linux/arch/x86/xen/time.c

511 lines
12 KiB
C
Raw Normal View History

/*
* 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>
#include <linux/kernel_stat.h>
#include <linux/math64.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 09:04:11 +01:00
#include <linux/gfp.h>
#include <asm/pvclock.h>
#include <asm/xen/hypervisor.h>
#include <asm/xen/hypercall.h>
#include <xen/events.h>
#include <xen/features.h>
#include <xen/interface/xen.h>
#include <xen/interface/vcpu.h>
#include "xen-ops.h"
/* Xen may fire a timer up to this many ns early */
#define TIMER_SLOP 100000
#define NS_PER_TICK (1000000000LL / HZ)
/* runstate info updated by Xen */
static DEFINE_PER_CPU(struct vcpu_runstate_info, xen_runstate);
/* snapshots of runstate info */
static DEFINE_PER_CPU(struct vcpu_runstate_info, xen_runstate_snapshot);
/* unused ns of stolen and blocked time */
static DEFINE_PER_CPU(u64, xen_residual_stolen);
static DEFINE_PER_CPU(u64, xen_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;
BUG_ON(preemptible());
state = &__get_cpu_var(xen_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);
}
/* return true when a vcpu could run but has no real cpu to run on */
bool xen_vcpu_stolen(int vcpu)
{
return per_cpu(xen_runstate, vcpu).state == RUNSTATE_runnable;
}
void xen_setup_runstate_info(int cpu)
{
struct vcpu_register_runstate_memory_area area;
area.addr.v = &per_cpu(xen_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(xen_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,
2008-12-31 15:11:38 +01:00
including any left-overs from last time. */
stolen = runnable + offline + __this_cpu_read(xen_residual_stolen);
if (stolen < 0)
stolen = 0;
ticks = iter_div_u64_rem(stolen, NS_PER_TICK, &stolen);
__this_cpu_write(xen_residual_stolen, stolen);
2008-12-31 15:11:38 +01:00
account_steal_ticks(ticks);
/* Add the appropriate number of ticks of blocked time,
2008-12-31 15:11:38 +01:00
including any left-overs from last time. */
blocked += __this_cpu_read(xen_residual_blocked);
if (blocked < 0)
blocked = 0;
ticks = iter_div_u64_rem(blocked, NS_PER_TICK, &blocked);
__this_cpu_write(xen_residual_blocked, blocked);
2008-12-31 15:11:38 +01:00
account_idle_ticks(ticks);
}
/* Get the TSC speed from Xen */
static unsigned long xen_tsc_khz(void)
{
struct pvclock_vcpu_time_info *info =
&HYPERVISOR_shared_info->vcpu_info[0].time;
return pvclock_tsc_khz(info);
}
cycle_t xen_clocksource_read(void)
{
struct pvclock_vcpu_time_info *src;
cycle_t ret;
src = &get_cpu_var(xen_vcpu)->time;
ret = pvclock_clocksource_read(src);
put_cpu_var(xen_vcpu);
return ret;
}
static cycle_t xen_clocksource_get_cycles(struct clocksource *cs)
{
return xen_clocksource_read();
}
static void xen_read_wallclock(struct timespec *ts)
{
struct shared_info *s = HYPERVISOR_shared_info;
struct pvclock_wall_clock *wall_clock = &(s->wc);
struct pvclock_vcpu_time_info *vcpu_time;
vcpu_time = &get_cpu_var(xen_vcpu)->time;
pvclock_read_wallclock(wall_clock, vcpu_time, ts);
put_cpu_var(xen_vcpu);
}
static unsigned long xen_get_wallclock(void)
{
struct timespec ts;
xen_read_wallclock(&ts);
return ts.tv_sec;
}
static 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_get_cycles,
.mask = ~0,
.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:
case CLOCK_EVT_MODE_RESUME:
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;
case CLOCK_EVT_MODE_RESUME:
break;
}
}
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;
}
do_stolen_accounting();
return ret;
}
void xen_setup_timer(int cpu)
{
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|IRQF_TIMER|
IRQF_FORCE_RESUME,
name, NULL);
evt = &per_cpu(xen_clock_events, cpu);
memcpy(evt, xen_clockevent, sizeof(*evt));
evt->cpumask = cpumask_of(cpu);
evt->irq = irq;
}
void xen_teardown_timer(int cpu)
{
struct clock_event_device *evt;
BUG_ON(cpu == 0);
evt = &per_cpu(xen_clock_events, cpu);
unbind_from_irqhandler(evt->irq, NULL);
}
void xen_setup_cpu_clockevents(void)
{
BUG_ON(preemptible());
clockevents_register_device(&__get_cpu_var(xen_clock_events));
}
void xen_timer_resume(void)
{
int cpu;
pvclock_resume();
if (xen_clockevent != &xen_vcpuop_clockevent)
return;
for_each_online_cpu(cpu) {
if (HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL))
BUG();
}
}
static const struct pv_time_ops xen_time_ops __initconst = {
.sched_clock = xen_clocksource_read,
};
static void __init xen_time_init(void)
{
int cpu = smp_processor_id();
struct timespec tp;
clocksource_register_hz(&xen_clocksource, NSEC_PER_SEC);
if (HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL) == 0) {
/* Successfully turned off 100Hz tick, so we have the
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(&tp);
do_settimeofday(&tp);
setup_force_cpu_cap(X86_FEATURE_TSC);
xen_setup_runstate_info(cpu);
xen_setup_timer(cpu);
xen_setup_cpu_clockevents();
}
void __init xen_init_time_ops(void)
{
pv_time_ops = xen_time_ops;
x86_init.timers.timer_init = xen_time_init;
x86_init.timers.setup_percpu_clockev = x86_init_noop;
x86_cpuinit.setup_percpu_clockev = x86_init_noop;
x86_platform.calibrate_tsc = xen_tsc_khz;
x86_platform.get_wallclock = xen_get_wallclock;
x86_platform.set_wallclock = xen_set_wallclock;
}
#ifdef CONFIG_XEN_PVHVM
static void xen_hvm_setup_cpu_clockevents(void)
{
int cpu = smp_processor_id();
xen_setup_runstate_info(cpu);
xen_setup_timer(cpu);
xen_setup_cpu_clockevents();
}
void __init xen_hvm_init_time_ops(void)
{
/* vector callback is needed otherwise we cannot receive interrupts
* on cpu > 0 and at this point we don't know how many cpus are
* available */
if (!xen_have_vector_callback)
return;
if (!xen_feature(XENFEAT_hvm_safe_pvclock)) {
printk(KERN_INFO "Xen doesn't support pvclock on HVM,"
"disable pv timer\n");
return;
}
pv_time_ops = xen_time_ops;
x86_init.timers.setup_percpu_clockev = xen_time_init;
x86_cpuinit.setup_percpu_clockev = xen_hvm_setup_cpu_clockevents;
x86_platform.calibrate_tsc = xen_tsc_khz;
x86_platform.get_wallclock = xen_get_wallclock;
x86_platform.set_wallclock = xen_set_wallclock;
}
#endif