are 64-bit unsigned numbers. | |
715 | */ | |
6b847d79 | 716 | notrace unsigned long long sched_clock(void) |
1da177e4 | 717 | { |
96c44507 PM |
718 | if (__USE_RTC()) |
719 | return get_rtc(); | |
fc9069fe | 720 | return mulhdu(get_tb() - boot_tb, tb_to_ns_scale) << tb_to_ns_shift; |
1da177e4 LT |
721 | } |
722 | ||
4be1b297 CB |
723 | |
724 | #ifdef CONFIG_PPC_PSERIES | |
725 | ||
726 | /* | |
727 | * Running clock - attempts to give a view of time passing for a virtualised | |
728 | * kernels. | |
729 | * Uses the VTB register if available otherwise a next best guess. | |
730 | */ | |
731 | unsigned long long running_clock(void) | |
732 | { | |
733 | /* | |
734 | * Don't read the VTB as a host since KVM does not switch in host | |
735 | * timebase into the VTB when it takes a guest off the CPU, reading the | |
736 | * VTB would result in reading 'last switched out' guest VTB. | |
737 | * | |
738 | * Host kernels are often compiled with CONFIG_PPC_PSERIES checked, it | |
739 | * would be unsafe to rely only on the #ifdef above. | |
740 | */ | |
741 | if (firmware_has_feature(FW_FEATURE_LPAR) && | |
742 | cpu_has_feature(CPU_FTR_ARCH_207S)) | |
743 | return mulhdu(get_vtb() - boot_tb, tb_to_ns_scale) << tb_to_ns_shift; | |
744 | ||
745 | /* | |
746 | * This is a next best approximation without a VTB. | |
747 | * On a host which is running bare metal there should never be any stolen | |
748 | * time and on a host which doesn't do any virtualisation TB *should* equal | |
749 | * VTB so it makes no difference anyway. | |
750 | */ | |
9f3768e0 | 751 | return local_clock() - kcpustat_this_cpu->cpustat[CPUTIME_STEAL]; |
4be1b297 CB |
752 | } |
753 | #endif | |
754 | ||
0bb474a4 | 755 | static int __init get_freq(char *name, int cells, unsigned long *val) |
10f7e7c1 AB |
756 | { |
757 | struct device_node *cpu; | |
6f7aba7b | 758 | const __be32 *fp; |
0bb474a4 | 759 | int found = 0; |
10f7e7c1 | 760 | |
0bb474a4 | 761 | /* The cpu node should have timebase and clock frequency properties */ |
10f7e7c1 AB |
762 | cpu = of_find_node_by_type(NULL, "cpu"); |
763 | ||
d8a8188d | 764 | if (cpu) { |
e2eb6392 | 765 | fp = of_get_property(cpu, name, NULL); |
d8a8188d | 766 | if (fp) { |
0bb474a4 | 767 | found = 1; |
a4dc7ff0 | 768 | *val = of_read_ulong(fp, cells); |
10f7e7c1 | 769 | } |
0bb474a4 AB |
770 | |
771 | of_node_put(cpu); | |
10f7e7c1 | 772 | } |
0bb474a4 AB |
773 | |
774 | return found; | |
775 | } | |
776 | ||
e51df2c1 | 777 | static void start_cpu_decrementer(void) |
77c0a700 BH |
778 | { |
779 | #if defined(CONFIG_BOOKE) || defined(CONFIG_40x) | |
6e2f03e2 IM |
780 | unsigned int tcr; |
781 | ||
77c0a700 BH |
782 | /* Clear any pending timer interrupts */ |
783 | mtspr(SPRN_TSR, TSR_ENW | TSR_WIS | TSR_DIS | TSR_FIS); | |
784 | ||
6e2f03e2 IM |
785 | tcr = mfspr(SPRN_TCR); |
786 | /* | |
787 | * The watchdog may have already been enabled by u-boot. So leave | |
788 | * TRC[WP] (Watchdog Period) alone. | |
789 | */ | |
790 | tcr &= TCR_WP_MASK; /* Clear all bits except for TCR[WP] */ | |
791 | tcr |= TCR_DIE; /* Enable decrementer */ | |
792 | mtspr(SPRN_TCR, tcr); | |
793 | #endif | |
77c0a700 BH |
794 | } |
795 | ||
0bb474a4 AB |
796 | void __init generic_calibrate_decr(void) |
797 | { | |
798 | ppc_tb_freq = DEFAULT_TB_FREQ; /* hardcoded default */ | |
799 | ||
800 | if (!get_freq("ibm,extended-timebase-frequency", 2, &ppc_tb_freq) && | |
801 | !get_freq("timebase-frequency", 1, &ppc_tb_freq)) { | |
802 | ||
10f7e7c1 AB |
803 | printk(KERN_ERR "WARNING: Estimating decrementer frequency " |
804 | "(not found)\n"); | |
0bb474a4 | 805 | } |
10f7e7c1 | 806 | |
0bb474a4 AB |
807 | ppc_proc_freq = DEFAULT_PROC_FREQ; /* hardcoded default */ |
808 | ||
809 | if (!get_freq("ibm,extended-clock-frequency", 2, &ppc_proc_freq) && | |
810 | !get_freq("clock-frequency", 1, &ppc_proc_freq)) { | |
811 | ||
812 | printk(KERN_ERR "WARNING: Estimating processor frequency " | |
813 | "(not found)\n"); | |
10f7e7c1 | 814 | } |
10f7e7c1 | 815 | } |
10f7e7c1 | 816 | |
aa3be5f3 | 817 | int update_persistent_clock(struct timespec now) |
f2783c15 PM |
818 | { |
819 | struct rtc_time tm; | |
820 | ||
aa3be5f3 | 821 | if (!ppc_md.set_rtc_time) |
023f333a | 822 | return -ENODEV; |
aa3be5f3 TB |
823 | |
824 | to_tm(now.tv_sec + 1 + timezone_offset, &tm); | |
825 | tm.tm_year -= 1900; | |
826 | tm.tm_mon -= 1; | |
827 | ||
828 | return ppc_md.set_rtc_time(&tm); | |
829 | } | |
830 | ||
978d7eb3 | 831 | static void __read_persistent_clock(struct timespec *ts) |
aa3be5f3 TB |
832 | { |
833 | struct rtc_time tm; | |
834 | static int first = 1; | |
835 | ||
d90246cd | 836 | ts->tv_nsec = 0; |
aa3be5f3 TB |
837 | /* XXX this is a litle fragile but will work okay in the short term */ |
838 | if (first) { | |
839 | first = 0; | |
840 | if (ppc_md.time_init) | |
841 | timezone_offset = ppc_md.time_init(); | |
842 | ||
843 | /* get_boot_time() isn't guaranteed to be safe to call late */ | |
d90246cd MS |
844 | if (ppc_md.get_boot_time) { |
845 | ts->tv_sec = ppc_md.get_boot_time() - timezone_offset; | |
846 | return; | |
847 | } | |
848 | } | |
849 | if (!ppc_md.get_rtc_time) { | |
850 | ts->tv_sec = 0; | |
851 | return; | |
aa3be5f3 | 852 | } |
f2783c15 | 853 | ppc_md.get_rtc_time(&tm); |
978d7eb3 | 854 | |
d4f587c6 MS |
855 | ts->tv_sec = mktime(tm.tm_year+1900, tm.tm_mon+1, tm.tm_mday, |
856 | tm.tm_hour, tm.tm_min, tm.tm_sec); | |
f2783c15 PM |
857 | } |
858 | ||
978d7eb3 BH |
859 | void read_persistent_clock(struct timespec *ts) |
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 | |
b919ee82 AB |
1014 | printk_once(KERN_DEBUG "clockevent: %s mult[%x] shift[%d] cpu[%d]\n", |
1015 | dec->name, dec->mult, dec->shift, cpu); | |
d831d0b8 TB |
1016 | |
1017 | clockevents_register_device(dec); | |
1018 | } | |
1019 | ||
79901024 OH |
1020 | static void enable_large_decrementer(void) |
1021 | { | |
1022 | if (!cpu_has_feature(CPU_FTR_ARCH_300)) | |
1023 | return; | |
1024 | ||
1025 | if (decrementer_max <= DECREMENTER_DEFAULT_MAX) | |
1026 | return; | |
1027 | ||
1028 | /* | |
1029 | * If we're running as the hypervisor we need to enable the LD manually | |
1030 | * otherwise firmware should have done it for us. | |
1031 | */ | |
1032 | if (cpu_has_feature(CPU_FTR_HVMODE)) | |
1033 | mtspr(SPRN_LPCR, mfspr(SPRN_LPCR) | LPCR_LD); | |
1034 | } | |
1035 | ||
1036 | static void __init set_decrementer_max(void) | |
1037 | { | |
1038 | struct device_node *cpu; | |
1039 | u32 bits = 32; | |
1040 | ||
1041 | /* Prior to ISAv3 the decrementer is always 32 bit */ | |
1042 | if (!cpu_has_feature(CPU_FTR_ARCH_300)) | |
1043 | return; | |
1044 | ||
1045 | cpu = of_find_node_by_type(NULL, "cpu"); | |
1046 | ||
1047 | if (of_property_read_u32(cpu, "ibm,dec-bits", &bits) == 0) { | |
1048 | if (bits > 64 || bits < 32) { | |
1049 | pr_warn("time_init: firmware supplied invalid ibm,dec-bits"); | |
1050 | bits = 32; | |
1051 | } | |
1052 | ||
1053 | /* calculate the signed maximum given this many bits */ | |
1054 | decrementer_max = (1ul << (bits - 1)) - 1; | |
1055 | } | |
1056 | ||
1057 | of_node_put(cpu); | |
1058 | ||
1059 | pr_info("time_init: %u bit decrementer (max: %llx)\n", | |
1060 | bits, decrementer_max); | |
1061 | } | |
1062 | ||
c481887f | 1063 | static void __init init_decrementer_clockevent(void) |
d831d0b8 TB |
1064 | { |
1065 | int cpu = smp_processor_id(); | |
1066 | ||
d8afc6fd AB |
1067 | clockevents_calc_mult_shift(&decrementer_clockevent, ppc_tb_freq, 4); |
1068 | ||
d831d0b8 | 1069 | decrementer_clockevent.max_delta_ns = |
79901024 | 1070 | clockevent_delta2ns(decrementer_max, &decrementer_clockevent); |
115631c3 | 1071 | decrementer_clockevent.max_delta_ticks = decrementer_max; |
43875cc0 PM |
1072 | decrementer_clockevent.min_delta_ns = |
1073 | clockevent_delta2ns(2, &decrementer_clockevent); | |
115631c3 | 1074 | decrementer_clockevent.min_delta_ticks = 2; |
d831d0b8 TB |
1075 | |
1076 | register_decrementer_clockevent(cpu); | |
1077 | } | |
1078 | ||
1079 | void secondary_cpu_time_init(void) | |
1080 | { | |
79901024 OH |
1081 | /* Enable and test the large decrementer for this cpu */ |
1082 | enable_large_decrementer(); | |
1083 | ||
77c0a700 BH |
1084 | /* Start the decrementer on CPUs that have manual control |
1085 | * such as BookE | |
1086 | */ | |
1087 | start_cpu_decrementer(); | |
1088 | ||
d831d0b8 TB |
1089 | /* FIME: Should make unrelatred change to move snapshot_timebase |
1090 | * call here ! */ | |
1091 | register_decrementer_clockevent(smp_processor_id()); | |
1092 | } | |
1093 | ||
f2783c15 | 1094 | /* This function is only called on the boot processor */ |
1da177e4 LT |
1095 | void __init time_init(void) |
1096 | { | |
1da177e4 | 1097 | struct div_result res; |
d75d68cf | 1098 | u64 scale; |
f2783c15 PM |
1099 | unsigned shift; |
1100 | ||
96c44507 PM |
1101 | if (__USE_RTC()) { |
1102 | /* 601 processor: dec counts down by 128 every 128ns */ | |
1103 | ppc_tb_freq = 1000000000; | |
96c44507 PM |
1104 | } else { |
1105 | /* Normal PowerPC with timebase register */ | |
1106 | ppc_md.calibrate_decr(); | |
224ad80a | 1107 | printk(KERN_DEBUG "time_init: decrementer frequency = %lu.%.6lu MHz\n", |
96c44507 | 1108 | ppc_tb_freq / 1000000, ppc_tb_freq % 1000000); |
224ad80a | 1109 | printk(KERN_DEBUG "time_init: processor frequency = %lu.%.6lu MHz\n", |
96c44507 | 1110 | ppc_proc_freq / 1000000, ppc_proc_freq % 1000000); |
96c44507 | 1111 | } |
374e99d4 PM |
1112 | |
1113 | tb_ticks_per_jiffy = ppc_tb_freq / HZ; | |
092b8f34 | 1114 | tb_ticks_per_sec = ppc_tb_freq; |
374e99d4 | 1115 | tb_ticks_per_usec = ppc_tb_freq / 1000000; |
c6622f63 | 1116 | calc_cputime_factors(); |
092b8f34 | 1117 | |
1da177e4 LT |
1118 | /* |
1119 | * Compute scale factor for sched_clock. | |
1120 | * The calibrate_decr() function has set tb_ticks_per_sec, | |
1121 | * which is the timebase frequency. | |
1122 | * We compute 1e9 * 2^64 / tb_ticks_per_sec and interpret | |
1123 | * the 128-bit result as a 64.64 fixed-point number. | |
1124 | * We then shift that number right until it is less than 1.0, | |
1125 | * giving us the scale factor and shift count to use in | |
1126 | * sched_clock(). | |
1127 | */ | |
1128 | div128_by_32(1000000000, 0, tb_ticks_per_sec, &res); | |
1129 | scale = res.result_low; | |
1130 | for (shift = 0; res.result_high != 0; ++shift) { | |
1131 | scale = (scale >> 1) | (res.result_high << 63); | |
1132 | res.result_high >>= 1; | |
1133 | } | |
1134 | tb_to_ns_scale = scale; | |
1135 | tb_to_ns_shift = shift; | |
fc9069fe | 1136 | /* Save the current timebase to pretty up CONFIG_PRINTK_TIME */ |
c27da339 | 1137 | boot_tb = get_tb_or_rtc(); |
1da177e4 | 1138 | |
092b8f34 | 1139 | /* If platform provided a timezone (pmac), we correct the time */ |
621692cb | 1140 | if (timezone_offset) { |
092b8f34 PM |
1141 | sys_tz.tz_minuteswest = -timezone_offset / 60; |
1142 | sys_tz.tz_dsttime = 0; | |
621692cb | 1143 | } |
092b8f34 | 1144 | |
a7f290da BH |
1145 | vdso_data->tb_update_count = 0; |
1146 | vdso_data->tb_ticks_per_sec = tb_ticks_per_sec; | |
1da177e4 | 1147 | |
79901024 OH |
1148 | /* initialise and enable the large decrementer (if we have one) */ |
1149 | set_decrementer_max(); | |
1150 | enable_large_decrementer(); | |
1151 | ||
77c0a700 BH |
1152 | /* Start the decrementer on CPUs that have manual control |
1153 | * such as BookE | |
1154 | */ | |
1155 | start_cpu_decrementer(); | |
1156 | ||
f5339277 SR |
1157 | /* Register the clocksource */ |
1158 | clocksource_init(); | |
4a4cfe38 | 1159 | |
d831d0b8 | 1160 | init_decrementer_clockevent(); |
0d948730 | 1161 | tick_setup_hrtimer_broadcast(); |
f0d37300 KH |
1162 | |
1163 | #ifdef CONFIG_COMMON_CLK | |
1164 | of_clk_init(NULL); | |
1165 | #endif | |
1da177e4 LT |
1166 | } |
1167 | ||
1da177e4 | 1168 | |
1da177e4 LT |
1169 | #define FEBRUARY 2 |
1170 | #define STARTOFTIME 1970 | |
1171 | #define SECDAY 86400L | |
1172 | #define SECYR (SECDAY * 365) | |
f2783c15 PM |
1173 | #define leapyear(year) ((year) % 4 == 0 && \ |
1174 | ((year) % 100 != 0 || (year) % 400 == 0)) | |
1da177e4 LT |
1175 | #define days_in_year(a) (leapyear(a) ? 366 : 365) |
1176 | #define days_in_month(a) (month_days[(a) - 1]) | |
1177 | ||
1178 | static int month_days[12] = { | |
1179 | 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 | |
1180 | }; | |
1181 | ||
1da177e4 LT |
1182 | void to_tm(int tim, struct rtc_time * tm) |
1183 | { | |
1184 | register int i; | |
1185 | register long hms, day; | |
1186 | ||
1187 | day = tim / SECDAY; | |
1188 | hms = tim % SECDAY; | |
1189 | ||
1190 | /* Hours, minutes, seconds are easy */ | |
1191 | tm->tm_hour = hms / 3600; | |
1192 | tm->tm_min = (hms % 3600) / 60; | |
1193 | tm->tm_sec = (hms % 3600) % 60; | |
1194 | ||
1195 | /* Number of years in days */ | |
1196 | for (i = STARTOFTIME; day >= days_in_year(i); i++) | |
1197 | day -= days_in_year(i); | |
1198 | tm->tm_year = i; | |
1199 | ||
1200 | /* Number of months in days left */ | |
1201 | if (leapyear(tm->tm_year)) | |
1202 | days_in_month(FEBRUARY) = 29; | |
1203 | for (i = 1; day >= days_in_month(i); i++) | |
1204 | day -= days_in_month(i); | |
1205 | days_in_month(FEBRUARY) = 28; | |
1206 | tm->tm_mon = i; | |
1207 | ||
1208 | /* Days are what is left over (+1) from all that. */ | |
1209 | tm->tm_mday = day + 1; | |
1210 | ||
1211 | /* | |
00b912b0 | 1212 | * No-one uses the day of the week. |
1da177e4 | 1213 | */ |
00b912b0 | 1214 | tm->tm_wday = -1; |
1da177e4 | 1215 | } |
e1802b06 | 1216 | EXPORT_SYMBOL(to_tm); |
1da177e4 | 1217 | |
1da177e4 LT |
1218 | /* |
1219 | * Divide a 128-bit dividend by a 32-bit divisor, leaving a 128 bit | |
1220 | * result. | |
1221 | */ | |
f2783c15 PM |
1222 | void div128_by_32(u64 dividend_high, u64 dividend_low, |
1223 | unsigned divisor, struct div_result *dr) | |
1da177e4 | 1224 | { |
f2783c15 PM |
1225 | unsigned long a, b, c, d; |
1226 | unsigned long w, x, y, z; | |
1227 | u64 ra, rb, rc; | |
1da177e4 LT |
1228 | |
1229 | a = dividend_high >> 32; | |
1230 | b = dividend_high & 0xffffffff; | |
1231 | c = dividend_low >> 32; | |
1232 | d = dividend_low & 0xffffffff; | |
1233 | ||
f2783c15 PM |
1234 | w = a / divisor; |
1235 | ra = ((u64)(a - (w * divisor)) << 32) + b; | |
1236 | ||
f2783c15 PM |
1237 | rb = ((u64) do_div(ra, divisor) << 32) + c; |
1238 | x = ra; | |
1da177e4 | 1239 | |
f2783c15 PM |
1240 | rc = ((u64) do_div(rb, divisor) << 32) + d; |
1241 | y = rb; | |
1242 | ||
1243 | do_div(rc, divisor); | |
1244 | z = rc; | |
1da177e4 | 1245 | |
f2783c15 PM |
1246 | dr->result_high = ((u64)w << 32) + x; |
1247 | dr->result_low = ((u64)y << 32) + z; | |
1da177e4 LT |
1248 | |
1249 | } | |
bcd68a70 | 1250 | |
177996e6 BH |
1251 | /* We don't need to calibrate delay, we use the CPU timebase for that */ |
1252 | void calibrate_delay(void) | |
1253 | { | |
1254 | /* Some generic code (such as spinlock debug) use loops_per_jiffy | |
1255 | * as the number of __delay(1) in a jiffy, so make it so | |
1256 | */ | |
1257 | loops_per_jiffy = tb_ticks_per_jiffy; | |
1258 | } | |
1259 | ||
169047f4 AB |
1260 | #if IS_ENABLED(CONFIG_RTC_DRV_GENERIC) |
1261 | static int rtc_generic_get_time(struct device *dev, struct rtc_time *tm) | |
1262 | { | |
1263 | ppc_md.get_rtc_time(tm); | |
890ae797 | 1264 | return 0; |
169047f4 AB |
1265 | } |
1266 | ||
1267 | static int rtc_generic_set_time(struct device *dev, struct rtc_time *tm) | |
1268 | { | |
1269 | if (!ppc_md.set_rtc_time) | |
1270 | return -EOPNOTSUPP; | |
1271 | ||
1272 | if (ppc_md.set_rtc_time(tm) < 0) | |
1273 | return -EOPNOTSUPP; | |
1274 | ||
1275 | return 0; | |
1276 | } | |
1277 | ||
1278 | static const struct rtc_class_ops rtc_generic_ops = { | |
1279 | .read_time = rtc_generic_get_time, | |
1280 | .set_time = rtc_generic_set_time, | |
1281 | }; | |
1282 | ||
bcd68a70 GU |
1283 | static int __init rtc_init(void) |
1284 | { | |
1285 | struct platform_device *pdev; | |
1286 | ||
1287 | if (!ppc_md.get_rtc_time) | |
1288 | return -ENODEV; | |
1289 | ||
169047f4 AB |
1290 | pdev = platform_device_register_data(NULL, "rtc-generic", -1, |
1291 | &rtc_generic_ops, | |
1292 | sizeof(rtc_generic_ops)); | |
bcd68a70 | 1293 | |
8c6ffba0 | 1294 | return PTR_ERR_OR_ZERO(pdev); |
bcd68a70 GU |
1295 | } |
1296 | ||
8f6b9512 | 1297 | device_initcall(rtc_init); |
169047f4 | 1298 | #endif |