* [PATCH 24/28] ia64/xen: time paravirtualization
2008-02-21 23:11 [PATCH 24/28] ia64/xen: time paravirtualization Jeremy Fitzhardinge
2008-02-22 3:57 ` Isaku Yamahata
@ 2008-02-22 5:10 ` Isaku Yamahata
1 sibling, 0 replies; 3+ messages in thread
From: Isaku Yamahata @ 2008-02-22 5:10 UTC (permalink / raw)
To: linux-ia64
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
---
arch/ia64/kernel/time.c | 204 +++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 204 insertions(+), 0 deletions(-)
diff --git a/arch/ia64/kernel/time.c b/arch/ia64/kernel/time.c
index 17fda52..1bb0362 100644
--- a/arch/ia64/kernel/time.c
+++ b/arch/ia64/kernel/time.c
@@ -29,6 +29,14 @@
#include <asm/sections.h>
#include <asm/system.h>
+#include <asm/xen/hypervisor.h>
+#ifdef CONFIG_XEN
+#include <linux/kernel_stat.h>
+#include <linux/posix-timers.h>
+#include <xen/interface/vcpu.h>
+#include <asm/percpu.h>
+#endif
+
#include "fsyscall_gtod_data.h"
static cycle_t itc_get_cycles(void);
@@ -38,6 +46,17 @@ struct fsyscall_gtod_data_t fsyscall_gtod_data = {
};
struct itc_jitter_data_t itc_jitter_data;
+#ifdef CONFIG_XEN
+static void itc_jitter_data_reset(void)
+{
+ u64 lcycle, ret;
+
+ do {
+ lcycle = itc_jitter_data.itc_lastcycle;
+ ret = cmpxchg(&itc_jitter_data.itc_lastcycle, lcycle, 0);
+ } while (unlikely(ret != lcycle));
+}
+#endif
volatile int time_keeper_id = 0; /* smp_processor_id() of time-keeper */
@@ -56,13 +75,105 @@ static struct clocksource clocksource_itc = {
.mult = 0, /*to be calculated*/
.shift = 16,
.flags = CLOCK_SOURCE_IS_CONTINUOUS,
+#ifdef CONFIG_XEN
+ .resume = itc_jitter_data_reset,
+#endif
};
static struct clocksource *itc_clocksource;
+#ifdef CONFIG_XEN
+DEFINE_PER_CPU(struct vcpu_runstate_info, runstate);
+DEFINE_PER_CPU(unsigned long, processed_stolen_time);
+DEFINE_PER_CPU(unsigned long, processed_blocked_time);
+#define NS_PER_TICK (1000000000LL/HZ)
+static unsigned long
+consider_steal_time(unsigned long new_itm)
+{
+ unsigned long stolen, blocked, sched_time;
+ unsigned long delta_itm = 0, stolentick = 0;
+ int cpu = smp_processor_id();
+ struct vcpu_runstate_info *runstate;
+ struct task_struct *p = current;
+
+ runstate = &per_cpu(runstate, smp_processor_id());
+
+ do {
+ sched_time = runstate->state_entry_time;
+ mb();
+ stolen = runstate->time[RUNSTATE_runnable] +
+ runstate->time[RUNSTATE_offline] -
+ per_cpu(processed_stolen_time, cpu);
+ blocked = runstate->time[RUNSTATE_blocked] -
+ per_cpu(processed_blocked_time, cpu);
+ mb();
+ } while (sched_time != runstate->state_entry_time);
+
+ /*
+ * Check for vcpu migration effect
+ * In this case, itc value is reversed.
+ * This causes huge stolen value.
+ * This function just checks and reject this effect.
+ */
+ if (!time_after_eq(runstate->time[RUNSTATE_blocked],
+ per_cpu(processed_blocked_time, cpu)))
+ blocked = 0;
+
+ if (!time_after_eq(runstate->time[RUNSTATE_runnable] +
+ runstate->time[RUNSTATE_offline],
+ per_cpu(processed_stolen_time, cpu)))
+ stolen = 0;
+
+ if (!time_after(delta_itm + new_itm, ia64_get_itc()))
+ stolentick = ia64_get_itc() - delta_itm - new_itm;
+
+ do_div(stolentick, NS_PER_TICK);
+ stolentick++;
+
+ do_div(stolen, NS_PER_TICK);
+
+ if (stolen > stolentick)
+ stolen = stolentick;
+
+ stolentick -= stolen;
+ do_div(blocked, NS_PER_TICK);
+
+ if (blocked > stolentick)
+ blocked = stolentick;
+
+ if (stolen > 0 || blocked > 0) {
+ account_steal_time(NULL, jiffies_to_cputime(stolen));
+ account_steal_time(idle_task(cpu), jiffies_to_cputime(blocked));
+ run_local_timers();
+
+ if (rcu_pending(cpu))
+ rcu_check_callbacks(cpu, user_mode(get_irq_regs()));
+
+ scheduler_tick();
+ run_posix_cpu_timers(p);
+ delta_itm += local_cpu_data->itm_delta * (stolen + blocked);
+
+ if (cpu = time_keeper_id) {
+ write_seqlock(&xtime_lock);
+ do_timer(stolen + blocked);
+ local_cpu_data->itm_next = delta_itm + new_itm;
+ write_sequnlock(&xtime_lock);
+ } else {
+ local_cpu_data->itm_next = delta_itm + new_itm;
+ }
+ per_cpu(processed_stolen_time, cpu) += NS_PER_TICK * stolen;
+ per_cpu(processed_blocked_time, cpu) += NS_PER_TICK * blocked;
+ }
+ return delta_itm;
+}
+#else
+#define consider_steal_time(new_itm) (0)
+#endif
+
static irqreturn_t
timer_interrupt (int irq, void *dev_id)
{
unsigned long new_itm;
+ unsigned long delta_itm; /* XEN */
if (unlikely(cpu_is_offline(smp_processor_id()))) {
return IRQ_HANDLED;
@@ -78,6 +189,13 @@ timer_interrupt (int irq, void *dev_id)
profile_tick(CPU_PROFILING);
+ if (is_running_on_xen()) {
+ delta_itm = consider_steal_time(new_itm);
+ new_itm += delta_itm;
+ if (time_after(new_itm, ia64_get_itc()) && delta_itm)
+ goto skip_process_time_accounting;
+ }
+
while (1) {
update_process_times(user_mode(get_irq_regs()));
@@ -107,6 +225,8 @@ timer_interrupt (int irq, void *dev_id)
local_irq_disable();
}
+skip_process_time_accounting: /* XEN */
+
do {
/*
* If we're too close to the next clock tick for
@@ -161,6 +281,84 @@ static int __init nojitter_setup(char *str)
__setup("nojitter", nojitter_setup);
+#ifdef CONFIG_XEN
+/* taken from i386/kernel/time-xen.c */
+static void init_missing_ticks_accounting(int cpu)
+{
+ struct vcpu_register_runstate_memory_area area;
+ struct vcpu_runstate_info *runstate = &per_cpu(runstate, cpu);
+ int rc;
+
+ memset(runstate, 0, sizeof(*runstate));
+
+ area.addr.v = runstate;
+ rc = HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area, cpu,
+ &area);
+ WARN_ON(rc && rc != -ENOSYS);
+
+ per_cpu(processed_blocked_time, cpu) = runstate->time[RUNSTATE_blocked];
+ per_cpu(processed_stolen_time, cpu) = runstate->time[RUNSTATE_runnable]
+ + runstate->time[RUNSTATE_offline];
+}
+
+static int xen_ia64_settimefoday_after_resume;
+
+static int __init __xen_ia64_settimeofday_after_resume(char *str)
+{
+ xen_ia64_settimefoday_after_resume = 1;
+ return 1;
+}
+
+__setup("xen_ia64_settimefoday_after_resume",
+ __xen_ia64_settimeofday_after_resume);
+
+/* Called after suspend, to resume time. */
+void
+time_resume(void)
+{
+ unsigned int cpu;
+
+ /* Just trigger a tick. */
+ ia64_cpu_local_tick();
+
+ if (xen_ia64_settimefoday_after_resume) {
+ /* do_settimeofday() resets timer interplator */
+ struct timespec xen_time;
+ int ret;
+ efi_gettimeofday(&xen_time);
+
+ ret = do_settimeofday(&xen_time);
+ WARN_ON(ret);
+ } else {
+#if 0
+ /* adjust EFI time */
+ struct timespec my_time = CURRENT_TIME;
+ struct timespec xen_time;
+ static timespec diff;
+ struct xen_domctl domctl;
+ int ret;
+
+ efi_gettimeofday(&xen_time);
+ diff = timespec_sub(&xen_time, &my_time);
+ domctl.cmd = XEN_DOMCTL_settimeoffset;
+ domctl.domain = DOMID_SELF;
+ domctl.u.settimeoffset.timeoffset_seconds = diff.tv_sec;
+ ret = HYPERVISOR_domctl_op(&domctl);
+ WARN_ON(ret);
+#endif
+ /* itc_clocksource remembers the last timer status in
+ * itc_jitter_data. Forget it */
+ clocksource_resume();
+ }
+
+ for_each_online_cpu(cpu)
+ init_missing_ticks_accounting(cpu);
+
+ touch_softlockup_watchdog();
+}
+#else
+#define init_missing_ticks_accounting(cpu) do {} while (0)
+#endif
void __devinit
ia64_init_itm (void)
@@ -256,6 +454,12 @@ ia64_init_itm (void)
*/
clocksource_itc.rating = 50;
+ if (is_running_on_xen())
+ init_missing_ticks_accounting(smp_processor_id());
+
+ /* avoid softlock up message when cpu is unplug and plugged again. */
+ touch_softlockup_watchdog();
+
/* Setup the CPU local timer tick */
ia64_cpu_local_tick();
--
1.5.3
^ permalink raw reply related [flat|nested] 3+ messages in thread