xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
/*
|
|
|
|
* Xen time implementation.
|
|
|
|
*
|
|
|
|
* This is implemented in terms of a clocksource driver which uses
|
|
|
|
* the hypervisor clock as a nanosecond timebase, and a clockevent
|
|
|
|
* driver which uses the hypervisor's timer mechanism.
|
|
|
|
*
|
|
|
|
* Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007
|
|
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/clocksource.h>
|
|
|
|
#include <linux/clockchips.h>
|
2007-07-18 01:37:05 +00:00
|
|
|
#include <linux/kernel_stat.h>
|
2008-06-12 08:47:56 +00:00
|
|
|
#include <linux/math64.h>
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
|
2008-06-03 14:17:30 +00:00
|
|
|
#include <asm/pvclock.h>
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
#include <asm/xen/hypervisor.h>
|
|
|
|
#include <asm/xen/hypercall.h>
|
|
|
|
|
|
|
|
#include <xen/events.h>
|
|
|
|
#include <xen/interface/xen.h>
|
|
|
|
#include <xen/interface/vcpu.h>
|
|
|
|
|
|
|
|
#include "xen-ops.h"
|
|
|
|
|
|
|
|
#define XEN_SHIFT 22
|
|
|
|
|
|
|
|
/* Xen may fire a timer up to this many ns early */
|
|
|
|
#define TIMER_SLOP 100000
|
2007-07-18 01:37:05 +00:00
|
|
|
#define NS_PER_TICK (1000000000LL / HZ)
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
|
2007-07-18 01:37:05 +00:00
|
|
|
/* runstate info updated by Xen */
|
|
|
|
static DEFINE_PER_CPU(struct vcpu_runstate_info, runstate);
|
|
|
|
|
|
|
|
/* snapshots of runstate info */
|
|
|
|
static DEFINE_PER_CPU(struct vcpu_runstate_info, runstate_snapshot);
|
|
|
|
|
|
|
|
/* unused ns of stolen and blocked time */
|
|
|
|
static DEFINE_PER_CPU(u64, residual_stolen);
|
|
|
|
static DEFINE_PER_CPU(u64, residual_blocked);
|
|
|
|
|
|
|
|
/* return an consistent snapshot of 64-bit time/counter value */
|
|
|
|
static u64 get64(const u64 *p)
|
|
|
|
{
|
|
|
|
u64 ret;
|
|
|
|
|
|
|
|
if (BITS_PER_LONG < 64) {
|
|
|
|
u32 *p32 = (u32 *)p;
|
|
|
|
u32 h, l;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read high then low, and then make sure high is
|
|
|
|
* still the same; this will only loop if low wraps
|
|
|
|
* and carries into high.
|
|
|
|
* XXX some clean way to make this endian-proof?
|
|
|
|
*/
|
|
|
|
do {
|
|
|
|
h = p32[1];
|
|
|
|
barrier();
|
|
|
|
l = p32[0];
|
|
|
|
barrier();
|
|
|
|
} while (p32[1] != h);
|
|
|
|
|
|
|
|
ret = (((u64)h) << 32) | l;
|
|
|
|
} else
|
|
|
|
ret = *p;
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Runstate accounting
|
|
|
|
*/
|
|
|
|
static void get_runstate_snapshot(struct vcpu_runstate_info *res)
|
|
|
|
{
|
|
|
|
u64 state_time;
|
|
|
|
struct vcpu_runstate_info *state;
|
|
|
|
|
2007-07-18 01:37:06 +00:00
|
|
|
BUG_ON(preemptible());
|
2007-07-18 01:37:05 +00:00
|
|
|
|
|
|
|
state = &__get_cpu_var(runstate);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The runstate info is always updated by the hypervisor on
|
|
|
|
* the current CPU, so there's no need to use anything
|
|
|
|
* stronger than a compiler barrier when fetching it.
|
|
|
|
*/
|
|
|
|
do {
|
|
|
|
state_time = get64(&state->state_entry_time);
|
|
|
|
barrier();
|
|
|
|
*res = *state;
|
|
|
|
barrier();
|
|
|
|
} while (get64(&state->state_entry_time) != state_time);
|
|
|
|
}
|
|
|
|
|
2007-10-16 18:51:30 +00:00
|
|
|
/* return true when a vcpu could run but has no real cpu to run on */
|
|
|
|
bool xen_vcpu_stolen(int vcpu)
|
|
|
|
{
|
|
|
|
return per_cpu(runstate, vcpu).state == RUNSTATE_runnable;
|
|
|
|
}
|
|
|
|
|
2007-07-18 01:37:05 +00:00
|
|
|
static void setup_runstate_info(int cpu)
|
|
|
|
{
|
|
|
|
struct vcpu_register_runstate_memory_area area;
|
|
|
|
|
|
|
|
area.addr.v = &per_cpu(runstate, cpu);
|
|
|
|
|
|
|
|
if (HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area,
|
|
|
|
cpu, &area))
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
static void do_stolen_accounting(void)
|
|
|
|
{
|
|
|
|
struct vcpu_runstate_info state;
|
|
|
|
struct vcpu_runstate_info *snap;
|
|
|
|
s64 blocked, runnable, offline, stolen;
|
|
|
|
cputime_t ticks;
|
|
|
|
|
|
|
|
get_runstate_snapshot(&state);
|
|
|
|
|
|
|
|
WARN_ON(state.state != RUNSTATE_running);
|
|
|
|
|
|
|
|
snap = &__get_cpu_var(runstate_snapshot);
|
|
|
|
|
|
|
|
/* work out how much time the VCPU has not been runn*ing* */
|
|
|
|
blocked = state.time[RUNSTATE_blocked] - snap->time[RUNSTATE_blocked];
|
|
|
|
runnable = state.time[RUNSTATE_runnable] - snap->time[RUNSTATE_runnable];
|
|
|
|
offline = state.time[RUNSTATE_offline] - snap->time[RUNSTATE_offline];
|
|
|
|
|
|
|
|
*snap = state;
|
|
|
|
|
|
|
|
/* Add the appropriate number of ticks of stolen time,
|
|
|
|
including any left-overs from last time. Passing NULL to
|
|
|
|
account_steal_time accounts the time as stolen. */
|
|
|
|
stolen = runnable + offline + __get_cpu_var(residual_stolen);
|
|
|
|
|
|
|
|
if (stolen < 0)
|
|
|
|
stolen = 0;
|
|
|
|
|
2008-06-12 08:47:56 +00:00
|
|
|
ticks = iter_div_u64_rem(stolen, NS_PER_TICK, &stolen);
|
2007-07-18 01:37:05 +00:00
|
|
|
__get_cpu_var(residual_stolen) = stolen;
|
|
|
|
account_steal_time(NULL, ticks);
|
|
|
|
|
|
|
|
/* Add the appropriate number of ticks of blocked time,
|
|
|
|
including any left-overs from last time. Passing idle to
|
|
|
|
account_steal_time accounts the time as idle/wait. */
|
|
|
|
blocked += __get_cpu_var(residual_blocked);
|
|
|
|
|
|
|
|
if (blocked < 0)
|
|
|
|
blocked = 0;
|
|
|
|
|
2008-06-12 08:47:56 +00:00
|
|
|
ticks = iter_div_u64_rem(blocked, NS_PER_TICK, &blocked);
|
2007-07-18 01:37:05 +00:00
|
|
|
__get_cpu_var(residual_blocked) = blocked;
|
|
|
|
account_steal_time(idle_task(smp_processor_id()), ticks);
|
|
|
|
}
|
|
|
|
|
2007-07-18 01:37:05 +00:00
|
|
|
/*
|
|
|
|
* Xen sched_clock implementation. Returns the number of unstolen
|
|
|
|
* nanoseconds, which is nanoseconds the VCPU spent in RUNNING+BLOCKED
|
|
|
|
* states.
|
|
|
|
*/
|
|
|
|
unsigned long long xen_sched_clock(void)
|
|
|
|
{
|
|
|
|
struct vcpu_runstate_info state;
|
2007-07-18 01:37:06 +00:00
|
|
|
cycle_t now;
|
|
|
|
u64 ret;
|
2007-07-18 01:37:05 +00:00
|
|
|
s64 offset;
|
|
|
|
|
2007-07-18 01:37:06 +00:00
|
|
|
/*
|
|
|
|
* Ideally sched_clock should be called on a per-cpu basis
|
|
|
|
* anyway, so preempt should already be disabled, but that's
|
|
|
|
* not current practice at the moment.
|
|
|
|
*/
|
|
|
|
preempt_disable();
|
|
|
|
|
|
|
|
now = xen_clocksource_read();
|
|
|
|
|
2007-07-18 01:37:05 +00:00
|
|
|
get_runstate_snapshot(&state);
|
|
|
|
|
|
|
|
WARN_ON(state.state != RUNSTATE_running);
|
|
|
|
|
|
|
|
offset = now - state.state_entry_time;
|
|
|
|
if (offset < 0)
|
|
|
|
offset = 0;
|
|
|
|
|
2007-07-18 01:37:06 +00:00
|
|
|
ret = state.time[RUNSTATE_blocked] +
|
2007-07-18 01:37:05 +00:00
|
|
|
state.time[RUNSTATE_running] +
|
|
|
|
offset;
|
2007-07-18 01:37:06 +00:00
|
|
|
|
|
|
|
preempt_enable();
|
|
|
|
|
|
|
|
return ret;
|
2007-07-18 01:37:05 +00:00
|
|
|
}
|
2007-07-18 01:37:05 +00:00
|
|
|
|
|
|
|
|
2008-07-01 18:43:36 +00:00
|
|
|
/* Get the TSC speed from Xen */
|
|
|
|
unsigned long xen_tsc_khz(void)
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
{
|
2008-02-09 22:24:08 +00:00
|
|
|
u64 xen_khz = 1000000ULL << 32;
|
2008-06-03 14:17:30 +00:00
|
|
|
const struct pvclock_vcpu_time_info *info =
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
&HYPERVISOR_shared_info->vcpu_info[0].time;
|
|
|
|
|
2008-02-09 22:24:08 +00:00
|
|
|
do_div(xen_khz, info->tsc_to_system_mul);
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
if (info->tsc_shift < 0)
|
2008-02-09 22:24:08 +00:00
|
|
|
xen_khz <<= -info->tsc_shift;
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
else
|
2008-02-09 22:24:08 +00:00
|
|
|
xen_khz >>= info->tsc_shift;
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
|
2008-02-09 22:24:08 +00:00
|
|
|
return xen_khz;
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
}
|
|
|
|
|
2008-08-21 20:17:56 +00:00
|
|
|
cycle_t xen_clocksource_read(void)
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
{
|
2008-06-03 14:17:30 +00:00
|
|
|
struct pvclock_vcpu_time_info *src;
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
cycle_t ret;
|
|
|
|
|
2008-06-03 14:17:30 +00:00
|
|
|
src = &get_cpu_var(xen_vcpu)->time;
|
|
|
|
ret = pvclock_clocksource_read(src);
|
|
|
|
put_cpu_var(xen_vcpu);
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void xen_read_wallclock(struct timespec *ts)
|
|
|
|
{
|
2008-06-03 14:17:30 +00:00
|
|
|
struct shared_info *s = HYPERVISOR_shared_info;
|
|
|
|
struct pvclock_wall_clock *wall_clock = &(s->wc);
|
|
|
|
struct pvclock_vcpu_time_info *vcpu_time;
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
|
2008-06-03 14:17:30 +00:00
|
|
|
vcpu_time = &get_cpu_var(xen_vcpu)->time;
|
|
|
|
pvclock_read_wallclock(wall_clock, vcpu_time, ts);
|
|
|
|
put_cpu_var(xen_vcpu);
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
unsigned long xen_get_wallclock(void)
|
|
|
|
{
|
|
|
|
struct timespec ts;
|
|
|
|
|
|
|
|
xen_read_wallclock(&ts);
|
|
|
|
return ts.tv_sec;
|
|
|
|
}
|
|
|
|
|
|
|
|
int xen_set_wallclock(unsigned long now)
|
|
|
|
{
|
|
|
|
/* do nothing for domU */
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct clocksource xen_clocksource __read_mostly = {
|
|
|
|
.name = "xen",
|
|
|
|
.rating = 400,
|
|
|
|
.read = xen_clocksource_read,
|
|
|
|
.mask = ~0,
|
|
|
|
.mult = 1<<XEN_SHIFT, /* time directly in nanoseconds */
|
|
|
|
.shift = XEN_SHIFT,
|
|
|
|
.flags = CLOCK_SOURCE_IS_CONTINUOUS,
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
Xen clockevent implementation
|
|
|
|
|
|
|
|
Xen has two clockevent implementations:
|
|
|
|
|
|
|
|
The old timer_op one works with all released versions of Xen prior
|
|
|
|
to version 3.0.4. This version of the hypervisor provides a
|
|
|
|
single-shot timer with nanosecond resolution. However, sharing the
|
|
|
|
same event channel is a 100Hz tick which is delivered while the
|
|
|
|
vcpu is running. We don't care about or use this tick, but it will
|
|
|
|
cause the core time code to think the timer fired too soon, and
|
|
|
|
will end up resetting it each time. It could be filtered, but
|
|
|
|
doing so has complications when the ktime clocksource is not yet
|
|
|
|
the xen clocksource (ie, at boot time).
|
|
|
|
|
|
|
|
The new vcpu_op-based timer interface allows the tick timer period
|
|
|
|
to be changed or turned off. The tick timer is not useful as a
|
|
|
|
periodic timer because events are only delivered to running vcpus.
|
|
|
|
The one-shot timer can report when a timeout is in the past, so
|
|
|
|
set_next_event is capable of returning -ETIME when appropriate.
|
|
|
|
This interface is used when available.
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
Get a hypervisor absolute time. In theory we could maintain an
|
|
|
|
offset between the kernel's time and the hypervisor's time, and
|
|
|
|
apply that to a kernel's absolute timeout. Unfortunately the
|
|
|
|
hypervisor and kernel times can drift even if the kernel is using
|
|
|
|
the Xen clocksource, because ntp can warp the kernel's clocksource.
|
|
|
|
*/
|
|
|
|
static s64 get_abs_timeout(unsigned long delta)
|
|
|
|
{
|
|
|
|
return xen_clocksource_read() + delta;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void xen_timerop_set_mode(enum clock_event_mode mode,
|
|
|
|
struct clock_event_device *evt)
|
|
|
|
{
|
|
|
|
switch (mode) {
|
|
|
|
case CLOCK_EVT_MODE_PERIODIC:
|
|
|
|
/* unsupported */
|
|
|
|
WARN_ON(1);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case CLOCK_EVT_MODE_ONESHOT:
|
2007-07-21 11:37:34 +00:00
|
|
|
case CLOCK_EVT_MODE_RESUME:
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case CLOCK_EVT_MODE_UNUSED:
|
|
|
|
case CLOCK_EVT_MODE_SHUTDOWN:
|
|
|
|
HYPERVISOR_set_timer_op(0); /* cancel timeout */
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int xen_timerop_set_next_event(unsigned long delta,
|
|
|
|
struct clock_event_device *evt)
|
|
|
|
{
|
|
|
|
WARN_ON(evt->mode != CLOCK_EVT_MODE_ONESHOT);
|
|
|
|
|
|
|
|
if (HYPERVISOR_set_timer_op(get_abs_timeout(delta)) < 0)
|
|
|
|
BUG();
|
|
|
|
|
|
|
|
/* We may have missed the deadline, but there's no real way of
|
|
|
|
knowing for sure. If the event was in the past, then we'll
|
|
|
|
get an immediate interrupt. */
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct clock_event_device xen_timerop_clockevent = {
|
|
|
|
.name = "xen",
|
|
|
|
.features = CLOCK_EVT_FEAT_ONESHOT,
|
|
|
|
|
|
|
|
.max_delta_ns = 0xffffffff,
|
|
|
|
.min_delta_ns = TIMER_SLOP,
|
|
|
|
|
|
|
|
.mult = 1,
|
|
|
|
.shift = 0,
|
|
|
|
.rating = 500,
|
|
|
|
|
|
|
|
.set_mode = xen_timerop_set_mode,
|
|
|
|
.set_next_event = xen_timerop_set_next_event,
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
static void xen_vcpuop_set_mode(enum clock_event_mode mode,
|
|
|
|
struct clock_event_device *evt)
|
|
|
|
{
|
|
|
|
int cpu = smp_processor_id();
|
|
|
|
|
|
|
|
switch (mode) {
|
|
|
|
case CLOCK_EVT_MODE_PERIODIC:
|
|
|
|
WARN_ON(1); /* unsupported */
|
|
|
|
break;
|
|
|
|
|
|
|
|
case CLOCK_EVT_MODE_ONESHOT:
|
|
|
|
if (HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL))
|
|
|
|
BUG();
|
|
|
|
break;
|
|
|
|
|
|
|
|
case CLOCK_EVT_MODE_UNUSED:
|
|
|
|
case CLOCK_EVT_MODE_SHUTDOWN:
|
|
|
|
if (HYPERVISOR_vcpu_op(VCPUOP_stop_singleshot_timer, cpu, NULL) ||
|
|
|
|
HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL))
|
|
|
|
BUG();
|
|
|
|
break;
|
2007-07-21 11:37:34 +00:00
|
|
|
case CLOCK_EVT_MODE_RESUME:
|
|
|
|
break;
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int xen_vcpuop_set_next_event(unsigned long delta,
|
|
|
|
struct clock_event_device *evt)
|
|
|
|
{
|
|
|
|
int cpu = smp_processor_id();
|
|
|
|
struct vcpu_set_singleshot_timer single;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
WARN_ON(evt->mode != CLOCK_EVT_MODE_ONESHOT);
|
|
|
|
|
|
|
|
single.timeout_abs_ns = get_abs_timeout(delta);
|
|
|
|
single.flags = VCPU_SSHOTTMR_future;
|
|
|
|
|
|
|
|
ret = HYPERVISOR_vcpu_op(VCPUOP_set_singleshot_timer, cpu, &single);
|
|
|
|
|
|
|
|
BUG_ON(ret != 0 && ret != -ETIME);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct clock_event_device xen_vcpuop_clockevent = {
|
|
|
|
.name = "xen",
|
|
|
|
.features = CLOCK_EVT_FEAT_ONESHOT,
|
|
|
|
|
|
|
|
.max_delta_ns = 0xffffffff,
|
|
|
|
.min_delta_ns = TIMER_SLOP,
|
|
|
|
|
|
|
|
.mult = 1,
|
|
|
|
.shift = 0,
|
|
|
|
.rating = 500,
|
|
|
|
|
|
|
|
.set_mode = xen_vcpuop_set_mode,
|
|
|
|
.set_next_event = xen_vcpuop_set_next_event,
|
|
|
|
};
|
|
|
|
|
|
|
|
static const struct clock_event_device *xen_clockevent =
|
|
|
|
&xen_timerop_clockevent;
|
|
|
|
static DEFINE_PER_CPU(struct clock_event_device, xen_clock_events);
|
|
|
|
|
|
|
|
static irqreturn_t xen_timer_interrupt(int irq, void *dev_id)
|
|
|
|
{
|
|
|
|
struct clock_event_device *evt = &__get_cpu_var(xen_clock_events);
|
|
|
|
irqreturn_t ret;
|
|
|
|
|
|
|
|
ret = IRQ_NONE;
|
|
|
|
if (evt->event_handler) {
|
|
|
|
evt->event_handler(evt);
|
|
|
|
ret = IRQ_HANDLED;
|
|
|
|
}
|
|
|
|
|
2007-07-18 01:37:05 +00:00
|
|
|
do_stolen_accounting();
|
|
|
|
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2007-07-18 01:37:06 +00:00
|
|
|
void xen_setup_timer(int cpu)
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
{
|
|
|
|
const char *name;
|
|
|
|
struct clock_event_device *evt;
|
|
|
|
int irq;
|
|
|
|
|
|
|
|
printk(KERN_INFO "installing Xen timer for CPU %d\n", cpu);
|
|
|
|
|
|
|
|
name = kasprintf(GFP_KERNEL, "timer%d", cpu);
|
|
|
|
if (!name)
|
|
|
|
name = "<timer kasprintf failed>";
|
|
|
|
|
|
|
|
irq = bind_virq_to_irqhandler(VIRQ_TIMER, cpu, xen_timer_interrupt,
|
|
|
|
IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
|
|
|
|
name, NULL);
|
|
|
|
|
2007-07-18 01:37:06 +00:00
|
|
|
evt = &per_cpu(xen_clock_events, cpu);
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
memcpy(evt, xen_clockevent, sizeof(*evt));
|
|
|
|
|
|
|
|
evt->cpumask = cpumask_of_cpu(cpu);
|
|
|
|
evt->irq = irq;
|
|
|
|
|
2007-07-18 01:37:05 +00:00
|
|
|
setup_runstate_info(cpu);
|
2007-07-18 01:37:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void xen_setup_cpu_clockevents(void)
|
|
|
|
{
|
|
|
|
BUG_ON(preemptible());
|
2007-07-18 01:37:05 +00:00
|
|
|
|
2007-07-18 01:37:06 +00:00
|
|
|
clockevents_register_device(&__get_cpu_var(xen_clock_events));
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
}
|
|
|
|
|
2008-05-31 00:33:03 +00:00
|
|
|
void xen_timer_resume(void)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
if (xen_clockevent != &xen_vcpuop_clockevent)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for_each_online_cpu(cpu) {
|
|
|
|
if (HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL))
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
__init void xen_time_init(void)
|
|
|
|
{
|
|
|
|
int cpu = smp_processor_id();
|
|
|
|
|
|
|
|
clocksource_register(&xen_clocksource);
|
|
|
|
|
|
|
|
if (HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL) == 0) {
|
2007-07-18 01:37:05 +00:00
|
|
|
/* Successfully turned off 100Hz tick, so we have the
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
vcpuop-based timer interface */
|
|
|
|
printk(KERN_DEBUG "Xen: using vcpuop timer interface\n");
|
|
|
|
xen_clockevent = &xen_vcpuop_clockevent;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set initial system time with full resolution */
|
|
|
|
xen_read_wallclock(&xtime);
|
|
|
|
set_normalized_timespec(&wall_to_monotonic,
|
|
|
|
-xtime.tv_sec, -xtime.tv_nsec);
|
|
|
|
|
2008-01-30 12:33:20 +00:00
|
|
|
setup_force_cpu_cap(X86_FEATURE_TSC);
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
|
|
|
|
xen_setup_timer(cpu);
|
2007-07-18 01:37:06 +00:00
|
|
|
xen_setup_cpu_clockevents();
|
xen: time implementation
Xen maintains a base clock which measures nanoseconds since system
boot. This is provided to guests via a shared page which contains a
base time in ns, a tsc timestamp at that point and tsc frequency
parameters. Guests can compute the current time by reading the tsc
and using it to extrapolate the current time from the basetime. The
hypervisor makes sure that the frequency parameters are updated
regularly, paricularly if the tsc changes rate or stops.
This is implemented as a clocksource, so the interface to the rest of
the kernel is a simple clocksource which simply returns the current
time directly in nanoseconds.
Xen also provides a simple timer mechanism, which allows a timeout to
be set in the future. When that time arrives, a timer event is sent
to the guest. There are two timer interfaces:
- An old one which also delivers a stream of (unused) ticks at 100Hz,
and on the same event, the actual timer events. The 100Hz ticks
cause a lot of spurious wakeups, but are basically harmless.
- The new timer interface doesn't have the 100Hz ticks, and can also
fail if the specified time is in the past.
This code presents the Xen timer as a clockevent driver, and uses the
new interface by preference.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2007-07-18 01:37:05 +00:00
|
|
|
}
|