>>> On 10.09.15 at 08:06, <ko...@jp.fujitsu.com> wrote:
> The ACPI PM timer is sometimes broken on live migration.
> Since vcpu->arch.hvm_vcpu.guest_time is always zero in other than
> "delay for missed ticks mode". Even in "delay for missed ticks mode",
> vcpu's guest_time field is not valid (i.e. zero) when
> the state of vcpu is "blocked". (see pt_save_timer function)

Better, but still leaving open why we should not use
v->arch.hvm_vcpu.guest_time when non-zero. In particular your
reference to a blocked vCPU is bogus, as in that case the field will
continue to be zero.

The fear I have with completely ignoring the field is that we may
lose guest time precision in the delay-for-missed-ticks mode. Otoh
I appreciate that you patch gets things in line with vHPET; the
question here is whether the vHPET save logic should also take
that field into account when non-zero.

> The original author (Tim Deegan) of pmtimer_save() must have intended
> that it saves the last scheduled time of the vcpu. Unfortunately it was
> already implied this bug. FYI, there is no other timer mode than 
> "delay for missed ticks mode" then.

I'm not sure how to interpret this paragraph. It would have been
helpful anyway if you had Cc-ed Tim right away...

Jan


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

Reply via email to