Glauber de Oliveira Costa | 790c73f | 2008-02-15 17:52:48 -0200 | [diff] [blame^] | 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/arch_hooks.h> |
| 22 | #include <asm/msr.h> |
| 23 | #include <asm/apic.h> |
| 24 | #include <linux/percpu.h> |
| 25 | |
| 26 | #define KVM_SCALE 22 |
| 27 | |
| 28 | static int kvmclock = 1; |
| 29 | |
| 30 | static int parse_no_kvmclock(char *arg) |
| 31 | { |
| 32 | kvmclock = 0; |
| 33 | return 0; |
| 34 | } |
| 35 | early_param("no-kvmclock", parse_no_kvmclock); |
| 36 | |
| 37 | /* The hypervisor will put information about time periodically here */ |
| 38 | static DEFINE_PER_CPU_SHARED_ALIGNED(struct kvm_vcpu_time_info, hv_clock); |
| 39 | #define get_clock(cpu, field) per_cpu(hv_clock, cpu).field |
| 40 | |
| 41 | static inline u64 kvm_get_delta(u64 last_tsc) |
| 42 | { |
| 43 | int cpu = smp_processor_id(); |
| 44 | u64 delta = native_read_tsc() - last_tsc; |
| 45 | return (delta * get_clock(cpu, tsc_to_system_mul)) >> KVM_SCALE; |
| 46 | } |
| 47 | |
| 48 | static struct kvm_wall_clock wall_clock; |
| 49 | static cycle_t kvm_clock_read(void); |
| 50 | /* |
| 51 | * The wallclock is the time of day when we booted. Since then, some time may |
| 52 | * have elapsed since the hypervisor wrote the data. So we try to account for |
| 53 | * that with system time |
| 54 | */ |
| 55 | unsigned long kvm_get_wallclock(void) |
| 56 | { |
| 57 | u32 wc_sec, wc_nsec; |
| 58 | u64 delta; |
| 59 | struct timespec ts; |
| 60 | int version, nsec; |
| 61 | int low, high; |
| 62 | |
| 63 | low = (int)__pa(&wall_clock); |
| 64 | high = ((u64)__pa(&wall_clock) >> 32); |
| 65 | |
| 66 | delta = kvm_clock_read(); |
| 67 | |
| 68 | native_write_msr(MSR_KVM_WALL_CLOCK, low, high); |
| 69 | do { |
| 70 | version = wall_clock.wc_version; |
| 71 | rmb(); |
| 72 | wc_sec = wall_clock.wc_sec; |
| 73 | wc_nsec = wall_clock.wc_nsec; |
| 74 | rmb(); |
| 75 | } while ((wall_clock.wc_version != version) || (version & 1)); |
| 76 | |
| 77 | delta = kvm_clock_read() - delta; |
| 78 | delta += wc_nsec; |
| 79 | nsec = do_div(delta, NSEC_PER_SEC); |
| 80 | set_normalized_timespec(&ts, wc_sec + delta, nsec); |
| 81 | /* |
| 82 | * Of all mechanisms of time adjustment I've tested, this one |
| 83 | * was the champion! |
| 84 | */ |
| 85 | return ts.tv_sec + 1; |
| 86 | } |
| 87 | |
| 88 | int kvm_set_wallclock(unsigned long now) |
| 89 | { |
| 90 | return 0; |
| 91 | } |
| 92 | |
| 93 | /* |
| 94 | * This is our read_clock function. The host puts an tsc timestamp each time |
| 95 | * it updates a new time. Without the tsc adjustment, we can have a situation |
| 96 | * in which a vcpu starts to run earlier (smaller system_time), but probes |
| 97 | * time later (compared to another vcpu), leading to backwards time |
| 98 | */ |
| 99 | static cycle_t kvm_clock_read(void) |
| 100 | { |
| 101 | u64 last_tsc, now; |
| 102 | int cpu; |
| 103 | |
| 104 | preempt_disable(); |
| 105 | cpu = smp_processor_id(); |
| 106 | |
| 107 | last_tsc = get_clock(cpu, tsc_timestamp); |
| 108 | now = get_clock(cpu, system_time); |
| 109 | |
| 110 | now += kvm_get_delta(last_tsc); |
| 111 | preempt_enable(); |
| 112 | |
| 113 | return now; |
| 114 | } |
| 115 | static struct clocksource kvm_clock = { |
| 116 | .name = "kvm-clock", |
| 117 | .read = kvm_clock_read, |
| 118 | .rating = 400, |
| 119 | .mask = CLOCKSOURCE_MASK(64), |
| 120 | .mult = 1 << KVM_SCALE, |
| 121 | .shift = KVM_SCALE, |
| 122 | .flags = CLOCK_SOURCE_IS_CONTINUOUS, |
| 123 | }; |
| 124 | |
| 125 | static int kvm_register_clock(void) |
| 126 | { |
| 127 | int cpu = smp_processor_id(); |
| 128 | int low, high; |
| 129 | low = (int)__pa(&per_cpu(hv_clock, cpu)) | 1; |
| 130 | high = ((u64)__pa(&per_cpu(hv_clock, cpu)) >> 32); |
| 131 | |
| 132 | return native_write_msr_safe(MSR_KVM_SYSTEM_TIME, low, high); |
| 133 | } |
| 134 | |
| 135 | static void kvm_setup_secondary_clock(void) |
| 136 | { |
| 137 | /* |
| 138 | * Now that the first cpu already had this clocksource initialized, |
| 139 | * we shouldn't fail. |
| 140 | */ |
| 141 | WARN_ON(kvm_register_clock()); |
| 142 | /* ok, done with our trickery, call native */ |
| 143 | setup_secondary_APIC_clock(); |
| 144 | } |
| 145 | |
| 146 | void __init kvmclock_init(void) |
| 147 | { |
| 148 | if (!kvm_para_available()) |
| 149 | return; |
| 150 | |
| 151 | if (kvmclock && kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE)) { |
| 152 | if (kvm_register_clock()) |
| 153 | return; |
| 154 | pv_time_ops.get_wallclock = kvm_get_wallclock; |
| 155 | pv_time_ops.set_wallclock = kvm_set_wallclock; |
| 156 | pv_time_ops.sched_clock = kvm_clock_read; |
| 157 | pv_apic_ops.setup_secondary_clock = kvm_setup_secondary_clock; |
| 158 | clocksource_register(&kvm_clock); |
| 159 | } |
| 160 | } |