initial commit with v2.6.32.60
[linux-2.6.32.60-moxart.git] / arch / x86 / kernel / pvclock.c
blob929047cf9622b0721e575a342304db94119d625a
1 /* paravirtual clock -- common code used by kvm/xen
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.
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.
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
18 #include <linux/kernel.h>
19 #include <linux/percpu.h>
20 #include <asm/pvclock.h>
23 * These are perodically updated
24 * xen: magic shared_info page
25 * kvm: gpa registered via msr
26 * and then copied here.
28 struct pvclock_shadow_time {
29 u64 tsc_timestamp; /* TSC at last update of time vals. */
30 u64 system_timestamp; /* Time, in nanosecs, since boot. */
31 u32 tsc_to_nsec_mul;
32 int tsc_shift;
33 u32 version;
37 * Scale a 64-bit delta by scaling and multiplying by a 32-bit fraction,
38 * yielding a 64-bit result.
40 static inline u64 scale_delta(u64 delta, u32 mul_frac, int shift)
42 u64 product;
43 #ifdef __i386__
44 u32 tmp1, tmp2;
45 #endif
47 if (shift < 0)
48 delta >>= -shift;
49 else
50 delta <<= shift;
52 #ifdef __i386__
53 __asm__ (
54 "mul %5 ; "
55 "mov %4,%%eax ; "
56 "mov %%edx,%4 ; "
57 "mul %5 ; "
58 "xor %5,%5 ; "
59 "add %4,%%eax ; "
60 "adc %5,%%edx ; "
61 : "=A" (product), "=r" (tmp1), "=r" (tmp2)
62 : "a" ((u32)delta), "1" ((u32)(delta >> 32)), "2" (mul_frac) );
63 #elif defined(__x86_64__)
64 __asm__ (
65 "mul %%rdx ; shrd $32,%%rdx,%%rax"
66 : "=a" (product) : "0" (delta), "d" ((u64)mul_frac) );
67 #else
68 #error implement me!
69 #endif
71 return product;
74 static u64 pvclock_get_nsec_offset(struct pvclock_shadow_time *shadow)
76 u64 delta = native_read_tsc() - shadow->tsc_timestamp;
77 return pvclock_scale_delta(delta, shadow->tsc_to_nsec_mul,
78 shadow->tsc_shift);
82 * Reads a consistent set of time-base values from hypervisor,
83 * into a shadow data area.
85 static unsigned pvclock_get_time_values(struct pvclock_shadow_time *dst,
86 struct pvclock_vcpu_time_info *src)
88 do {
89 dst->version = src->version;
90 rmb(); /* fetch version before data */
91 dst->tsc_timestamp = src->tsc_timestamp;
92 dst->system_timestamp = src->system_time;
93 dst->tsc_to_nsec_mul = src->tsc_to_system_mul;
94 dst->tsc_shift = src->tsc_shift;
95 rmb(); /* test version after fetching data */
96 } while ((src->version & 1) || (dst->version != src->version));
98 return dst->version;
101 unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src)
103 u64 pv_tsc_khz = 1000000ULL << 32;
105 do_div(pv_tsc_khz, src->tsc_to_system_mul);
106 if (src->tsc_shift < 0)
107 pv_tsc_khz <<= -src->tsc_shift;
108 else
109 pv_tsc_khz >>= src->tsc_shift;
110 return pv_tsc_khz;
113 static atomic64_t last_value = ATOMIC64_INIT(0);
115 void pvclock_resume(void)
117 atomic64_set(&last_value, 0);
120 cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
122 struct pvclock_shadow_time shadow;
123 unsigned version;
124 cycle_t ret, offset;
125 u64 last;
127 do {
128 version = pvclock_get_time_values(&shadow, src);
129 barrier();
130 offset = pvclock_get_nsec_offset(&shadow);
131 ret = shadow.system_timestamp + offset;
132 barrier();
133 } while (version != src->version);
136 * Assumption here is that last_value, a global accumulator, always goes
137 * forward. If we are less than that, we should not be much smaller.
138 * We assume there is an error marging we're inside, and then the correction
139 * does not sacrifice accuracy.
141 * For reads: global may have changed between test and return,
142 * but this means someone else updated poked the clock at a later time.
143 * We just need to make sure we are not seeing a backwards event.
145 * For updates: last_value = ret is not enough, since two vcpus could be
146 * updating at the same time, and one of them could be slightly behind,
147 * making the assumption that last_value always go forward fail to hold.
149 last = atomic64_read(&last_value);
150 do {
151 if (ret < last)
152 return last;
153 last = atomic64_cmpxchg(&last_value, last, ret);
154 } while (unlikely(last != ret));
156 return ret;
159 void pvclock_read_wallclock(struct pvclock_wall_clock *wall_clock,
160 struct pvclock_vcpu_time_info *vcpu_time,
161 struct timespec *ts)
163 u32 version;
164 u64 delta;
165 struct timespec now;
167 /* get wallclock at system boot */
168 do {
169 version = wall_clock->version;
170 rmb(); /* fetch version before time */
171 now.tv_sec = wall_clock->sec;
172 now.tv_nsec = wall_clock->nsec;
173 rmb(); /* fetch time before checking version */
174 } while ((wall_clock->version & 1) || (version != wall_clock->version));
176 delta = pvclock_clocksource_read(vcpu_time); /* time since system boot */
177 delta += now.tv_sec * (u64)NSEC_PER_SEC + now.tv_nsec;
179 now.tv_nsec = do_div(delta, NSEC_PER_SEC);
180 now.tv_sec = delta;
182 set_normalized_timespec(ts, now.tv_sec, now.tv_nsec);