Re: [PATCH] x86/kvm: fix condition to update kvm master clocks

[Marcelo Tosatti <mtosatti@redhat.com>]

On Thu, May 26, 2016 at 05:49:55PM +0300, Roman Kagan wrote:
> The condition to schedule per-VM master clock updates is inversed; as a
> result the master clocks are never updated and the kvm_clock drift WRT
> host's NTP-disciplined clock (which was the motivation to introduce this
> machinery in the first place) still remains.
> 
> Fix it, and reword the comment to make it more apparent what the desired
> behavior is.
> 
> Cc: Owen Hofmann <osh@google.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Marcelo Tosatti <mtosatti@redhat.com>
> Signed-off-by: Roman Kagan <rkagan@virtuozzo.com>
> ---
>  arch/x86/kvm/x86.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index c805cf4..d8f591c 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -5810,10 +5810,10 @@ static int pvclock_gtod_notify(struct notifier_block *nb, unsigned long unused,
>  
>  	update_pvclock_gtod(tk);
>  
> -	/* disable master clock if host does not trust, or does not
> -	 * use, TSC clocksource
> +	/* only schedule per-VM master clock updates if the host uses TSC and
> +	 * there's at least one VM in need of an update
>  	 */
> -	if (gtod->clock.vclock_mode != VCLOCK_TSC &&
> +	if (gtod->clock.vclock_mode == VCLOCK_TSC &&
>  	    atomic_read(&kvm_guest_has_master_clock) != 0)
>  		queue_work(system_long_wq, &pvclock_gtod_work);
>  
> -- 
> 2.5.5

NAK, as stated this leaves clocks out of sync between execution of
pvclock_gtod_notify and pvclock_gtod_work execution.