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

[David Hildenbrand <david@redhat.com>]

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@redhat.com>
> Suggested-by: Christian Borntraeger <borntraeger@de.ibm.com>
> ---
>  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