[Qemu-devel] [RFC 1/2] kvmclock: remove cpu_synchronize_all_states call

[Liang Li <liang.z.li@intel.com>]

The commit 317b0a6d8 fixed an issue which caused by the outdated env->tsc
value, but the fix lead to cpu_synchronize_all_states called twice during
live migration. The cpu_synchronize_all_states takes about 300us, which is
a bit expensive.
This patch tries to fix the issue in another way, and there is only one
cpu_synchronize_all_states call during live migration.

Signed-off-by: Liang Li <liang.z.li@intel.com>
---
 cpus.c              |  8 ++++++++
 hw/i386/kvm/clock.c | 17 -----------------
 2 files changed, 8 insertions(+), 17 deletions(-)

diff --git a/cpus.c b/cpus.c
index c1e74d9..cd5abef 100644
--- a/cpus.c
+++ b/cpus.c
@@ -634,6 +634,14 @@ static int do_vm_stop(RunState state)
         pause_all_vcpus();
         runstate_set(state);
         vm_state_notify(0, state);
+        switch (state) {
+        case RUN_STATE_PAUSED:
+        case RUN_STATE_SUSPENDED:
+            cpu_synchronize_all_states();
+            break;
+        default:
+            break;
+        }
         qapi_event_send_stop(&error_abort);
     }
 
diff --git a/hw/i386/kvm/clock.c b/hw/i386/kvm/clock.c
index efdf165..a9db073 100644
--- a/hw/i386/kvm/clock.c
+++ b/hw/i386/kvm/clock.c
@@ -17,7 +17,6 @@
 #include "qemu/host-utils.h"
 #include "sysemu/sysemu.h"
 #include "sysemu/kvm.h"
-#include "sysemu/cpus.h"
 #include "hw/sysbus.h"
 #include "hw/kvm/clock.h"
 
@@ -125,22 +124,6 @@ static void kvmclock_vm_state_change(void *opaque, int running,
             return;
         }
 
-        cpu_synchronize_all_states();
-        /* In theory, the cpu_synchronize_all_states() call above wouldn't
-         * affect the rest of the code, as the VCPU state inside CPUState
-         * is supposed to always match the VCPU state on the kernel side.
-         *
-         * In practice, calling cpu_synchronize_state() too soon will load the
-         * kernel-side APIC state into X86CPU.apic_state too early, APIC state
-         * won't be reloaded later because CPUState.vcpu_dirty==true, and
-         * outdated APIC state may be migrated to another host.
-         *
-         * The real fix would be to make sure outdated APIC state is read
-         * from the kernel again when necessary. While this is not fixed, we
-         * need the cpu_clean_all_dirty() call below.
-         */
-        cpu_clean_all_dirty();
-
         ret = kvm_vm_ioctl(kvm_state, KVM_GET_CLOCK, &data);
         if (ret < 0) {
             fprintf(stderr, "KVM_GET_CLOCK failed: %s\n", strerror(ret));
-- 
1.9.1