x86: Use xen_vcpuop_clockevent, xen_clocksource and xen wallclock.

Use xen_vcpuop_clockevent instead of hpet and APIC timers as main
clockevent device on all vcpus, use the xen wallclock time as wallclock
instead of rtc and use xen_clocksource as clocksource.
The pv clock algorithm needs to work correctly for the xen_clocksource
and xen wallclock to be usable, only modern Xen versions offer a
reliable pv clock in HVM guests (XENFEAT_hvm_safe_pvclock).

Using the hpet as clocksource means a VMEXIT every time we read/write to
the hpet mmio addresses, pvclock give us a better rating without
VMEXITs. Same goes for the xen wallclock and xen_vcpuop_clockevent

Signed-off-by: Stefano Stabellini <[e-mail]>
Signed-off-by: Don Dutile <[e-mail]>
Signed-off-by: Jeremy Fitzhardinge <[e-mail]>

arch/x86/xen/enlighten.c

···
 	.patch = xen_patch,
 };
 
-static const struct pv_time_ops xen_time_ops __initdata = {
-	.sched_clock = xen_sched_clock,
-};
-
 static const struct pv_cpu_ops xen_cpu_ops __initdata = {
 	.cpuid = xen_cpuid,
 
···
 	/* Install Xen paravirt ops */
 	pv_info = xen_info;
 	pv_init_ops = xen_init_ops;
-	pv_time_ops = xen_time_ops;
 	pv_cpu_ops = xen_cpu_ops;
 	pv_apic_ops = xen_apic_ops;
 
···
 	x86_init.oem.arch_setup = xen_arch_setup;
 	x86_init.oem.banner = xen_banner;
 
-	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;
+	xen_init_time_ops();
 
 	/*
 	 * Set up some pagetable state before starting to set any ptes.
···
 	register_cpu_notifier(&xen_hvm_cpu_notifier);
 	have_vcpu_info_placement = 0;
 	x86_init.irqs.intr_init = xen_init_IRQ;
+	xen_hvm_init_time_ops();
 }
 
 static bool __init xen_hvm_platform(void)

arch/x86/xen/suspend.c

···
 
 void xen_hvm_post_suspend(int suspend_cancelled)
 {
+	int cpu;
 	xen_hvm_init_shared_info();
 	xen_callback_vector();
+	if (xen_feature(XENFEAT_hvm_safe_pvclock)) {
+		for_each_online_cpu(cpu) {
+			xen_setup_runstate_info(cpu);
+		}
+	}
 }
 
 void xen_post_suspend(int suspend_cancelled)

arch/x86/xen/time.c

···
 #include <asm/xen/hypercall.h>
 
 #include <xen/events.h>
+#include <xen/features.h>
 #include <xen/interface/xen.h>
 #include <xen/interface/vcpu.h>
 
···
  * nanoseconds, which is nanoseconds the VCPU spent in RUNNING+BLOCKED
  * states.
  */
-unsigned long long xen_sched_clock(void)
+static unsigned long long xen_sched_clock(void)
 {
 	struct vcpu_runstate_info state;
 	cycle_t now;
···
 
 
 /* Get the TSC speed from Xen */
-unsigned long xen_tsc_khz(void)
+static unsigned long xen_tsc_khz(void)
 {
 	struct pvclock_vcpu_time_info *info =
 		&HYPERVISOR_shared_info->vcpu_info[0].time;
···
 	put_cpu_var(xen_vcpu);
 }
 
-unsigned long xen_get_wallclock(void)
+static unsigned long xen_get_wallclock(void)
 {
 	struct timespec ts;
 
···
 	return ts.tv_sec;
 }
 
-int xen_set_wallclock(unsigned long now)
+static int xen_set_wallclock(unsigned long now)
 {
 	/* do nothing for domU */
 	return -1;
···
 	}
 }
 
-__init void xen_time_init(void)
+static const struct pv_time_ops xen_time_ops __initdata = {
+	.sched_clock = xen_sched_clock,
+};
+
+static __init void xen_time_init(void)
 {
 	int cpu = smp_processor_id();
 	struct timespec tp;
···
 	xen_setup_timer(cpu);
 	xen_setup_cpu_clockevents();
 }
+
+__init void 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;
+}
+
+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();
+}
+
+__init void xen_hvm_init_time_ops(void)
+{
+	/* vector callback is needed otherwise we cannot receive interrupts
+	 * on cpu > 0 */
+	if (!xen_have_vector_callback && num_present_cpus() > 1)
+		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;
+}
+

arch/x86/xen/xen-ops.h

···
 void xen_teardown_timer(int cpu);
 cycle_t xen_clocksource_read(void);
 void xen_setup_cpu_clockevents(void);
-unsigned long xen_tsc_khz(void);
-void __init xen_time_init(void);
-unsigned long xen_get_wallclock(void);
-int xen_set_wallclock(unsigned long time);
-unsigned long long xen_sched_clock(void);
+void __init xen_init_time_ops(void);
+void __init xen_hvm_init_time_ops(void);
 
 irqreturn_t xen_debug_interrupt(int irq, void *dev_id);
 

include/xen/interface/features.h

···
 /* x86: Does this Xen host support the HVM callback vector type? */
 #define XENFEAT_hvm_callback_vector        8
 
+/* x86: pvclock algorithm is safe to use on HVM */
+#define XENFEAT_hvm_safe_pvclock           9
+
 #define XENFEAT_NR_SUBMAPS 1
 
 #endif /* __XEN_PUBLIC_FEATURES_H__ */