From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:58748) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Wrh64-0008Da-Js for qemu-devel@nongnu.org; Tue, 03 Jun 2014 01:17:22 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Wrh5y-0004XQ-FW for qemu-devel@nongnu.org; Tue, 03 Jun 2014 01:17:16 -0400 Received: from mx1.redhat.com ([209.132.183.28]:50082) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Wrh5y-0004XB-7I for qemu-devel@nongnu.org; Tue, 03 Jun 2014 01:17:10 -0400 Date: Tue, 3 Jun 2014 02:16:30 -0300 From: Marcelo Tosatti Message-ID: <20140603051630.GA2289@amt.cnet> References: <1400253321-9239-1-git-send-email-agraf@suse.de> <538CDF30.9050902@beyond.pl> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <538CDF30.9050902@beyond.pl> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v2] kvmclock: Ensure time in migration never goes backward List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Marcin =?utf-8?Q?Gibu=C5=82a?= Cc: pbonzini@redhat.com, Alexander Graf , qemu-devel@nongnu.org On Mon, Jun 02, 2014 at 10:31:44PM +0200, Marcin Gibu=C5=82a wrote: > >+ cpu_physical_memory_read(kvmclock_struct_pa, &time, sizeof(time))= ; > >+ > >+ delta =3D migration_tsc - time.tsc_timestamp; >=20 > Hi, >=20 > when I was testing live storage migration with libvirt I found out > that this patch can cause virtual machine to hang when completing > mirror job. >=20 > This is (probably) because kvmclock_current_nsec() is called twice > in a row and on second call time.tsc_timestamp is larger than > migration_tsc. This causes delta to be huge and sets timer to > invalid value. >=20 > The double call happens when switching from old to new disk > (pivoting in libvirt's nomenclature). >=20 > Example values: >=20 > First call: migration_tsc: 12052203518652476, time_tsc: > 12052203301565676, delta 108543400 >=20 > Second call: migration_tsc: 12052203518652476, time_tsc: > 12052204478600322, delta 9223372036374801885 >=20 > Perhaps it is worth adding: >=20 > if (time.tsc_timestamp > migration_tsc) { > return 0; > } >=20 > there? Untested though... Hi Marcin, Can you give this patch a try? Should read the guest TSC values after stopping the VM. diff --git a/hw/i386/kvm/clock.c b/hw/i386/kvm/clock.c index 6f4ed28a..bef2504 100644 --- a/hw/i386/kvm/clock.c +++ b/hw/i386/kvm/clock.c @@ -17,6 +17,7 @@ #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" =20 @@ -65,6 +66,7 @@ static uint64_t kvmclock_current_nsec(KVMClockState *s) =20 cpu_physical_memory_read(kvmclock_struct_pa, &time, sizeof(time)); =20 + assert(time.tsc_timestamp <=3D migration_tsc); delta =3D migration_tsc - time.tsc_timestamp; if (time.tsc_shift < 0) { delta >>=3D -time.tsc_shift; @@ -123,6 +125,8 @@ static void kvmclock_vm_state_change(void *opaque, in= t running, if (s->clock_valid) { return; } + + cpu_synchronize_all_states(); ret =3D kvm_vm_ioctl(kvm_state, KVM_GET_CLOCK, &data); if (ret < 0) { fprintf(stderr, "KVM_GET_CLOCK failed: %s\n", strerror(ret))= ;