On 13/03/2018 18:48, Vitaly Kuznetsov wrote:
> Changes since v1:
> - Merged patches 1 and 2, skip savesegment() for FS and FS for x86_64
> as save_fsgs() already does that [Paolo Bonzini]
> - Rename save_current_fsgs() -> save_fsgs_for_kvm() [Andy Lutomirski]
> - Trimmed comments [Andy Lutomirski]
> - Add Andy's A-b to what is now PATCH2
>
> Some time ago Paolo suggested to take a look at probably unneeded expensive
> rdmsrs for FS/GS base MSR in vmx_save_host_state(). This is called on every
> vcpu run when we need to handle vmexit in userspace.
>
> vmx_save_host_state() is always called in a very well defined context
> (ioctl from userspace) so we may try to get the required values for FS/GS
> bases from in-kernel variables and avoid expensive rdmsrs.
>
> My debug shows we're shaving off 240 cpu cycles (E5-2603 v3).
>
> Vitaly Kuznetsov (2):
> x86/kvm/vmx: read MSR_{FS,KERNEL_GS}_BASE from current->thread
> x86/kvm/vmx: avoid expensive rdmsr for MSR_GS_BASE
>
> arch/x86/include/asm/processor.h | 10 ++++++++++
> arch/x86/kernel/cpu/common.c | 3 ++-
> arch/x86/kernel/process_64.c | 14 ++++++++++++++
> arch/x86/kvm/vmx.c | 16 ++++++++++++----
> 4 files changed, 38 insertions(+), 5 deletions(-)
>
Queued, thanks.
Paolo
