1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #include <sys/types.h> 28 #include <sys/clock.h> 29 #include <sys/panic.h> 30 #include <sys/atomic.h> 31 #include <sys/hypervisor.h> 32 33 #include <sys/archsystm.h> 34 35 /* 36 * On the hypervisor, we have a virtualized system time based upon the 37 * information provided for each VCPU, which is updated every time it is 38 * scheduled onto a real CPU. Thus, none of the traditional code in 39 * i86pc/os/timestamp.c applies, our gethrtime() implementation is run through 40 * the PSM, and there is no scaling step to apply. 41 * 42 * However, the platform does not guarantee monotonicity; thus we have to fake 43 * this up, which is a deeply unpleasant thing to have to do. 44 * 45 * Note that the virtualized interface still relies on the current TSC to 46 * calculate the time in nanoseconds since the VCPU was scheduled, and is thus 47 * subject to all the problems with that. For the most part, the hypervisor is 48 * supposed to deal with them. 49 * 50 * Another wrinkle involves suspend/resume/migration. If we come back and time 51 * is apparently less, we may have resumed on a different machine or on the 52 * same machine after a reboot. In this case we need to maintain an addend to 53 * ensure time continues reasonably. Otherwise we could end up taking a very 54 * long time to expire cyclics in the heap. Thus we have two functions: 55 * 56 * xpv_getsystime() 57 * 58 * The unadulterated system time from the hypervisor. This is only to be 59 * used when programming the hypervisor (setting a timer or calculating 60 * the TOD). 61 * 62 * xpv_gethrtime() 63 * 64 * This is the monotonic hrtime counter to be used by everything else such 65 * as the cyclic subsystem. We should never pass an hrtime directly into 66 * a hypervisor interface, as hrtime_addend may well be non-zero. 67 */ 68 69 int hrtime_fake_mt = 1; 70 static volatile hrtime_t hrtime_last; 71 static hrtime_t hrtime_suspend_time; 72 static hrtime_t hrtime_addend; 73 74 /* 75 * These functions are used in DTrace probe context, and must be removed from 76 * fbt consideration. Currently fbt ignores all weak symbols, so this will 77 * achieve that. 78 */ 79 #pragma weak xpv_gethrtime = dtrace_xpv_gethrtime 80 #pragma weak xpv_getsystime = dtrace_xpv_getsystime 81 #pragma weak dtrace_gethrtime = dtrace_xpv_gethrtime 82 #pragma weak tsc_read = dtrace_xpv_gethrtime 83 84 hrtime_t 85 dtrace_xpv_getsystime(void) 86 { 87 vcpu_time_info_t *src; 88 vcpu_time_info_t __vti, *dst = &__vti; 89 uint64_t tsc_delta; 90 kthread_t *t = curthread; 91 uint64_t tsc; 92 hrtime_t result; 93 94 /* 95 * This stops us from wandering off the virtual cpu. 96 */ 97 t->t_preempt++; 98 99 src = &CPU->cpu_m.mcpu_vcpu_info->time; 100 101 /* 102 * Loop until version has not been changed during our update, and a Xen 103 * update is not under way (lowest bit is set). 104 */ 105 do { 106 dst->version = src->version; 107 108 membar_consumer(); 109 110 dst->tsc_timestamp = src->tsc_timestamp; 111 dst->system_time = src->system_time; 112 dst->tsc_to_system_mul = src->tsc_to_system_mul; 113 dst->tsc_shift = src->tsc_shift; 114 115 /* 116 * Note that this use of the -actual- TSC register 117 * should probably be the SOLE one in the system on this 118 * paravirtualized platform. 119 */ 120 tsc = __rdtsc_insn(); 121 tsc_delta = tsc - dst->tsc_timestamp; 122 123 membar_consumer(); 124 125 } while ((src->version & 1) | (dst->version ^ src->version)); 126 127 if (dst->tsc_shift >= 0) 128 tsc_delta <<= dst->tsc_shift; 129 else if (dst->tsc_shift < 0) 130 tsc_delta >>= -dst->tsc_shift; 131 132 result = dst->system_time + 133 ((uint64_t)(tsc_delta * (uint64_t)dst->tsc_to_system_mul) >> 32); 134 135 t->t_preempt--; 136 137 return (result); 138 } 139 140 hrtime_t 141 dtrace_xpv_gethrtime(void) 142 { 143 hrtime_t result = xpv_getsystime() + hrtime_addend; 144 145 if (hrtime_fake_mt) { 146 hrtime_t last; 147 do { 148 last = hrtime_last; 149 if (result < last) 150 result = last + 1; 151 } while (atomic_cas_64((volatile uint64_t *)&hrtime_last, 152 last, result) != last); 153 } 154 155 return (result); 156 } 157 158 void 159 xpv_time_suspend(void) 160 { 161 hrtime_suspend_time = xpv_getsystime(); 162 } 163 164 void 165 xpv_time_resume(void) 166 { 167 hrtime_t delta = xpv_getsystime() - hrtime_suspend_time; 168 169 if (delta < 0) 170 hrtime_addend += -delta; 171 } 172