From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:39034) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Xlxoa-0000oj-4S for qemu-devel@nongnu.org; Wed, 05 Nov 2014 05:27:52 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1XlxoV-0006R0-Cr for qemu-devel@nongnu.org; Wed, 05 Nov 2014 05:27:48 -0500 Received: from mx1.redhat.com ([209.132.183.28]:56909) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1XlxoV-0006Qm-4V for qemu-devel@nongnu.org; Wed, 05 Nov 2014 05:27:43 -0500 Message-ID: <5459FB96.9020208@redhat.com> Date: Wed, 05 Nov 2014 11:27:34 +0100 From: Paolo Bonzini MIME-Version: 1.0 References: <1415036734-24310-1-git-send-email-ehabkost@redhat.com> In-Reply-To: <1415036734-24310-1-git-send-email-ehabkost@redhat.com> Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH] kvmclock: Add comment explaining why we need cpu_clean_all_dirty() List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Eduardo Habkost , qemu-devel@nongnu.org Cc: Andrey Korolyov , =?UTF-8?B?TWFyY2luIEdpYnXFgmE=?= , Marcelo Tosatti 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(). >=20 > Signed-off-by: Eduardo Habkost > Cc: Andrey Korolyov > Cc: Marcin Gibu=C5=82a > Cc: Marcelo Tosatti > Cc: Paolo Bonzini > --- > hw/i386/kvm/clock.c | 14 ++++++++++++++ > 1 file changed, 14 insertions(+) >=20 > 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, > } > =20 > cpu_synchronize_all_states(); > + /* In theory, the cpu_synchronize_all_states() call above woul= dn't > + * affect the rest of the code, as the VCPU state inside CPUSt= ate > + * is supposed to always match the VCPU state on the kernel si= de. > + * > + * In practice, calling cpu_synchronize_state() too soon will = load the > + * kernel-side APIC state into X86CPU.apic_state too early, AP= IC state > + * won't be reloaded later because CPUState.vcpu_dirty=3D=3Dtr= ue, and > + * outdated APIC state may be migrated to another host. > + * > + * The real fix would be to make sure outdated APIC state is r= ead > + * from the kernel again when necessary. While this is not fix= ed, we > + * need the cpu_clean_all_dirty() call below. > + */ > cpu_clean_all_dirty(); > + > ret =3D kvm_vm_ioctl(kvm_state, KVM_GET_CLOCK, &data); > if (ret < 0) { > fprintf(stderr, "KVM_GET_CLOCK failed: %s\n", strerror(ret= )); >=20 Thanks, applying to uq/master. Paolo