From: Glauber Costa <glommer@redhat.com>
Date: Tue, 3 Nov 2009 17:41:17 -0500
Subject: [x86] kvm: don't ask HV for tsc khz if not using kvmclock
Message-id: <1257270077-25607-1-git-send-email-glommer@redhat.com>
Patchwork-id: 21300
O-Subject: [PATCH] BZ531268 don't ask hypervisor for tsc khz if not using
	kvmclock
Bugzilla: 531268
RH-Acked-by: Rik van Riel <riel@redhat.com>
RH-Acked-by: Marcelo Tosatti <mtosatti@redhat.com>

In my last patches, we started using the upstream version of kvm_get_tsc_khz,
which depends on kvmclock being already initialized. The version that we had,
did not depend on that, and thus, could be always used.

However, now that we don't initialize tsc_khz ourselves, this leads to a crash.
Keeping the behaviour of always being able to use tsc_khz is too complex,
and does not match upstream anyway. So I propose we make it conditional
on kvm-clock being used.

I am sending as a separate patch, since the last series were already triple-acked,
and are correct on its own.

Signed-off-by: Glauber Costa <glommer@redhat.com>

diff --git a/arch/i386/kernel/kvmclock.c b/arch/i386/kernel/kvmclock.c
index 7ed58c0..4c96682 100644
--- a/arch/i386/kernel/kvmclock.c
+++ b/arch/i386/kernel/kvmclock.c
@@ -87,9 +87,12 @@ cycle_t kvm_clock_read(void)
  */
 unsigned long kvm_get_tsc_khz(void)
 {
-	struct pvclock_vcpu_time_info *src;
-	src = &per_cpu(hv_clock, 0);
-	return pvclock_tsc_khz(src);
+	if (use_kvm_time > 0) {
+		struct pvclock_vcpu_time_info *src;
+		src = &per_cpu(hv_clock, 0);
+		return pvclock_tsc_khz(src);
+	} else
+		return 0; /* indicates we should not use this value */
 }