pvclock.c 4.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167
  1. /* paravirtual clock -- common code used by kvm/xen
  2. This program is free software; you can redistribute it and/or modify
  3. it under the terms of the GNU General Public License as published by
  4. the Free Software Foundation; either version 2 of the License, or
  5. (at your option) any later version.
  6. This program is distributed in the hope that it will be useful,
  7. but WITHOUT ANY WARRANTY; without even the implied warranty of
  8. MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  9. GNU General Public License for more details.
  10. You should have received a copy of the GNU General Public License
  11. along with this program; if not, write to the Free Software
  12. Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  13. */
  14. #include <linux/kernel.h>
  15. #include <linux/percpu.h>
  16. #include <linux/notifier.h>
  17. #include <linux/sched.h>
  18. #include <linux/gfp.h>
  19. #include <linux/bootmem.h>
  20. #include <asm/fixmap.h>
  21. #include <asm/pvclock.h>
  22. static u8 valid_flags __read_mostly = 0;
  23. void pvclock_set_flags(u8 flags)
  24. {
  25. valid_flags = flags;
  26. }
  27. unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src)
  28. {
  29. u64 pv_tsc_khz = 1000000ULL << 32;
  30. do_div(pv_tsc_khz, src->tsc_to_system_mul);
  31. if (src->tsc_shift < 0)
  32. pv_tsc_khz <<= -src->tsc_shift;
  33. else
  34. pv_tsc_khz >>= src->tsc_shift;
  35. return pv_tsc_khz;
  36. }
  37. void pvclock_touch_watchdogs(void)
  38. {
  39. touch_softlockup_watchdog_sync();
  40. clocksource_touch_watchdog();
  41. rcu_cpu_stall_reset();
  42. reset_hung_task_detector();
  43. }
  44. static atomic64_t last_value = ATOMIC64_INIT(0);
  45. void pvclock_resume(void)
  46. {
  47. atomic64_set(&last_value, 0);
  48. }
  49. u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src)
  50. {
  51. unsigned version;
  52. cycle_t ret;
  53. u8 flags;
  54. do {
  55. version = __pvclock_read_cycles(src, &ret, &flags);
  56. } while ((src->version & 1) || version != src->version);
  57. return flags & valid_flags;
  58. }
  59. cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
  60. {
  61. unsigned version;
  62. cycle_t ret;
  63. u64 last;
  64. u8 flags;
  65. do {
  66. version = __pvclock_read_cycles(src, &ret, &flags);
  67. } while ((src->version & 1) || version != src->version);
  68. if (unlikely((flags & PVCLOCK_GUEST_STOPPED) != 0)) {
  69. src->flags &= ~PVCLOCK_GUEST_STOPPED;
  70. pvclock_touch_watchdogs();
  71. }
  72. if ((valid_flags & PVCLOCK_TSC_STABLE_BIT) &&
  73. (flags & PVCLOCK_TSC_STABLE_BIT))
  74. return ret;
  75. /*
  76. * Assumption here is that last_value, a global accumulator, always goes
  77. * forward. If we are less than that, we should not be much smaller.
  78. * We assume there is an error marging we're inside, and then the correction
  79. * does not sacrifice accuracy.
  80. *
  81. * For reads: global may have changed between test and return,
  82. * but this means someone else updated poked the clock at a later time.
  83. * We just need to make sure we are not seeing a backwards event.
  84. *
  85. * For updates: last_value = ret is not enough, since two vcpus could be
  86. * updating at the same time, and one of them could be slightly behind,
  87. * making the assumption that last_value always go forward fail to hold.
  88. */
  89. last = atomic64_read(&last_value);
  90. do {
  91. if (ret < last)
  92. return last;
  93. last = atomic64_cmpxchg(&last_value, last, ret);
  94. } while (unlikely(last != ret));
  95. return ret;
  96. }
  97. void pvclock_read_wallclock(struct pvclock_wall_clock *wall_clock,
  98. struct pvclock_vcpu_time_info *vcpu_time,
  99. struct timespec *ts)
  100. {
  101. u32 version;
  102. u64 delta;
  103. struct timespec now;
  104. /* get wallclock at system boot */
  105. do {
  106. version = wall_clock->version;
  107. rmb(); /* fetch version before time */
  108. now.tv_sec = wall_clock->sec;
  109. now.tv_nsec = wall_clock->nsec;
  110. rmb(); /* fetch time before checking version */
  111. } while ((wall_clock->version & 1) || (version != wall_clock->version));
  112. delta = pvclock_clocksource_read(vcpu_time); /* time since system boot */
  113. delta += now.tv_sec * (u64)NSEC_PER_SEC + now.tv_nsec;
  114. now.tv_nsec = do_div(delta, NSEC_PER_SEC);
  115. now.tv_sec = delta;
  116. set_normalized_timespec(ts, now.tv_sec, now.tv_nsec);
  117. }
  118. #ifdef CONFIG_X86_64
  119. /*
  120. * Initialize the generic pvclock vsyscall state. This will allocate
  121. * a/some page(s) for the per-vcpu pvclock information, set up a
  122. * fixmap mapping for the page(s)
  123. */
  124. int __init pvclock_init_vsyscall(struct pvclock_vsyscall_time_info *i,
  125. int size)
  126. {
  127. int idx;
  128. WARN_ON (size != PVCLOCK_VSYSCALL_NR_PAGES*PAGE_SIZE);
  129. for (idx = 0; idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) {
  130. __set_fixmap(PVCLOCK_FIXMAP_BEGIN + idx,
  131. __pa(i) + (idx*PAGE_SIZE),
  132. PAGE_KERNEL_VVAR);
  133. }
  134. return 0;
  135. }
  136. #endif