| Message ID | 20180306205838.2710-1-sean.j.christopherson@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 06/03/2018 21:58, Sean Christopherson wrote: > Fast emulation of processor I/O for IN was disabled on x86 (both VMX > and SVM) some years ago due to a buggy implementation. The addition > of kvm_fast_pio_in(), used by SVM, re-introduced (functional!) fast > emulation of IN. Piggyback SVM's work and use kvm_fast_pio_in() on > VMX instead of performing full emulation of IN. > > Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> There's a little code duplication that would be nice to avoid, moving kvm_skip_emulated_instruction to kvm_fast_pio_{in,out}. Otherwise, Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Paolo > --- > arch/x86/kvm/vmx.c | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > > diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c > index 051dab74e4e9..c8a8391e95e5 100644 > --- a/arch/x86/kvm/vmx.c > +++ b/arch/x86/kvm/vmx.c > @@ -6231,15 +6231,15 @@ static int handle_io(struct kvm_vcpu *vcpu) > > exit_qualification = vmcs_readl(EXIT_QUALIFICATION); > string = (exit_qualification & 16) != 0; > - in = (exit_qualification & 8) != 0; > > ++vcpu->stat.io_exits; > > - if (string || in) > + if (string) > return emulate_instruction(vcpu, 0) == EMULATE_DONE; > > port = exit_qualification >> 16; > size = (exit_qualification & 7) + 1; > + in = (exit_qualification & 8) != 0; > > ret = kvm_skip_emulated_instruction(vcpu); > > @@ -6247,7 +6247,10 @@ static int handle_io(struct kvm_vcpu *vcpu) > * TODO: we might be squashing a KVM_GUESTDBG_SINGLESTEP-triggered > * KVM_EXIT_DEBUG here. > */ > - return kvm_fast_pio_out(vcpu, size, port) && ret; > + if (in) > + return kvm_fast_pio_in(vcpu, size, port) && ret; > + else > + return kvm_fast_pio_out(vcpu, size, port) && ret; > } > > static void >
On Wed, 2018-03-07 at 16:19 +0100, Paolo Bonzini wrote: > On 06/03/2018 21:58, Sean Christopherson wrote: > > > > Fast emulation of processor I/O for IN was disabled on x86 (both VMX > > and SVM) some years ago due to a buggy implementation. The addition > > of kvm_fast_pio_in(), used by SVM, re-introduced (functional!) fast > > emulation of IN. Piggyback SVM's work and use kvm_fast_pio_in() on > > VMX instead of performing full emulation of IN. > > > > Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> > There's a little code duplication that would be nice to avoid, moving > kvm_skip_emulated_instruction to kvm_fast_pio_{in,out}. Otherwise, What about adding kvm_fast_pio() to also avoid duplication of branching on 'in' versus 'out'? That seems worthy of a v2. > > Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> > > Paolo > > > > > --- > > arch/x86/kvm/vmx.c | 9 ++++++--- > > 1 file changed, 6 insertions(+), 3 deletions(-) > > > > diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c > > index 051dab74e4e9..c8a8391e95e5 100644 > > --- a/arch/x86/kvm/vmx.c > > +++ b/arch/x86/kvm/vmx.c > > @@ -6231,15 +6231,15 @@ static int handle_io(struct kvm_vcpu *vcpu) > > > > exit_qualification = vmcs_readl(EXIT_QUALIFICATION); > > string = (exit_qualification & 16) != 0; > > - in = (exit_qualification & 8) != 0; > > > > ++vcpu->stat.io_exits; > > > > - if (string || in) > > + if (string) > > return emulate_instruction(vcpu, 0) == EMULATE_DONE; > > > > port = exit_qualification >> 16; > > size = (exit_qualification & 7) + 1; > > + in = (exit_qualification & 8) != 0; > > > > ret = kvm_skip_emulated_instruction(vcpu); > > > > @@ -6247,7 +6247,10 @@ static int handle_io(struct kvm_vcpu *vcpu) > > * TODO: we might be squashing a KVM_GUESTDBG_SINGLESTEP-triggered > > * KVM_EXIT_DEBUG here. > > */ > > - return kvm_fast_pio_out(vcpu, size, port) && ret; > > + if (in) > > + return kvm_fast_pio_in(vcpu, size, port) && ret; > > + else > > + return kvm_fast_pio_out(vcpu, size, port) && ret; > > } > > > > static void > >
On 07/03/2018 16:35, Sean Christopherson wrote: > On Wed, 2018-03-07 at 16:19 +0100, Paolo Bonzini wrote: >> On 06/03/2018 21:58, Sean Christopherson wrote: >>> >>> Fast emulation of processor I/O for IN was disabled on x86 (both VMX >>> and SVM) some years ago due to a buggy implementation. The addition >>> of kvm_fast_pio_in(), used by SVM, re-introduced (functional!) fast >>> emulation of IN. Piggyback SVM's work and use kvm_fast_pio_in() on >>> VMX instead of performing full emulation of IN. >>> >>> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> >> There's a little code duplication that would be nice to avoid, moving >> kvm_skip_emulated_instruction to kvm_fast_pio_{in,out}. Otherwise, > > What about adding kvm_fast_pio() to also avoid duplication of branching > on 'in' versus 'out'? That seems worthy of a v2. If you would like to do that for a v2, that makes sense as well, indeed. Paolo >> >> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> >> >> Paolo >> >>> >>> --- >>> arch/x86/kvm/vmx.c | 9 ++++++--- >>> 1 file changed, 6 insertions(+), 3 deletions(-) >>> >>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c >>> index 051dab74e4e9..c8a8391e95e5 100644 >>> --- a/arch/x86/kvm/vmx.c >>> +++ b/arch/x86/kvm/vmx.c >>> @@ -6231,15 +6231,15 @@ static int handle_io(struct kvm_vcpu *vcpu) >>> >>> exit_qualification = vmcs_readl(EXIT_QUALIFICATION); >>> string = (exit_qualification & 16) != 0; >>> - in = (exit_qualification & 8) != 0; >>> >>> ++vcpu->stat.io_exits; >>> >>> - if (string || in) >>> + if (string) >>> return emulate_instruction(vcpu, 0) == EMULATE_DONE; >>> >>> port = exit_qualification >> 16; >>> size = (exit_qualification & 7) + 1; >>> + in = (exit_qualification & 8) != 0; >>> >>> ret = kvm_skip_emulated_instruction(vcpu); >>> >>> @@ -6247,7 +6247,10 @@ static int handle_io(struct kvm_vcpu *vcpu) >>> * TODO: we might be squashing a KVM_GUESTDBG_SINGLESTEP-triggered >>> * KVM_EXIT_DEBUG here. >>> */ >>> - return kvm_fast_pio_out(vcpu, size, port) && ret; >>> + if (in) >>> + return kvm_fast_pio_in(vcpu, size, port) && ret; >>> + else >>> + return kvm_fast_pio_out(vcpu, size, port) && ret; >>> } >>> >>> static void >>>
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 051dab74e4e9..c8a8391e95e5 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -6231,15 +6231,15 @@ static int handle_io(struct kvm_vcpu *vcpu) exit_qualification = vmcs_readl(EXIT_QUALIFICATION); string = (exit_qualification & 16) != 0; - in = (exit_qualification & 8) != 0; ++vcpu->stat.io_exits; - if (string || in) + if (string) return emulate_instruction(vcpu, 0) == EMULATE_DONE; port = exit_qualification >> 16; size = (exit_qualification & 7) + 1; + in = (exit_qualification & 8) != 0; ret = kvm_skip_emulated_instruction(vcpu); @@ -6247,7 +6247,10 @@ static int handle_io(struct kvm_vcpu *vcpu) * TODO: we might be squashing a KVM_GUESTDBG_SINGLESTEP-triggered * KVM_EXIT_DEBUG here. */ - return kvm_fast_pio_out(vcpu, size, port) && ret; + if (in) + return kvm_fast_pio_in(vcpu, size, port) && ret; + else + return kvm_fast_pio_out(vcpu, size, port) && ret; } static void
Fast emulation of processor I/O for IN was disabled on x86 (both VMX and SVM) some years ago due to a buggy implementation. The addition of kvm_fast_pio_in(), used by SVM, re-introduced (functional!) fast emulation of IN. Piggyback SVM's work and use kvm_fast_pio_in() on VMX instead of performing full emulation of IN. Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> --- arch/x86/kvm/vmx.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)