arch/x86/kernel/kvmclock.c

   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>
  21 #include <asm/pvclock.h>
  22 #include <asm/arch_hooks.h>
  23 #include <asm/msr.h>
  24 #include <asm/apic.h>
  25 #include <linux/percpu.h>
  26 #include <asm/reboot.h>
  27
  28 #define KVM_SCALE 22
  29
  30 static int kvmclock = 1;
  31
  32 static int parse_no_kvmclock(char *arg)
  33 {
  34         kvmclock = 0;
  35         return 0;
  36 }
  37 early_param("no-kvmclock", parse_no_kvmclock);
  38
  39 /* The hypervisor will put information about time periodically here */
  40 static DEFINE_PER_CPU_SHARED_ALIGNED(struct pvclock_vcpu_time_info, hv_clock);
  41 static struct pvclock_wall_clock wall_clock;
  42
  43 /*
  44  * The wallclock is the time of day when we booted. Since then, some time may
  45  * have elapsed since the hypervisor wrote the data. So we try to account for
  46  * that with system time
  47  */
  48 static unsigned long kvm_get_wallclock(void)
  49 {
  50         struct pvclock_vcpu_time_info *vcpu_time;
  51         struct timespec ts;
  52         int low, high;
  53
  54         low = (int)__pa(&wall_clock);
  55         high = ((u64)__pa(&wall_clock) >> 32);
  56         native_write_msr(MSR_KVM_WALL_CLOCK, low, high);
  57
  58         vcpu_time = &get_cpu_var(hv_clock);
  59         pvclock_read_wallclock(&wall_clock, vcpu_time, &ts);
  60         put_cpu_var(hv_clock);
  61
  62         return ts.tv_sec;
  63 }
  64
  65 static int kvm_set_wallclock(unsigned long now)
  66 {
  67         return -1;
  68 }
  69
  70 static cycle_t kvm_clock_read(void)
  71 {
  72         struct pvclock_vcpu_time_info *src;
  73         cycle_t ret;
  74
  75         src = &get_cpu_var(hv_clock);
  76         ret = pvclock_clocksource_read(src);
  77         put_cpu_var(hv_clock);
  78         return ret;
  79 }
  80
  81 /*
  82  * If we don't do that, there is the possibility that the guest
  83  * will calibrate under heavy load - thus, getting a lower lpj -
  84  * and execute the delays themselves without load. This is wrong,
  85  * because no delay loop can finish beforehand.
  86  * Any heuristics is subject to fail, because ultimately, a large
  87  * poll of guests can be running and trouble each other. So we preset
  88  * lpj here
  89  */
  90 static unsigned long kvm_get_tsc_khz(void)
  91 {
  92         return preset_lpj;
  93 }
  94
  95 static void kvm_get_preset_lpj(void)
  96 {
  97         struct pvclock_vcpu_time_info *src;
  98         unsigned long khz;
  99         u64 lpj;
 100
 101         src = &per_cpu(hv_clock, 0);
 102         khz = pvclock_tsc_khz(src);
 103
 104         lpj = ((u64)khz * 1000);
 105         do_div(lpj, HZ);
 106         preset_lpj = lpj;
 107 }
 108
 109 static struct clocksource kvm_clock = {
 110         .name = "kvm-clock",
 111         .read = kvm_clock_read,
 112         .rating = 400,
 113         .mask = CLOCKSOURCE_MASK(64),
 114         .mult = 1 << KVM_SCALE,
 115         .shift = KVM_SCALE,
 116         .flags = CLOCK_SOURCE_IS_CONTINUOUS,
 117 };
 118
 119 static int kvm_register_clock(char *txt)
 120 {
 121         int cpu = smp_processor_id();
 122         int low, high;
 123         low = (int)__pa(&per_cpu(hv_clock, cpu)) | 1;
 124         high = ((u64)__pa(&per_cpu(hv_clock, cpu)) >> 32);
 125         printk(KERN_INFO "kvm-clock: cpu %d, msr %x:%x, %s\n",
 126                cpu, high, low, txt);
 127         return native_write_msr_safe(MSR_KVM_SYSTEM_TIME, low, high);
 128 }
 129
 130 #ifdef CONFIG_X86_LOCAL_APIC
 131 static void __cpuinit kvm_setup_secondary_clock(void)
 132 {
 133         /*
 134          * Now that the first cpu already had this clocksource initialized,
 135          * we shouldn't fail.
 136          */
 137         WARN_ON(kvm_register_clock("secondary cpu clock"));
 138         /* ok, done with our trickery, call native */
 139         setup_secondary_APIC_clock();
 140 }
 141 #endif
 142
 143 #ifdef CONFIG_SMP
 144 static void __init kvm_smp_prepare_boot_cpu(void)
 145 {
 146         WARN_ON(kvm_register_clock("primary cpu clock"));
 147         native_smp_prepare_boot_cpu();
 148 }
 149 #endif
 150
 151 /*
 152  * After the clock is registered, the host will keep writing to the
 153  * registered memory location. If the guest happens to shutdown, this memory
 154  * won't be valid. In cases like kexec, in which you install a new kernel, this
 155  * means a random memory location will be kept being written. So before any
 156  * kind of shutdown from our side, we unregister the clock by writting anything
 157  * that does not have the 'enable' bit set in the msr
 158  */
 159 #ifdef CONFIG_KEXEC
 160 static void kvm_crash_shutdown(struct pt_regs *regs)
 161 {
 162         native_write_msr_safe(MSR_KVM_SYSTEM_TIME, 0, 0);
 163         native_machine_crash_shutdown(regs);
 164 }
 165 #endif
 166
 167 static void kvm_shutdown(void)
 168 {
 169         native_write_msr_safe(MSR_KVM_SYSTEM_TIME, 0, 0);
 170         native_machine_shutdown();
 171 }
 172
 173 void __init kvmclock_init(void)
 174 {
 175         if (!kvm_para_available())
 176                 return;
 177
 178         if (kvmclock && kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE)) {
 179                 if (kvm_register_clock("boot clock"))
 180                         return;
 181                 pv_time_ops.get_wallclock = kvm_get_wallclock;
 182                 pv_time_ops.set_wallclock = kvm_set_wallclock;
 183                 pv_time_ops.sched_clock = kvm_clock_read;
 184                 pv_time_ops.get_tsc_khz = kvm_get_tsc_khz;
 185 #ifdef CONFIG_X86_LOCAL_APIC
 186                 pv_apic_ops.setup_secondary_clock = kvm_setup_secondary_clock;
 187 #endif
 188 #ifdef CONFIG_SMP
 189                 smp_ops.smp_prepare_boot_cpu = kvm_smp_prepare_boot_cpu;
 190 #endif
 191                 machine_ops.shutdown  = kvm_shutdown;
 192 #ifdef CONFIG_KEXEC
 193                 machine_ops.crash_shutdown  = kvm_crash_shutdown;
 194 #endif
 195                 kvm_get_preset_lpj();
 196                 clocksource_register(&kvm_clock);
 197         }
 198 }