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