1 /* paravirtual clock -- common code used by kvm/xen 2 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. 7 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. 12 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 16 */ 17 18 #include <linux/kernel.h> 19 #include <linux/percpu.h> 20 #include <linux/notifier.h> 21 #include <linux/sched.h> 22 #include <linux/gfp.h> 23 #include <linux/bootmem.h> 24 #include <asm/fixmap.h> 25 #include <asm/pvclock.h> 26 27 static u8 valid_flags __read_mostly = 0; 28 29 void pvclock_set_flags(u8 flags) 30 { 31 valid_flags = flags; 32 } 33 34 unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src) 35 { 36 u64 pv_tsc_khz = 1000000ULL << 32; 37 38 do_div(pv_tsc_khz, src->tsc_to_system_mul); 39 if (src->tsc_shift < 0) 40 pv_tsc_khz <<= -src->tsc_shift; 41 else 42 pv_tsc_khz >>= src->tsc_shift; 43 return pv_tsc_khz; 44 } 45 46 void pvclock_touch_watchdogs(void) 47 { 48 touch_softlockup_watchdog_sync(); 49 clocksource_touch_watchdog(); 50 rcu_cpu_stall_reset(); 51 reset_hung_task_detector(); 52 } 53 54 static atomic64_t last_value = ATOMIC64_INIT(0); 55 56 void pvclock_resume(void) 57 { 58 atomic64_set(&last_value, 0); 59 } 60 61 u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src) 62 { 63 unsigned version; 64 cycle_t ret; 65 u8 flags; 66 67 do { 68 version = __pvclock_read_cycles(src, &ret, &flags); 69 } while ((src->version & 1) || version != src->version); 70 71 return flags & valid_flags; 72 } 73 74 cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src) 75 { 76 unsigned version; 77 cycle_t ret; 78 u64 last; 79 u8 flags; 80 81 do { 82 version = __pvclock_read_cycles(src, &ret, &flags); 83 } while ((src->version & 1) || version != src->version); 84 85 if (unlikely((flags & PVCLOCK_GUEST_STOPPED) != 0)) { 86 src->flags &= ~PVCLOCK_GUEST_STOPPED; 87 pvclock_touch_watchdogs(); 88 } 89 90 if ((valid_flags & PVCLOCK_TSC_STABLE_BIT) && 91 (flags & PVCLOCK_TSC_STABLE_BIT)) 92 return ret; 93 94 /* 95 * Assumption here is that last_value, a global accumulator, always goes 96 * forward. If we are less than that, we should not be much smaller. 97 * We assume there is an error marging we're inside, and then the correction 98 * does not sacrifice accuracy. 99 * 100 * For reads: global may have changed between test and return, 101 * but this means someone else updated poked the clock at a later time. 102 * We just need to make sure we are not seeing a backwards event. 103 * 104 * For updates: last_value = ret is not enough, since two vcpus could be 105 * updating at the same time, and one of them could be slightly behind, 106 * making the assumption that last_value always go forward fail to hold. 107 */ 108 last = atomic64_read(&last_value); 109 do { 110 if (ret < last) 111 return last; 112 last = atomic64_cmpxchg(&last_value, last, ret); 113 } while (unlikely(last != ret)); 114 115 return ret; 116 } 117 118 void pvclock_read_wallclock(struct pvclock_wall_clock *wall_clock, 119 struct pvclock_vcpu_time_info *vcpu_time, 120 struct timespec *ts) 121 { 122 u32 version; 123 u64 delta; 124 struct timespec now; 125 126 /* get wallclock at system boot */ 127 do { 128 version = wall_clock->version; 129 rmb(); /* fetch version before time */ 130 now.tv_sec = wall_clock->sec; 131 now.tv_nsec = wall_clock->nsec; 132 rmb(); /* fetch time before checking version */ 133 } while ((wall_clock->version & 1) || (version != wall_clock->version)); 134 135 delta = pvclock_clocksource_read(vcpu_time); /* time since system boot */ 136 delta += now.tv_sec * (u64)NSEC_PER_SEC + now.tv_nsec; 137 138 now.tv_nsec = do_div(delta, NSEC_PER_SEC); 139 now.tv_sec = delta; 140 141 set_normalized_timespec(ts, now.tv_sec, now.tv_nsec); 142 } 143 144 static struct pvclock_vsyscall_time_info *pvclock_vdso_info; 145 146 static struct pvclock_vsyscall_time_info * 147 pvclock_get_vsyscall_user_time_info(int cpu) 148 { 149 if (!pvclock_vdso_info) { 150 BUG(); 151 return NULL; 152 } 153 154 return &pvclock_vdso_info[cpu]; 155 } 156 157 struct pvclock_vcpu_time_info *pvclock_get_vsyscall_time_info(int cpu) 158 { 159 return &pvclock_get_vsyscall_user_time_info(cpu)->pvti; 160 } 161 162 #ifdef CONFIG_X86_64 163 static int pvclock_task_migrate(struct notifier_block *nb, unsigned long l, 164 void *v) 165 { 166 struct task_migration_notifier *mn = v; 167 struct pvclock_vsyscall_time_info *pvti; 168 169 pvti = pvclock_get_vsyscall_user_time_info(mn->from_cpu); 170 171 /* this is NULL when pvclock vsyscall is not initialized */ 172 if (unlikely(pvti == NULL)) 173 return NOTIFY_DONE; 174 175 pvti->migrate_count++; 176 177 return NOTIFY_DONE; 178 } 179 180 static struct notifier_block pvclock_migrate = { 181 .notifier_call = pvclock_task_migrate, 182 }; 183 184 /* 185 * Initialize the generic pvclock vsyscall state. This will allocate 186 * a/some page(s) for the per-vcpu pvclock information, set up a 187 * fixmap mapping for the page(s) 188 */ 189 190 int __init pvclock_init_vsyscall(struct pvclock_vsyscall_time_info *i, 191 int size) 192 { 193 int idx; 194 195 WARN_ON (size != PVCLOCK_VSYSCALL_NR_PAGES*PAGE_SIZE); 196 197 pvclock_vdso_info = i; 198 199 for (idx = 0; idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) { 200 __set_fixmap(PVCLOCK_FIXMAP_BEGIN + idx, 201 __pa(i) + (idx*PAGE_SIZE), 202 PAGE_KERNEL_VVAR); 203 } 204 205 206 register_task_migration_notifier(&pvclock_migrate); 207 208 return 0; 209 } 210 #endif 211