blob: 2f355d229a587771680b28080d92fd06f345d7e7 [file] [log] [blame]
Gerd Hoffmann7af192c2008-06-03 16:17:29 +02001/* paravirtual clock -- common code used by kvm/xen
2
3 This program is free software; you can redistribute it and/or modify
4 it under the terms of the GNU General Public License as published by
5 the Free Software Foundation; either version 2 of the License, or
6 (at your option) any later version.
7
8 This program is distributed in the hope that it will be useful,
9 but WITHOUT ANY WARRANTY; without even the implied warranty of
10 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 GNU General Public License for more details.
12
13 You should have received a copy of the GNU General Public License
14 along with this program; if not, write to the Free Software
15 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
16*/
17
18#include <linux/kernel.h>
19#include <linux/percpu.h>
Marcelo Tosatti71056ae2012-11-27 23:28:55 -020020#include <linux/notifier.h>
21#include <linux/sched.h>
22#include <linux/gfp.h>
23#include <linux/bootmem.h>
24#include <asm/fixmap.h>
Gerd Hoffmann7af192c2008-06-03 16:17:29 +020025#include <asm/pvclock.h>
26
Glauber Costa424c32f2010-05-11 12:17:39 -040027static u8 valid_flags __read_mostly = 0;
28
29void pvclock_set_flags(u8 flags)
30{
31 valid_flags = flags;
32}
33
Glauber Costa3807f342008-07-28 11:47:52 -030034unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src)
35{
Harvey Harrisona0854602008-09-23 11:01:45 -070036 u64 pv_tsc_khz = 1000000ULL << 32;
Glauber Costa3807f342008-07-28 11:47:52 -030037
Harvey Harrisona0854602008-09-23 11:01:45 -070038 do_div(pv_tsc_khz, src->tsc_to_system_mul);
Glauber Costa3807f342008-07-28 11:47:52 -030039 if (src->tsc_shift < 0)
Harvey Harrisona0854602008-09-23 11:01:45 -070040 pv_tsc_khz <<= -src->tsc_shift;
Glauber Costa3807f342008-07-28 11:47:52 -030041 else
Harvey Harrisona0854602008-09-23 11:01:45 -070042 pv_tsc_khz >>= src->tsc_shift;
43 return pv_tsc_khz;
Glauber Costa3807f342008-07-28 11:47:52 -030044}
45
Marcelo Tosattid63285e2013-10-11 21:39:25 -030046void pvclock_touch_watchdogs(void)
47{
48 touch_softlockup_watchdog_sync();
49 clocksource_touch_watchdog();
50 rcu_cpu_stall_reset();
Marcelo Tosatti8b414522013-10-11 21:39:26 -030051 reset_hung_task_detector();
Marcelo Tosattid63285e2013-10-11 21:39:25 -030052}
53
Glauber Costa489fb492010-05-11 12:17:40 -040054static atomic64_t last_value = ATOMIC64_INIT(0);
55
Jeremy Fitzhardingee7a34812010-10-25 16:53:46 -070056void pvclock_resume(void)
57{
58 atomic64_set(&last_value, 0);
59}
60
Marcelo Tosatti26979022012-11-27 23:28:52 -020061u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src)
62{
63 unsigned version;
64 cycle_t ret;
65 u8 flags;
66
67 do {
68 version = __pvclock_read_cycles(src, &ret, &flags);
69 } while ((src->version & 1) || version != src->version);
70
71 return flags & valid_flags;
72}
73
Gerd Hoffmann7af192c2008-06-03 16:17:29 +020074cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
75{
Gerd Hoffmann7af192c2008-06-03 16:17:29 +020076 unsigned version;
Marcelo Tosattidce2db02012-11-27 23:28:51 -020077 cycle_t ret;
Glauber Costa489fb492010-05-11 12:17:40 -040078 u64 last;
Marcelo Tosatti42b56372012-11-27 23:28:50 -020079 u8 flags;
Gerd Hoffmann7af192c2008-06-03 16:17:29 +020080
81 do {
Marcelo Tosattidce2db02012-11-27 23:28:51 -020082 version = __pvclock_read_cycles(src, &ret, &flags);
Marcelo Tosatti42b56372012-11-27 23:28:50 -020083 } while ((src->version & 1) || version != src->version);
Gerd Hoffmann7af192c2008-06-03 16:17:29 +020084
Marcelo Tosattid63285e2013-10-11 21:39:25 -030085 if (unlikely((flags & PVCLOCK_GUEST_STOPPED) != 0)) {
86 src->flags &= ~PVCLOCK_GUEST_STOPPED;
87 pvclock_touch_watchdogs();
88 }
89
Glauber Costa3a0d7252010-05-11 12:17:45 -040090 if ((valid_flags & PVCLOCK_TSC_STABLE_BIT) &&
Marcelo Tosatti42b56372012-11-27 23:28:50 -020091 (flags & PVCLOCK_TSC_STABLE_BIT))
Glauber Costa3a0d7252010-05-11 12:17:45 -040092 return ret;
93
Glauber Costa489fb492010-05-11 12:17:40 -040094 /*
95 * Assumption here is that last_value, a global accumulator, always goes
96 * forward. If we are less than that, we should not be much smaller.
97 * We assume there is an error marging we're inside, and then the correction
98 * does not sacrifice accuracy.
99 *
100 * For reads: global may have changed between test and return,
101 * but this means someone else updated poked the clock at a later time.
102 * We just need to make sure we are not seeing a backwards event.
103 *
104 * For updates: last_value = ret is not enough, since two vcpus could be
105 * updating at the same time, and one of them could be slightly behind,
106 * making the assumption that last_value always go forward fail to hold.
107 */
108 last = atomic64_read(&last_value);
109 do {
110 if (ret < last)
111 return last;
112 last = atomic64_cmpxchg(&last_value, last, ret);
113 } while (unlikely(last != ret));
114
Gerd Hoffmann7af192c2008-06-03 16:17:29 +0200115 return ret;
116}
117
118void pvclock_read_wallclock(struct pvclock_wall_clock *wall_clock,
119 struct pvclock_vcpu_time_info *vcpu_time,
120 struct timespec *ts)
121{
122 u32 version;
123 u64 delta;
124 struct timespec now;
125
126 /* get wallclock at system boot */
127 do {
128 version = wall_clock->version;
129 rmb(); /* fetch version before time */
130 now.tv_sec = wall_clock->sec;
131 now.tv_nsec = wall_clock->nsec;
132 rmb(); /* fetch time before checking version */
133 } while ((wall_clock->version & 1) || (version != wall_clock->version));
134
135 delta = pvclock_clocksource_read(vcpu_time); /* time since system boot */
136 delta += now.tv_sec * (u64)NSEC_PER_SEC + now.tv_nsec;
137
138 now.tv_nsec = do_div(delta, NSEC_PER_SEC);
139 now.tv_sec = delta;
140
141 set_normalized_timespec(ts, now.tv_sec, now.tv_nsec);
142}
Marcelo Tosatti71056ae2012-11-27 23:28:55 -0200143
Marcelo Tosatti71056ae2012-11-27 23:28:55 -0200144#ifdef CONFIG_X86_64
Marcelo Tosatti71056ae2012-11-27 23:28:55 -0200145/*
146 * Initialize the generic pvclock vsyscall state. This will allocate
147 * a/some page(s) for the per-vcpu pvclock information, set up a
148 * fixmap mapping for the page(s)
149 */
150
151int __init pvclock_init_vsyscall(struct pvclock_vsyscall_time_info *i,
152 int size)
153{
154 int idx;
155
156 WARN_ON (size != PVCLOCK_VSYSCALL_NR_PAGES*PAGE_SIZE);
157
Marcelo Tosatti71056ae2012-11-27 23:28:55 -0200158 for (idx = 0; idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) {
159 __set_fixmap(PVCLOCK_FIXMAP_BEGIN + idx,
Peter Hurley3d2a80a2013-02-27 15:28:28 -0500160 __pa(i) + (idx*PAGE_SIZE),
Marcelo Tosatti71056ae2012-11-27 23:28:55 -0200161 PAGE_KERNEL_VVAR);
162 }
163
Marcelo Tosatti71056ae2012-11-27 23:28:55 -0200164 return 0;
165}
166#endif