Re: [Qemu-devel] [PATCH v2] kvmclock: Ensure time in migration never goes backward

[Paolo Bonzini <pbonzini@redhat.com>]

Il 03/06/2014 07:16, Marcelo Tosatti ha scritto:
> On Mon, Jun 02, 2014 at 10:31:44PM +0200, Marcin Gibuła wrote:
>>> +    cpu_physical_memory_read(kvmclock_struct_pa, &time, sizeof(time));
>>> +
>>> +    delta = migration_tsc - time.tsc_timestamp;
>>
>> Hi,
>>
>> 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.
>>
>> 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.
>>
>> The double call happens when switching from old to new disk
>> (pivoting in libvirt's nomenclature).
>>
>> Example values:
>>
>> First call: migration_tsc: 12052203518652476, time_tsc:
>> 12052203301565676, delta 108543400
>>
>> Second call: migration_tsc: 12052203518652476, time_tsc:
>> 12052204478600322, delta 9223372036374801885
>>
>> Perhaps it is worth adding:
>>
>> if (time.tsc_timestamp > migration_tsc) {
>>     return 0;
>> }
>>
>> 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"
>
> @@ -65,6 +66,7 @@ static uint64_t kvmclock_current_nsec(KVMClockState *s)
>
>      cpu_physical_memory_read(kvmclock_struct_pa, &time, sizeof(time));
>
> +    assert(time.tsc_timestamp <= migration_tsc);
>      delta = migration_tsc - time.tsc_timestamp;
>      if (time.tsc_shift < 0) {
>          delta >>= -time.tsc_shift;
> @@ -123,6 +125,8 @@ static void kvmclock_vm_state_change(void *opaque, int running,
>          if (s->clock_valid) {
>              return;
>          }
> +
> +        cpu_synchronize_all_states();
>          ret = kvm_vm_ioctl(kvm_state, KVM_GET_CLOCK, &data);
>          if (ret < 0) {
>              fprintf(stderr, "KVM_GET_CLOCK failed: %s\n", strerror(ret));
>

Can you send it with a real commit message?

Paolo