1 /* 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>
20 #include <linux/notifier.h>
21 #include <linux/sched.h>
22 #include <linux/gfp.h>
23 #include <linux/bootmem.h>
24 #include <asm/fixmap.h>
25 #include <asm/pvclock.h>
26
27 static u8 valid_flags __read_mostly = 0;
28
pvclock_set_flags(u8 flags)29 void pvclock_set_flags(u8 flags)
30 {
31 valid_flags = flags;
32 }
33
pvclock_tsc_khz(struct pvclock_vcpu_time_info * src)34 unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src)
35 {
36 u64 pv_tsc_khz = 1000000ULL << 32;
37
38 do_div(pv_tsc_khz, src->tsc_to_system_mul);
39 if (src->tsc_shift < 0)
40 pv_tsc_khz <<= -src->tsc_shift;
41 else
42 pv_tsc_khz >>= src->tsc_shift;
43 return pv_tsc_khz;
44 }
45
pvclock_touch_watchdogs(void)46 void pvclock_touch_watchdogs(void)
47 {
48 touch_softlockup_watchdog_sync();
49 clocksource_touch_watchdog();
50 rcu_cpu_stall_reset();
51 reset_hung_task_detector();
52 }
53
54 static atomic64_t last_value = ATOMIC64_INIT(0);
55
pvclock_resume(void)56 void pvclock_resume(void)
57 {
58 atomic64_set(&last_value, 0);
59 }
60
pvclock_read_flags(struct pvclock_vcpu_time_info * src)61 u8 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 /* Make sure that the version double-check is last. */
70 smp_rmb();
71 } while ((src->version & 1) || version != src->version);
72
73 return flags & valid_flags;
74 }
75
pvclock_clocksource_read(struct pvclock_vcpu_time_info * src)76 cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
77 {
78 unsigned version;
79 cycle_t ret;
80 u64 last;
81 u8 flags;
82
83 do {
84 version = __pvclock_read_cycles(src, &ret, &flags);
85 /* Make sure that the version double-check is last. */
86 smp_rmb();
87 } while ((src->version & 1) || version != src->version);
88
89 if (unlikely((flags & PVCLOCK_GUEST_STOPPED) != 0)) {
90 src->flags &= ~PVCLOCK_GUEST_STOPPED;
91 pvclock_touch_watchdogs();
92 }
93
94 if ((valid_flags & PVCLOCK_TSC_STABLE_BIT) &&
95 (flags & PVCLOCK_TSC_STABLE_BIT))
96 return ret;
97
98 /*
99 * Assumption here is that last_value, a global accumulator, always goes
100 * forward. If we are less than that, we should not be much smaller.
101 * We assume there is an error marging we're inside, and then the correction
102 * does not sacrifice accuracy.
103 *
104 * For reads: global may have changed between test and return,
105 * but this means someone else updated poked the clock at a later time.
106 * We just need to make sure we are not seeing a backwards event.
107 *
108 * For updates: last_value = ret is not enough, since two vcpus could be
109 * updating at the same time, and one of them could be slightly behind,
110 * making the assumption that last_value always go forward fail to hold.
111 */
112 last = atomic64_read(&last_value);
113 do {
114 if (ret < last)
115 return last;
116 last = atomic64_cmpxchg(&last_value, last, ret);
117 } while (unlikely(last != ret));
118
119 return ret;
120 }
121
pvclock_read_wallclock(struct pvclock_wall_clock * wall_clock,struct pvclock_vcpu_time_info * vcpu_time,struct timespec * ts)122 void pvclock_read_wallclock(struct pvclock_wall_clock *wall_clock,
123 struct pvclock_vcpu_time_info *vcpu_time,
124 struct timespec *ts)
125 {
126 u32 version;
127 u64 delta;
128 struct timespec now;
129
130 /* get wallclock at system boot */
131 do {
132 version = wall_clock->version;
133 rmb(); /* fetch version before time */
134 now.tv_sec = wall_clock->sec;
135 now.tv_nsec = wall_clock->nsec;
136 rmb(); /* fetch time before checking version */
137 } while ((wall_clock->version & 1) || (version != wall_clock->version));
138
139 delta = pvclock_clocksource_read(vcpu_time); /* time since system boot */
140 delta += now.tv_sec * (u64)NSEC_PER_SEC + now.tv_nsec;
141
142 now.tv_nsec = do_div(delta, NSEC_PER_SEC);
143 now.tv_sec = delta;
144
145 set_normalized_timespec(ts, now.tv_sec, now.tv_nsec);
146 }
147
148 #ifdef CONFIG_X86_64
149 /*
150 * Initialize the generic pvclock vsyscall state. This will allocate
151 * a/some page(s) for the per-vcpu pvclock information, set up a
152 * fixmap mapping for the page(s)
153 */
154
pvclock_init_vsyscall(struct pvclock_vsyscall_time_info * i,int size)155 int __init pvclock_init_vsyscall(struct pvclock_vsyscall_time_info *i,
156 int size)
157 {
158 int idx;
159
160 WARN_ON (size != PVCLOCK_VSYSCALL_NR_PAGES*PAGE_SIZE);
161
162 for (idx = 0; idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) {
163 __set_fixmap(PVCLOCK_FIXMAP_BEGIN + idx,
164 __pa(i) + (idx*PAGE_SIZE),
165 PAGE_KERNEL_VVAR);
166 }
167
168 return 0;
169 }
170 #endif
171