x86: KVM guest: Use the paravirt clocksource structs and functions
authorGerd Hoffmann <kraxel@redhat.com>
Tue, 3 Jun 2008 14:17:32 +0000 (16:17 +0200)
committerAvi Kivity <avi@qumranet.com>
Tue, 24 Jun 2008 18:02:33 +0000 (21:02 +0300)
This patch updates the kvm host code to use the pvclock structs
and functions, thereby making it compatible with Xen.

The patch also fixes an initialization bug: on SMP systems the
per-cpu has two different locations early at boot and after CPU
bringup.  kvmclock must take that in account when registering the
physical address within the host.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
arch/x86/Kconfig
arch/x86/kernel/kvmclock.c

index f94bca6ff47f33df81763cc2e9b6841c5a126e63..e0edaaa6920af63832979a04d4f525acafe64e8a 100644 (file)
@@ -383,6 +383,7 @@ config VMI
 config KVM_CLOCK
        bool "KVM paravirtualized clock"
        select PARAVIRT
+       select PARAVIRT_CLOCK
        depends on !(X86_VISWS || X86_VOYAGER)
        help
          Turning on this option will allow you to run a paravirtualized clock
index 08a30986d4725f25ebc316b196884dd8f5aac9d7..87edf1ceb1df32c3afd62d1de44801c286922f7e 100644 (file)
@@ -18,6 +18,7 @@
 
 #include <linux/clocksource.h>
 #include <linux/kvm_para.h>
+#include <asm/pvclock.h>
 #include <asm/arch_hooks.h>
 #include <asm/msr.h>
 #include <asm/apic.h>
@@ -36,18 +37,9 @@ static int parse_no_kvmclock(char *arg)
 early_param("no-kvmclock", parse_no_kvmclock);
 
 /* The hypervisor will put information about time periodically here */
-static DEFINE_PER_CPU_SHARED_ALIGNED(struct kvm_vcpu_time_info, hv_clock);
-#define get_clock(cpu, field) per_cpu(hv_clock, cpu).field
+static DEFINE_PER_CPU_SHARED_ALIGNED(struct pvclock_vcpu_time_info, hv_clock);
+static struct pvclock_wall_clock wall_clock;
 
-static inline u64 kvm_get_delta(u64 last_tsc)
-{
-       int cpu = smp_processor_id();
-       u64 delta = native_read_tsc() - last_tsc;
-       return (delta * get_clock(cpu, tsc_to_system_mul)) >> KVM_SCALE;
-}
-
-static struct kvm_wall_clock wall_clock;
-static cycle_t kvm_clock_read(void);
 /*
  * The wallclock is the time of day when we booted. Since then, some time may
  * have elapsed since the hypervisor wrote the data. So we try to account for
@@ -55,64 +47,37 @@ static cycle_t kvm_clock_read(void);
  */
 static unsigned long kvm_get_wallclock(void)
 {
-       u32 wc_sec, wc_nsec;
-       u64 delta;
+       struct pvclock_vcpu_time_info *vcpu_time;
        struct timespec ts;
-       int version, nsec;
        int low, high;
 
        low = (int)__pa(&wall_clock);
        high = ((u64)__pa(&wall_clock) >> 32);
+       native_write_msr(MSR_KVM_WALL_CLOCK, low, high);
 
-       delta = kvm_clock_read();
+       vcpu_time = &get_cpu_var(hv_clock);
+       pvclock_read_wallclock(&wall_clock, vcpu_time, &ts);
+       put_cpu_var(hv_clock);
 
-       native_write_msr(MSR_KVM_WALL_CLOCK, low, high);
-       do {
-               version = wall_clock.wc_version;
-               rmb();
-               wc_sec = wall_clock.wc_sec;
-               wc_nsec = wall_clock.wc_nsec;
-               rmb();
-       } while ((wall_clock.wc_version != version) || (version & 1));
-
-       delta = kvm_clock_read() - delta;
-       delta += wc_nsec;
-       nsec = do_div(delta, NSEC_PER_SEC);
-       set_normalized_timespec(&ts, wc_sec + delta, nsec);
-       /*
-        * Of all mechanisms of time adjustment I've tested, this one
-        * was the champion!
-        */
-       return ts.tv_sec + 1;
+       return ts.tv_sec;
 }
 
 static int kvm_set_wallclock(unsigned long now)
 {
-       return 0;
+       return -1;
 }
 
-/*
- * This is our read_clock function. The host puts an tsc timestamp each time
- * it updates a new time. Without the tsc adjustment, we can have a situation
- * in which a vcpu starts to run earlier (smaller system_time), but probes
- * time later (compared to another vcpu), leading to backwards time
- */
 static cycle_t kvm_clock_read(void)
 {
-       u64 last_tsc, now;
-       int cpu;
+       struct pvclock_vcpu_time_info *src;
+       cycle_t ret;
 
-       preempt_disable();
-       cpu = smp_processor_id();
-
-       last_tsc = get_clock(cpu, tsc_timestamp);
-       now = get_clock(cpu, system_time);
-
-       now += kvm_get_delta(last_tsc);
-       preempt_enable();
-
-       return now;
+       src = &get_cpu_var(hv_clock);
+       ret = pvclock_clocksource_read(src);
+       put_cpu_var(hv_clock);
+       return ret;
 }
+
 static struct clocksource kvm_clock = {
        .name = "kvm-clock",
        .read = kvm_clock_read,
@@ -123,13 +88,14 @@ static struct clocksource kvm_clock = {
        .flags = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static int kvm_register_clock(void)
+static int kvm_register_clock(char *txt)
 {
        int cpu = smp_processor_id();
        int low, high;
        low = (int)__pa(&per_cpu(hv_clock, cpu)) | 1;
        high = ((u64)__pa(&per_cpu(hv_clock, cpu)) >> 32);
-
+       printk(KERN_INFO "kvm-clock: cpu %d, msr %x:%x, %s\n",
+              cpu, high, low, txt);
        return native_write_msr_safe(MSR_KVM_SYSTEM_TIME, low, high);
 }
 
@@ -140,12 +106,20 @@ static void kvm_setup_secondary_clock(void)
         * Now that the first cpu already had this clocksource initialized,
         * we shouldn't fail.
         */
-       WARN_ON(kvm_register_clock());
+       WARN_ON(kvm_register_clock("secondary cpu clock"));
        /* ok, done with our trickery, call native */
        setup_secondary_APIC_clock();
 }
 #endif
 
+#ifdef CONFIG_SMP
+void __init kvm_smp_prepare_boot_cpu(void)
+{
+       WARN_ON(kvm_register_clock("primary cpu clock"));
+       native_smp_prepare_boot_cpu();
+}
+#endif
+
 /*
  * After the clock is registered, the host will keep writing to the
  * registered memory location. If the guest happens to shutdown, this memory
@@ -174,13 +148,16 @@ void __init kvmclock_init(void)
                return;
 
        if (kvmclock && kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE)) {
-               if (kvm_register_clock())
+               if (kvm_register_clock("boot clock"))
                        return;
                pv_time_ops.get_wallclock = kvm_get_wallclock;
                pv_time_ops.set_wallclock = kvm_set_wallclock;
                pv_time_ops.sched_clock = kvm_clock_read;
 #ifdef CONFIG_X86_LOCAL_APIC
                pv_apic_ops.setup_secondary_clock = kvm_setup_secondary_clock;
+#endif
+#ifdef CONFIG_SMP
+               smp_ops.smp_prepare_boot_cpu = kvm_smp_prepare_boot_cpu;
 #endif
                machine_ops.shutdown  = kvm_shutdown;
 #ifdef CONFIG_KEXEC