Re: [Xen-devel] [PATCH v2 5/5] x86: Rework MSR_TSC_AUX handling from scratch.

2018-02-26 Thread Jan Beulich
>>> On 23.02.18 at 16:51, wrote: > On 23/02/18 15:05, Jan Beulich wrote: > On 21.02.18 at 12:36, wrote: >>> --- a/xen/arch/x86/msr.c >>> +++ b/xen/arch/x86/msr.c >>> @@ -175,6 +175,13 @@ int guest_rdmsr(const struct vcpu *v, uint32_t msr,

Re: [Xen-devel] [PATCH v2 5/5] x86: Rework MSR_TSC_AUX handling from scratch.

2018-02-23 Thread Andrew Cooper
On 23/02/18 15:05, Jan Beulich wrote: On 21.02.18 at 12:36, wrote: >> There are many problems with MSR_TSC_AUX handling. >> >> To begin with, the RDPID instruction reads MSR_TSC_AUX, but it is only the >> RDTSCP feature which enumerates the MSR. Therefore, RDPID

Re: [Xen-devel] [PATCH v2 5/5] x86: Rework MSR_TSC_AUX handling from scratch.

2018-02-23 Thread Jan Beulich
>>> On 21.02.18 at 12:36, wrote: > There are many problems with MSR_TSC_AUX handling. > > To begin with, the RDPID instruction reads MSR_TSC_AUX, but it is only the > RDTSCP feature which enumerates the MSR. Therefore, RDPID functionally > depends on RDTSCP. Are you

Re: [Xen-devel] [PATCH v2 5/5] x86: Rework MSR_TSC_AUX handling from scratch.

2018-02-21 Thread Roger Pau Monné
On Wed, Feb 21, 2018 at 11:36:15AM +, Andrew Cooper wrote: > There are many problems with MSR_TSC_AUX handling. > > To begin with, the RDPID instruction reads MSR_TSC_AUX, but it is only the > RDTSCP feature which enumerates the MSR. Therefore, RDPID functionally > depends on RDTSCP. > >

[Xen-devel] [PATCH v2 5/5] x86: Rework MSR_TSC_AUX handling from scratch.

2018-02-21 Thread Andrew Cooper
There are many problems with MSR_TSC_AUX handling. To begin with, the RDPID instruction reads MSR_TSC_AUX, but it is only the RDTSCP feature which enumerates the MSR. Therefore, RDPID functionally depends on RDTSCP. For PV guests, we hide RDTSCP but advertise RDPID. We also silently drop