Re: [PATCH] x86,kvm: move qemu/guest FPU switching out to vcpu_run

From: David Hildenbrand
Date: Tue Nov 14 2017 - 11:58:18 EST


On 14.11.2017 06:12, Rik van Riel wrote:
> Currently, every time a VCPU is scheduled out, the host kernel will
> first save the guest FPU/xstate context, then load the qemu userspace
> FPU context, only to then immediately save the qemu userspace FPU
> context back to memory. When scheduling in a VCPU, the same extraneous
> FPU loads and saves are done.
>
> This could be avoided by moving from a model where the guest FPU is
> loaded and stored with preemption disabled, to a model where the
> qemu userspace FPU is swapped out for the guest FPU context for
> the duration of the KVM_RUN ioctl.
>
> This is done under the VCPU mutex, which is also taken when other
> tasks inspect the VCPU FPU context, so the code should already be
> safe for this change. That should come as no surprise, given that
> s390 already has this optimization.
>
> No performance changes were detected in quick ping-pong tests on
> my 4 socket system, which is expected since an FPU+xstate load is
> on the order of 0.1us, while ping-ponging between CPUs is on the
> order of 20us, and somewhat noisy.
>
> There may be other tests where performance changes are noticeable.
>
> Signed-off-by: Rik van Riel <riel@xxxxxxxxxx>
> Suggested-by: Christian Borntraeger <borntraeger@xxxxxxxxxx>
> ---
> arch/x86/include/asm/kvm_host.h | 13 +++++++++++++
> arch/x86/kvm/x86.c | 29 ++++++++++++-----------------
> 2 files changed, 25 insertions(+), 17 deletions(-)
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index c73e493adf07..92e66685249e 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h

We should also get rid of guest_fpu_loaded now, right?


> @@ -536,7 +536,20 @@ struct kvm_vcpu_arch {
> struct kvm_mmu_memory_cache mmu_page_cache;
> struct kvm_mmu_memory_cache mmu_page_header_cache;
>
> + /*
> + * QEMU userspace and the guest each have their own FPU state.
> + * In vcpu_run, we switch between the user and guest FPU contexts.
> + * While running a VCPU, the VCPU thread will have the guest FPU
> + * context.
> + *
> + * Note that while the PKRU state lives inside the fpu registers,
> + * it is switched out separately at VMENTER and VMEXIT time. The
> + * "guest_fpu" state here contains the guest FPU context, with the
> + * host PRKU bits.
> + */
> + struct fpu user_fpu;
> struct fpu guest_fpu;
> +
> u64 xcr0;
> u64 guest_supported_xcr0;
> u32 guest_xstate_size;
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 03869eb7fcd6..59912b20a830 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -2917,7 +2917,6 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
> srcu_read_unlock(&vcpu->kvm->srcu, idx);
> pagefault_enable();
> kvm_x86_ops->vcpu_put(vcpu);
> - kvm_put_guest_fpu(vcpu);
> vcpu->arch.last_host_tsc = rdtsc();
> }
>
> @@ -6908,7 +6907,6 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> preempt_disable();
>
> kvm_x86_ops->prepare_guest_switch(vcpu);
> - kvm_load_guest_fpu(vcpu);
>
> /*
> * Disable IRQs before setting IN_GUEST_MODE. Posted interrupt
> @@ -7095,6 +7093,8 @@ static int vcpu_run(struct kvm_vcpu *vcpu)
>
> vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
>
> + kvm_load_guest_fpu(vcpu);
> +
> for (;;) {
> if (kvm_vcpu_running(vcpu)) {
> r = vcpu_enter_guest(vcpu);
> @@ -7132,6 +7132,8 @@ static int vcpu_run(struct kvm_vcpu *vcpu)
> }
> }
>
> + kvm_put_guest_fpu(vcpu);
> +
> srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
>
> return r;
> @@ -7663,32 +7665,25 @@ static void fx_init(struct kvm_vcpu *vcpu)
> vcpu->arch.cr0 |= X86_CR0_ET;
> }
>
> +/* Swap (qemu) user FPU context for the guest FPU context. */
> void kvm_load_guest_fpu(struct kvm_vcpu *vcpu)
> {
> - if (vcpu->guest_fpu_loaded)
> - return;
> -
> - /*
> - * Restore all possible states in the guest,
> - * and assume host would use all available bits.
> - * Guest xcr0 would be loaded later.
> - */
> - vcpu->guest_fpu_loaded = 1;
> - __kernel_fpu_begin();
> + preempt_disable();
> + copy_fpregs_to_fpstate(&vcpu->arch.user_fpu);
> /* PKRU is separately restored in kvm_x86_ops->run. */
> __copy_kernel_to_fpregs(&vcpu->arch.guest_fpu.state,
> ~XFEATURE_MASK_PKRU);
> + preempt_enable();
> trace_kvm_fpu(1);
> }
>
> +/* When vcpu_run ends, restore user space FPU context. */
> void kvm_put_guest_fpu(struct kvm_vcpu *vcpu)
> {
> - if (!vcpu->guest_fpu_loaded)
> - return;
> -
> - vcpu->guest_fpu_loaded = 0;
> + preempt_disable();
> copy_fpregs_to_fpstate(&vcpu->arch.guest_fpu);
> - __kernel_fpu_end();
> + copy_kernel_to_fpregs(&vcpu->arch.user_fpu.state);
> + preempt_enable();
> ++vcpu->stat.fpu_reload;
> trace_kvm_fpu(0);
> }
>

emulator_get_fpu() does a kvm_load_guest_fpu(). Doesn't that mean that
this is now not needed anymore? (at least when emulator code is called
from inside the loop?)

Also, what about preempt_diable() at that point, still needed?


--

Thanks,

David / dhildenb