1 /* KVM paravirtual clock driver. A clocksource implementation
2 Copyright (C) 2008 Glauber de Oliveira Costa, Red Hat Inc.
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.
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.
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
19 #include <linux/clocksource.h>
20 #include <linux/kvm_para.h>
21 #include <asm/pvclock.h>
24 #include <linux/percpu.h>
25 #include <linux/hardirq.h>
26 #include <linux/memblock.h>
28 #include <asm/x86_init.h>
29 #include <asm/reboot.h>
31 static int kvmclock
= 1;
32 static int msr_kvm_system_time
= MSR_KVM_SYSTEM_TIME
;
33 static int msr_kvm_wall_clock
= MSR_KVM_WALL_CLOCK
;
35 static int parse_no_kvmclock(char *arg
)
40 early_param("no-kvmclock", parse_no_kvmclock
);
42 /* The hypervisor will put information about time periodically here */
43 static struct pvclock_vsyscall_time_info
*hv_clock
;
44 static struct pvclock_wall_clock wall_clock
;
47 * The wallclock is the time of day when we booted. Since then, some time may
48 * have elapsed since the hypervisor wrote the data. So we try to account for
49 * that with system time
51 static void kvm_get_wallclock(struct timespec
*now
)
53 struct pvclock_vcpu_time_info
*vcpu_time
;
57 low
= (int)__pa_symbol(&wall_clock
);
58 high
= ((u64
)__pa_symbol(&wall_clock
) >> 32);
60 native_write_msr(msr_kvm_wall_clock
, low
, high
);
63 cpu
= smp_processor_id();
65 vcpu_time
= &hv_clock
[cpu
].pvti
;
66 pvclock_read_wallclock(&wall_clock
, vcpu_time
, now
);
71 static int kvm_set_wallclock(const struct timespec
*now
)
76 static cycle_t
kvm_clock_read(void)
78 struct pvclock_vcpu_time_info
*src
;
82 preempt_disable_notrace();
83 cpu
= smp_processor_id();
84 src
= &hv_clock
[cpu
].pvti
;
85 ret
= pvclock_clocksource_read(src
);
86 preempt_enable_notrace();
90 static cycle_t
kvm_clock_get_cycles(struct clocksource
*cs
)
92 return kvm_clock_read();
96 * If we don't do that, there is the possibility that the guest
97 * will calibrate under heavy load - thus, getting a lower lpj -
98 * and execute the delays themselves without load. This is wrong,
99 * because no delay loop can finish beforehand.
100 * Any heuristics is subject to fail, because ultimately, a large
101 * poll of guests can be running and trouble each other. So we preset
104 static unsigned long kvm_get_tsc_khz(void)
106 struct pvclock_vcpu_time_info
*src
;
108 unsigned long tsc_khz
;
111 cpu
= smp_processor_id();
112 src
= &hv_clock
[cpu
].pvti
;
113 tsc_khz
= pvclock_tsc_khz(src
);
118 static void kvm_get_preset_lpj(void)
123 khz
= kvm_get_tsc_khz();
125 lpj
= ((u64
)khz
* 1000);
130 bool kvm_check_and_clear_guest_paused(void)
133 struct pvclock_vcpu_time_info
*src
;
134 int cpu
= smp_processor_id();
139 src
= &hv_clock
[cpu
].pvti
;
140 if ((src
->flags
& PVCLOCK_GUEST_STOPPED
) != 0) {
141 src
->flags
&= ~PVCLOCK_GUEST_STOPPED
;
142 pvclock_touch_watchdogs();
149 static struct clocksource kvm_clock
= {
151 .read
= kvm_clock_get_cycles
,
153 .mask
= CLOCKSOURCE_MASK(64),
154 .flags
= CLOCK_SOURCE_IS_CONTINUOUS
,
157 int kvm_register_clock(char *txt
)
159 int cpu
= smp_processor_id();
161 struct pvclock_vcpu_time_info
*src
;
166 src
= &hv_clock
[cpu
].pvti
;
167 low
= (int)slow_virt_to_phys(src
) | 1;
168 high
= ((u64
)slow_virt_to_phys(src
) >> 32);
169 ret
= native_write_msr_safe(msr_kvm_system_time
, low
, high
);
170 printk(KERN_INFO
"kvm-clock: cpu %d, msr %x:%x, %s\n",
171 cpu
, high
, low
, txt
);
176 static void kvm_save_sched_clock_state(void)
180 static void kvm_restore_sched_clock_state(void)
182 kvm_register_clock("primary cpu clock, resume");
185 #ifdef CONFIG_X86_LOCAL_APIC
186 static void kvm_setup_secondary_clock(void)
189 * Now that the first cpu already had this clocksource initialized,
192 WARN_ON(kvm_register_clock("secondary cpu clock"));
197 * After the clock is registered, the host will keep writing to the
198 * registered memory location. If the guest happens to shutdown, this memory
199 * won't be valid. In cases like kexec, in which you install a new kernel, this
200 * means a random memory location will be kept being written. So before any
201 * kind of shutdown from our side, we unregister the clock by writting anything
202 * that does not have the 'enable' bit set in the msr
205 static void kvm_crash_shutdown(struct pt_regs
*regs
)
207 native_write_msr(msr_kvm_system_time
, 0, 0);
208 kvm_disable_steal_time();
209 native_machine_crash_shutdown(regs
);
213 static void kvm_shutdown(void)
215 native_write_msr(msr_kvm_system_time
, 0, 0);
216 kvm_disable_steal_time();
217 native_machine_shutdown();
220 void __init
kvmclock_init(void)
225 size
= PAGE_ALIGN(sizeof(struct pvclock_vsyscall_time_info
)*NR_CPUS
);
227 if (!kvm_para_available())
230 if (kvmclock
&& kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE2
)) {
231 msr_kvm_system_time
= MSR_KVM_SYSTEM_TIME_NEW
;
232 msr_kvm_wall_clock
= MSR_KVM_WALL_CLOCK_NEW
;
233 } else if (!(kvmclock
&& kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE
)))
236 printk(KERN_INFO
"kvm-clock: Using msrs %x and %x",
237 msr_kvm_system_time
, msr_kvm_wall_clock
);
239 mem
= memblock_alloc(size
, PAGE_SIZE
);
242 hv_clock
= __va(mem
);
243 memset(hv_clock
, 0, size
);
245 if (kvm_register_clock("primary cpu clock")) {
247 memblock_free(mem
, size
);
250 pv_time_ops
.sched_clock
= kvm_clock_read
;
251 x86_platform
.calibrate_tsc
= kvm_get_tsc_khz
;
252 x86_platform
.get_wallclock
= kvm_get_wallclock
;
253 x86_platform
.set_wallclock
= kvm_set_wallclock
;
254 #ifdef CONFIG_X86_LOCAL_APIC
255 x86_cpuinit
.early_percpu_clock_init
=
256 kvm_setup_secondary_clock
;
258 x86_platform
.save_sched_clock_state
= kvm_save_sched_clock_state
;
259 x86_platform
.restore_sched_clock_state
= kvm_restore_sched_clock_state
;
260 machine_ops
.shutdown
= kvm_shutdown
;
262 machine_ops
.crash_shutdown
= kvm_crash_shutdown
;
264 kvm_get_preset_lpj();
265 clocksource_register_hz(&kvm_clock
, NSEC_PER_SEC
);
266 pv_info
.paravirt_enabled
= 1;
267 pv_info
.name
= "KVM";
269 if (kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE_STABLE_BIT
))
270 pvclock_set_flags(PVCLOCK_TSC_STABLE_BIT
);
273 int __init
kvm_setup_vsyscall_timeinfo(void)
279 struct pvclock_vcpu_time_info
*vcpu_time
;
285 size
= PAGE_ALIGN(sizeof(struct pvclock_vsyscall_time_info
)*NR_CPUS
);
288 cpu
= smp_processor_id();
290 vcpu_time
= &hv_clock
[cpu
].pvti
;
291 flags
= pvclock_read_flags(vcpu_time
);
293 if (!(flags
& PVCLOCK_TSC_STABLE_BIT
)) {
298 if ((ret
= pvclock_init_vsyscall(hv_clock
, size
))) {
305 kvm_clock
.archdata
.vclock_mode
= VCLOCK_PVCLOCK
;