Re: [patch 4/7] x86/kvmclock: Cleanup the code

2018-07-09 Thread Paolo Bonzini
On 09/07/2018 11:05, Peter Zijlstra wrote: > On Fri, Jul 06, 2018 at 06:13:11PM +0200, Thomas Gleixner wrote: >> -native_write_msr(msr_kvm_wall_clock, low, high); >> +wrmsrl(msr_kvm_wall_clock, slow_virt_to_phys(&wall_clock)); > Does it matter that you went from an explicit native WRMSR ins

Re: [patch 4/7] x86/kvmclock: Cleanup the code

2018-07-09 Thread Thomas Gleixner
On Mon, 9 Jul 2018, Peter Zijlstra wrote: > On Fri, Jul 06, 2018 at 06:13:11PM +0200, Thomas Gleixner wrote: > > - native_write_msr(msr_kvm_wall_clock, low, high); > > + wrmsrl(msr_kvm_wall_clock, slow_virt_to_phys(&wall_clock)); > > Does it matter that you went from an explicit native WRMSR i

Re: [patch 4/7] x86/kvmclock: Cleanup the code

2018-07-09 Thread Peter Zijlstra
On Fri, Jul 06, 2018 at 06:13:11PM +0200, Thomas Gleixner wrote: > - native_write_msr(msr_kvm_wall_clock, low, high); > + wrmsrl(msr_kvm_wall_clock, slow_virt_to_phys(&wall_clock)); Does it matter that you went from an explicit native WRMSR instruction to a potentially paravirt MSR thing?

Re: [patch 4/7] x86/kvmclock: Cleanup the code

2018-07-06 Thread Paolo Bonzini
> - Cleanup the mrs write for wall clock. s/mrs/MSR/ On 06/07/2018 18:13, Thomas Gleixner wrote: > @@ -195,6 +181,7 @@ static void kvm_register_clock(char *txt > if (!hv_clock) > return; > > + src = &hv_clock[cpu].pvti; > pa = slow_virt_to_phys(src) | 0x01ULL; >