Commit | Line | Data |
---|---|---|
790c73f6 GOC |
1 | /* KVM paravirtual clock driver. A clocksource implementation |
2 | Copyright (C) 2008 Glauber de Oliveira Costa, Red Hat Inc. | |
3 | ||
4 | This program is free software; you can redistribute it and/or modify | |
5 | it under the terms of the GNU General Public License as published by | |
6 | the Free Software Foundation; either version 2 of the License, or | |
7 | (at your option) any later version. | |
8 | ||
9 | This program is distributed in the hope that it will be useful, | |
10 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | GNU General Public License for more details. | |
13 | ||
14 | You should have received a copy of the GNU General Public License | |
15 | along with this program; if not, write to the Free Software | |
16 | Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
17 | */ | |
18 | ||
19 | #include <linux/clocksource.h> | |
20 | #include <linux/kvm_para.h> | |
f6e16d5a | 21 | #include <asm/pvclock.h> |
790c73f6 GOC |
22 | #include <asm/msr.h> |
23 | #include <asm/apic.h> | |
24 | #include <linux/percpu.h> | |
3b5d56b9 | 25 | #include <linux/hardirq.h> |
736decac TG |
26 | |
27 | #include <asm/x86_init.h> | |
1e977aa1 | 28 | #include <asm/reboot.h> |
790c73f6 | 29 | |
790c73f6 | 30 | static int kvmclock = 1; |
838815a7 GC |
31 | static int msr_kvm_system_time = MSR_KVM_SYSTEM_TIME; |
32 | static int msr_kvm_wall_clock = MSR_KVM_WALL_CLOCK; | |
790c73f6 GOC |
33 | |
34 | static int parse_no_kvmclock(char *arg) | |
35 | { | |
36 | kvmclock = 0; | |
37 | return 0; | |
38 | } | |
39 | early_param("no-kvmclock", parse_no_kvmclock); | |
40 | ||
41 | /* The hypervisor will put information about time periodically here */ | |
f6e16d5a GH |
42 | static DEFINE_PER_CPU_SHARED_ALIGNED(struct pvclock_vcpu_time_info, hv_clock); |
43 | static struct pvclock_wall_clock wall_clock; | |
790c73f6 | 44 | |
790c73f6 GOC |
45 | /* |
46 | * The wallclock is the time of day when we booted. Since then, some time may | |
47 | * have elapsed since the hypervisor wrote the data. So we try to account for | |
48 | * that with system time | |
49 | */ | |
2ddfd20e | 50 | static unsigned long kvm_get_wallclock(void) |
790c73f6 | 51 | { |
f6e16d5a | 52 | struct pvclock_vcpu_time_info *vcpu_time; |
790c73f6 | 53 | struct timespec ts; |
790c73f6 GOC |
54 | int low, high; |
55 | ||
a20316d2 GC |
56 | low = (int)__pa_symbol(&wall_clock); |
57 | high = ((u64)__pa_symbol(&wall_clock) >> 32); | |
838815a7 GC |
58 | |
59 | native_write_msr(msr_kvm_wall_clock, low, high); | |
790c73f6 | 60 | |
f6e16d5a GH |
61 | vcpu_time = &get_cpu_var(hv_clock); |
62 | pvclock_read_wallclock(&wall_clock, vcpu_time, &ts); | |
63 | put_cpu_var(hv_clock); | |
790c73f6 | 64 | |
f6e16d5a | 65 | return ts.tv_sec; |
790c73f6 GOC |
66 | } |
67 | ||
2ddfd20e | 68 | static int kvm_set_wallclock(unsigned long now) |
790c73f6 | 69 | { |
f6e16d5a | 70 | return -1; |
790c73f6 GOC |
71 | } |
72 | ||
790c73f6 GOC |
73 | static cycle_t kvm_clock_read(void) |
74 | { | |
f6e16d5a GH |
75 | struct pvclock_vcpu_time_info *src; |
76 | cycle_t ret; | |
790c73f6 | 77 | |
95ef1e52 AK |
78 | preempt_disable_notrace(); |
79 | src = &__get_cpu_var(hv_clock); | |
f6e16d5a | 80 | ret = pvclock_clocksource_read(src); |
95ef1e52 | 81 | preempt_enable_notrace(); |
f6e16d5a | 82 | return ret; |
790c73f6 | 83 | } |
f6e16d5a | 84 | |
8e19608e MD |
85 | static cycle_t kvm_clock_get_cycles(struct clocksource *cs) |
86 | { | |
87 | return kvm_clock_read(); | |
88 | } | |
89 | ||
0293615f GC |
90 | /* |
91 | * If we don't do that, there is the possibility that the guest | |
92 | * will calibrate under heavy load - thus, getting a lower lpj - | |
93 | * and execute the delays themselves without load. This is wrong, | |
94 | * because no delay loop can finish beforehand. | |
95 | * Any heuristics is subject to fail, because ultimately, a large | |
96 | * poll of guests can be running and trouble each other. So we preset | |
97 | * lpj here | |
98 | */ | |
99 | static unsigned long kvm_get_tsc_khz(void) | |
100 | { | |
e93353c9 EH |
101 | struct pvclock_vcpu_time_info *src; |
102 | src = &per_cpu(hv_clock, 0); | |
103 | return pvclock_tsc_khz(src); | |
0293615f GC |
104 | } |
105 | ||
106 | static void kvm_get_preset_lpj(void) | |
107 | { | |
0293615f GC |
108 | unsigned long khz; |
109 | u64 lpj; | |
110 | ||
e93353c9 | 111 | khz = kvm_get_tsc_khz(); |
0293615f GC |
112 | |
113 | lpj = ((u64)khz * 1000); | |
114 | do_div(lpj, HZ); | |
115 | preset_lpj = lpj; | |
116 | } | |
117 | ||
3b5d56b9 EM |
118 | bool kvm_check_and_clear_guest_paused(void) |
119 | { | |
120 | bool ret = false; | |
121 | struct pvclock_vcpu_time_info *src; | |
122 | ||
3b5d56b9 EM |
123 | src = &__get_cpu_var(hv_clock); |
124 | if ((src->flags & PVCLOCK_GUEST_STOPPED) != 0) { | |
125 | __this_cpu_and(hv_clock.flags, ~PVCLOCK_GUEST_STOPPED); | |
126 | ret = true; | |
127 | } | |
128 | ||
129 | return ret; | |
130 | } | |
3b5d56b9 | 131 | |
790c73f6 GOC |
132 | static struct clocksource kvm_clock = { |
133 | .name = "kvm-clock", | |
8e19608e | 134 | .read = kvm_clock_get_cycles, |
790c73f6 GOC |
135 | .rating = 400, |
136 | .mask = CLOCKSOURCE_MASK(64), | |
790c73f6 GOC |
137 | .flags = CLOCK_SOURCE_IS_CONTINUOUS, |
138 | }; | |
139 | ||
ca3f1017 | 140 | int kvm_register_clock(char *txt) |
790c73f6 GOC |
141 | { |
142 | int cpu = smp_processor_id(); | |
19b6a85b AK |
143 | int low, high, ret; |
144 | ||
790c73f6 GOC |
145 | low = (int)__pa(&per_cpu(hv_clock, cpu)) | 1; |
146 | high = ((u64)__pa(&per_cpu(hv_clock, cpu)) >> 32); | |
19b6a85b | 147 | ret = native_write_msr_safe(msr_kvm_system_time, low, high); |
f6e16d5a GH |
148 | printk(KERN_INFO "kvm-clock: cpu %d, msr %x:%x, %s\n", |
149 | cpu, high, low, txt); | |
838815a7 | 150 | |
19b6a85b | 151 | return ret; |
790c73f6 GOC |
152 | } |
153 | ||
b74f05d6 MT |
154 | static void kvm_save_sched_clock_state(void) |
155 | { | |
156 | } | |
157 | ||
158 | static void kvm_restore_sched_clock_state(void) | |
159 | { | |
160 | kvm_register_clock("primary cpu clock, resume"); | |
161 | } | |
162 | ||
b8ba5f10 | 163 | #ifdef CONFIG_X86_LOCAL_APIC |
23a14b9e | 164 | static void __cpuinit kvm_setup_secondary_clock(void) |
790c73f6 GOC |
165 | { |
166 | /* | |
167 | * Now that the first cpu already had this clocksource initialized, | |
168 | * we shouldn't fail. | |
169 | */ | |
f6e16d5a | 170 | WARN_ON(kvm_register_clock("secondary cpu clock")); |
790c73f6 | 171 | } |
b8ba5f10 | 172 | #endif |
790c73f6 | 173 | |
1e977aa1 GC |
174 | /* |
175 | * After the clock is registered, the host will keep writing to the | |
176 | * registered memory location. If the guest happens to shutdown, this memory | |
177 | * won't be valid. In cases like kexec, in which you install a new kernel, this | |
178 | * means a random memory location will be kept being written. So before any | |
179 | * kind of shutdown from our side, we unregister the clock by writting anything | |
180 | * that does not have the 'enable' bit set in the msr | |
181 | */ | |
182 | #ifdef CONFIG_KEXEC | |
183 | static void kvm_crash_shutdown(struct pt_regs *regs) | |
184 | { | |
838815a7 | 185 | native_write_msr(msr_kvm_system_time, 0, 0); |
d910f5c1 | 186 | kvm_disable_steal_time(); |
1e977aa1 GC |
187 | native_machine_crash_shutdown(regs); |
188 | } | |
189 | #endif | |
190 | ||
191 | static void kvm_shutdown(void) | |
192 | { | |
838815a7 | 193 | native_write_msr(msr_kvm_system_time, 0, 0); |
d910f5c1 | 194 | kvm_disable_steal_time(); |
1e977aa1 GC |
195 | native_machine_shutdown(); |
196 | } | |
197 | ||
790c73f6 GOC |
198 | void __init kvmclock_init(void) |
199 | { | |
200 | if (!kvm_para_available()) | |
201 | return; | |
202 | ||
838815a7 GC |
203 | if (kvmclock && kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE2)) { |
204 | msr_kvm_system_time = MSR_KVM_SYSTEM_TIME_NEW; | |
205 | msr_kvm_wall_clock = MSR_KVM_WALL_CLOCK_NEW; | |
206 | } else if (!(kvmclock && kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE))) | |
207 | return; | |
208 | ||
209 | printk(KERN_INFO "kvm-clock: Using msrs %x and %x", | |
210 | msr_kvm_system_time, msr_kvm_wall_clock); | |
211 | ||
212 | if (kvm_register_clock("boot clock")) | |
213 | return; | |
214 | pv_time_ops.sched_clock = kvm_clock_read; | |
215 | x86_platform.calibrate_tsc = kvm_get_tsc_khz; | |
216 | x86_platform.get_wallclock = kvm_get_wallclock; | |
217 | x86_platform.set_wallclock = kvm_set_wallclock; | |
b8ba5f10 | 218 | #ifdef CONFIG_X86_LOCAL_APIC |
df156f90 | 219 | x86_cpuinit.early_percpu_clock_init = |
838815a7 | 220 | kvm_setup_secondary_clock; |
b8ba5f10 | 221 | #endif |
b74f05d6 MT |
222 | x86_platform.save_sched_clock_state = kvm_save_sched_clock_state; |
223 | x86_platform.restore_sched_clock_state = kvm_restore_sched_clock_state; | |
838815a7 | 224 | machine_ops.shutdown = kvm_shutdown; |
1e977aa1 | 225 | #ifdef CONFIG_KEXEC |
838815a7 | 226 | machine_ops.crash_shutdown = kvm_crash_shutdown; |
1e977aa1 | 227 | #endif |
838815a7 | 228 | kvm_get_preset_lpj(); |
b01cc1b0 | 229 | clocksource_register_hz(&kvm_clock, NSEC_PER_SEC); |
838815a7 GC |
230 | pv_info.paravirt_enabled = 1; |
231 | pv_info.name = "KVM"; | |
3a0d7256 GC |
232 | |
233 | if (kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE_STABLE_BIT)) | |
234 | pvclock_set_flags(PVCLOCK_TSC_STABLE_BIT); | |
790c73f6 | 235 | } |