[Qemu-devel] [PATCH] kvmclock: Add comment explaining why we need cpu_clean_all_dirty()

[Qemu-devel] [PATCH] kvmclock: Add comment explaining why we need cpu_clean_all_dirty()

[Eduardo Habkost]

Try to explain why commit 317b0a6d8ba44e9bf8f9c3dbd776c4536843d82c
needed a cpu_clean_all_dirty() call just after calling
cpu_synchronize_all_states().

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Cc: Andrey Korolyov <andrey@xdel.ru>
Cc: Marcin Gibuła <m.gibula@beyond.pl>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
---
 hw/i386/kvm/clock.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/hw/i386/kvm/clock.c b/hw/i386/kvm/clock.c
index 1ac60d6..58be2bd 100644
--- a/hw/i386/kvm/clock.c
+++ b/hw/i386/kvm/clock.c
@@ -127,7 +127,21 @@ static void kvmclock_vm_state_change(void *opaque, int running,
         }
 
         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+         * outdated APIC state may be migrated to another host.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+         * The real fix would be to make sure outdated APIC state is read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));
-- 
1.9.3

Re: [Qemu-devel] [PATCH] kvmclock: Add comment explaining why we need cpu_clean_all_dirty()

[Paolo Bonzini]

On 03/11/2014 18:45, Eduardo Habkost wrote:
> Try to explain why commit 317b0a6d8ba44e9bf8f9c3dbd776c4536843d82c
> needed a cpu_clean_all_dirty() call just after calling
> cpu_synchronize_all_states().
> 
> Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
> Cc: Andrey Korolyov <andrey@xdel.ru>
> Cc: Marcin Gibuła <m.gibula@beyond.pl>
> Cc: Marcelo Tosatti <mtosatti@redhat.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  hw/i386/kvm/clock.c | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)
> 
> diff --git a/hw/i386/kvm/clock.c b/hw/i386/kvm/clock.c
> index 1ac60d6..58be2bd 100644
> --- a/hw/i386/kvm/clock.c
> +++ b/hw/i386/kvm/clock.c
> @@ -127,7 +127,21 @@ static void kvmclock_vm_state_change(void *opaque, int running,
>          }
>  
>          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));
> 

Thanks, applying to uq/master.

Paolo