arch/x86/kernel/pvclock.c

   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 <asm/pvclock.h>
  21
  22 /*
  23  * These are perodically updated
  24  *    xen: magic shared_info page
  25  *    kvm: gpa registered via msr
  26  * and then copied here.
  27  */
  28 struct pvclock_shadow_time {
  29         u64 tsc_timestamp;     /* TSC at last update of time vals.  */
  30         u64 system_timestamp;  /* Time, in nanosecs, since boot.    */
  31         u32 tsc_to_nsec_mul;
  32         int tsc_shift;
  33         u32 version;
  34 };
  35
  36 /*
  37  * Scale a 64-bit delta by scaling and multiplying by a 32-bit fraction,
  38  * yielding a 64-bit result.
  39  */
  40 static inline u64 scale_delta(u64 delta, u32 mul_frac, int shift)
  41 {
  42         u64 product;
  43 #ifdef __i386__
  44         u32 tmp1, tmp2;
  45 #endif
  46
  47         if (shift < 0)
  48                 delta >>= -shift;
  49         else
  50                 delta <<= shift;
  51
  52 #ifdef __i386__
  53         __asm__ (
  54                 "mul  %5       ; "
  55                 "mov  %4,%%eax ; "
  56                 "mov  %%edx,%4 ; "
  57                 "mul  %5       ; "
  58                 "xor  %5,%5    ; "
  59                 "add  %4,%%eax ; "
  60                 "adc  %5,%%edx ; "
  61                 : "=A" (product), "=r" (tmp1), "=r" (tmp2)
  62                 : "a" ((u32)delta), "1" ((u32)(delta >> 32)), "2" (mul_frac) );
  63 #elif __x86_64__
  64         __asm__ (
  65                 "mul %%rdx ; shrd $32,%%rdx,%%rax"
  66                 : "=a" (product) : "0" (delta), "d" ((u64)mul_frac) );
  67 #else
  68 #error implement me!
  69 #endif
  70
  71         return product;
  72 }
  73
  74 static u64 pvclock_get_nsec_offset(struct pvclock_shadow_time *shadow)
  75 {
  76         u64 delta = native_read_tsc() - shadow->tsc_timestamp;
  77         return scale_delta(delta, shadow->tsc_to_nsec_mul, shadow->tsc_shift);
  78 }
  79
  80 /*
  81  * Reads a consistent set of time-base values from hypervisor,
  82  * into a shadow data area.
  83  */
  84 static unsigned pvclock_get_time_values(struct pvclock_shadow_time *dst,
  85                                         struct pvclock_vcpu_time_info *src)
  86 {
  87         do {
  88                 dst->version = src->version;
  89                 rmb();          /* fetch version before data */
  90                 dst->tsc_timestamp     = src->tsc_timestamp;
  91                 dst->system_timestamp  = src->system_time;
  92                 dst->tsc_to_nsec_mul   = src->tsc_to_system_mul;
  93                 dst->tsc_shift         = src->tsc_shift;
  94                 rmb();          /* test version after fetching data */
  95         } while ((src->version & 1) || (dst->version != src->version));
  96
  97         return dst->version;
  98 }
  99
 100 unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src)
 101 {
 102         u64 pv_tsc_khz = 1000000ULL << 32;
 103
 104         do_div(pv_tsc_khz, src->tsc_to_system_mul);
 105         if (src->tsc_shift < 0)
 106                 pv_tsc_khz <<= -src->tsc_shift;
 107         else
 108                 pv_tsc_khz >>= src->tsc_shift;
 109         return pv_tsc_khz;
 110 }
 111
 112 cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
 113 {
 114         struct pvclock_shadow_time shadow;
 115         unsigned version;
 116         cycle_t ret, offset;
 117
 118         do {
 119                 version = pvclock_get_time_values(&shadow, src);
 120                 barrier();
 121                 offset = pvclock_get_nsec_offset(&shadow);
 122                 ret = shadow.system_timestamp + offset;
 123                 barrier();
 124         } while (version != src->version);
 125
 126         return ret;
 127 }
 128
 129 void pvclock_read_wallclock(struct pvclock_wall_clock *wall_clock,
 130                             struct pvclock_vcpu_time_info *vcpu_time,
 131                             struct timespec *ts)
 132 {
 133         u32 version;
 134         u64 delta;
 135         struct timespec now;
 136
 137         /* get wallclock at system boot */
 138         do {
 139                 version = wall_clock->version;
 140                 rmb();          /* fetch version before time */
 141                 now.tv_sec  = wall_clock->sec;
 142                 now.tv_nsec = wall_clock->nsec;
 143                 rmb();          /* fetch time before checking version */
 144         } while ((wall_clock->version & 1) || (version != wall_clock->version));
 145
 146         delta = pvclock_clocksource_read(vcpu_time);    /* time since system boot */
 147         delta += now.tv_sec * (u64)NSEC_PER_SEC + now.tv_nsec;
 148
 149         now.tv_nsec = do_div(delta, NSEC_PER_SEC);
 150         now.tv_sec = delta;
 151
 152         set_normalized_timespec(ts, now.tv_sec, now.tv_nsec);
 153 }