Re: [PATCH v3] x86/vtsc: update vcpu_time in hvm_set_guest_time

[George Dunlap <george.dunlap@eu.citrix.com>]

On 05/06/13 13:02, Diana Crisan wrote:
> On 04/06/13 16:32, Roger Pau Monne wrote:
>> When using a vtsc, hvm_set_guest_time changes hvm_vcpu.stime_offset,
>> which is used in the vcpu time structure to calculate the
>> tsc_timestamp, so after updating stime_offset we need to propagate the
>> change to vcpu_time in order for the guest to get the right time if
>> using the PV clock.
>>
>> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
>> Cc: Keir Fraser <keir@xen.org>
>> Cc: Jan Beulich <jbeulich@suse.com>
>> Cc: George Dunlap <george.dunlap@eu.citrix.com>
> Tested-By: Diana Crisan  <dcrisan@flexiant.com>
>> ---
>> Changes since v2:
>>   * Check for v == current instead of v->is_running to know if the vcpu
>>     is running.
>> Changes since v1:
>>   * Perform the call to update_vcpu_system_time in hvm_set_guest_time
>>     if the offset has changed and the vCPU is running.
>> ---
>>   xen/arch/x86/hvm/vpt.c |   13 ++++++++++++-
>>   1 files changed, 12 insertions(+), 1 deletions(-)
>>
>> diff --git a/xen/arch/x86/hvm/vpt.c b/xen/arch/x86/hvm/vpt.c
>> index 8dee662..4b1e1a3 100644
>> --- a/xen/arch/x86/hvm/vpt.c
>> +++ b/xen/arch/x86/hvm/vpt.c
>> @@ -57,7 +57,18 @@ u64 hvm_get_guest_time(struct vcpu *v)
>>     void hvm_set_guest_time(struct vcpu *v, u64 guest_time)
>>   {
>> -    v->arch.hvm_vcpu.stime_offset += guest_time - 
>> hvm_get_guest_time(v);
>> +    u64 offset = guest_time - hvm_get_guest_time(v);
>> +
>> +    if ( offset ) {
>> +        v->arch.hvm_vcpu.stime_offset += offset;
>> +        /*
>> +         * If hvm_vcpu.stime_offset is updated make sure to
>> +         * also update vcpu time, since this value is used to
>> +         * calculate the TSC.
>> +         */
>> +        if ( v == current )
>> +            update_vcpu_system_time(v);
>> +    }
>>   }
>>     static int pt_irq_vector(struct periodic_time *pt, enum 
>> hvm_intsrc src)
>
> I have tested this patch and I have not been able to reproduce the 
> guest stuck-clock problem we were seeing before (see *HVM Migration of 
> domU on Qemu-upstream DM causes stuck system clock with ACPI* for 
> details of the problem).
> Note: this was tested without the temporary fix to our problem ( which 
> is specifying tsc_mode='native_paravirt').

Great, thanks Diana.

  -George


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel