2 * Copyright (c) 1998-2003 Poul-Henning Kamp
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include "opt_clock.h"
31 #include "opt_kdtrace.h"
33 #include <sys/param.h>
36 #include <sys/malloc.h>
37 #include <sys/systm.h>
38 #include <sys/sysctl.h>
40 #include <sys/timetc.h>
41 #include <sys/kernel.h>
42 #include <sys/power.h>
44 #include <machine/clock.h>
45 #include <machine/md_var.h>
46 #include <machine/specialreg.h>
49 #include <sys/dtrace_bsd.h>
52 #include "cpufreq_if.h"
57 static eventhandler_tag tsc_levels_tag, tsc_pre_tag, tsc_post_tag;
59 SYSCTL_INT(_kern_timecounter, OID_AUTO, invariant_tsc, CTLFLAG_RDTUN,
60 &tsc_is_invariant, 0, "Indicates whether the TSC is P-state invariant");
61 TUNABLE_INT("kern.timecounter.invariant_tsc", &tsc_is_invariant);
65 SYSCTL_INT(_kern_timecounter, OID_AUTO, smp_tsc, CTLFLAG_RDTUN, &smp_tsc, 0,
66 "Indicates whether the TSC is safe to use in SMP mode");
67 TUNABLE_INT("kern.timecounter.smp_tsc", &smp_tsc);
70 static void tsc_freq_changed(void *arg, const struct cf_level *level,
72 static void tsc_freq_changing(void *arg, const struct cf_level *level,
74 static unsigned tsc_get_timecount(struct timecounter *tc);
75 static void tsc_levels_changed(void *arg, int unit);
77 static struct timecounter tsc_timecounter = {
78 tsc_get_timecount, /* get_timecount */
80 ~0u, /* counter_mask */
83 800, /* quality (adjusted in code) */
92 printf("Calibrating TSC clock ... ");
98 tsc_freq = tscval[1] - tscval[0];
100 printf("TSC clock: %lu Hz\n", tsc_freq);
103 * Inform CPU accounting about our boot-time clock rate. Once the
104 * system is finished booting, we will get the real max clock rate
105 * via tsc_freq_max(). This also will be updated if someone loads
106 * a cpufreq driver after boot that discovers a new max frequency.
108 set_cputicker(rdtsc, tsc_freq, 1);
110 /* Register to find out about changes in CPU frequency. */
111 tsc_pre_tag = EVENTHANDLER_REGISTER(cpufreq_pre_change,
112 tsc_freq_changing, NULL, EVENTHANDLER_PRI_FIRST);
113 tsc_post_tag = EVENTHANDLER_REGISTER(cpufreq_post_change,
114 tsc_freq_changed, NULL, EVENTHANDLER_PRI_FIRST);
115 tsc_levels_tag = EVENTHANDLER_REGISTER(cpufreq_levels_changed,
116 tsc_levels_changed, NULL, EVENTHANDLER_PRI_ANY);
125 * We can not use the TSC in SMP mode unless the TSCs on all CPUs
126 * are somehow synchronized. Some hardware configurations do
127 * this, but we have no way of determining whether this is the
128 * case, so we do not use the TSC in multi-processor systems
129 * unless the user indicated (by setting kern.timecounter.smp_tsc
130 * to 1) that he believes that his TSCs are synchronized.
132 if (mp_ncpus > 1 && !smp_tsc)
133 tsc_timecounter.tc_quality = -100;
136 if (tsc_freq != 0 && !tsc_is_broken) {
137 tsc_timecounter.tc_frequency = tsc_freq;
138 tc_init(&tsc_timecounter);
143 * When cpufreq levels change, find out about the (new) max frequency. We
144 * use this to update CPU accounting in case it got a lower estimate at boot.
147 tsc_levels_changed(void *arg, int unit)
150 struct cf_level *levels;
154 /* Only use values from the first CPU, assuming all are equal. */
158 /* Find the appropriate cpufreq device instance. */
159 cf_dev = devclass_get_device(devclass_find("cpufreq"), unit);
160 if (cf_dev == NULL) {
161 printf("tsc_levels_changed() called but no cpufreq device?\n");
165 /* Get settings from the device and find the max frequency. */
167 levels = malloc(count * sizeof(*levels), M_TEMP, M_NOWAIT);
170 error = CPUFREQ_LEVELS(cf_dev, levels, &count);
171 if (error == 0 && count != 0) {
172 max_freq = (uint64_t)levels[0].total_set.freq * 1000000;
173 set_cputicker(rdtsc, max_freq, 1);
175 printf("tsc_levels_changed: no max freq found\n");
176 free(levels, M_TEMP);
180 * If the TSC timecounter is in use, veto the pending change. It may be
181 * possible in the future to handle a dynamically-changing timecounter rate.
184 tsc_freq_changing(void *arg, const struct cf_level *level, int *status)
187 if (*status != 0 || timecounter != &tsc_timecounter ||
191 printf("timecounter TSC must not be in use when "
192 "changing frequencies; change denied\n");
196 /* Update TSC freq with the value indicated by the caller. */
198 tsc_freq_changed(void *arg, const struct cf_level *level, int status)
201 * If there was an error during the transition or
202 * TSC is P-state invariant, don't do anything.
204 if (status != 0 || tsc_is_invariant)
207 /* Total setting for this level gives the new frequency in MHz. */
208 tsc_freq = (uint64_t)level->total_set.freq * 1000000;
209 tsc_timecounter.tc_frequency = tsc_freq;
213 sysctl_machdep_tsc_freq(SYSCTL_HANDLER_ARGS)
218 if (tsc_timecounter.tc_frequency == 0)
221 error = sysctl_handle_quad(oidp, &freq, 0, req);
222 if (error == 0 && req->newptr != NULL) {
224 tsc_timecounter.tc_frequency = tsc_freq;
229 SYSCTL_PROC(_machdep, OID_AUTO, tsc_freq, CTLTYPE_QUAD | CTLFLAG_RW,
230 0, sizeof(u_int), sysctl_machdep_tsc_freq, "QU", "");
233 tsc_get_timecount(struct timecounter *tc)
240 * DTrace needs a high resolution time function which can
241 * be called from a probe context and guaranteed not to have
242 * instrumented with probes itself.
244 * Returns nanoseconds since boot.
249 return (rdtsc() * (uint64_t) 1000000000 / tsc_freq);