Re: [PATCH 1/4] KVM: x86: Handle TIF_NEED_FPU_LOAD in kvm_{load,put}_guest_fpu()

From: Dave Hansen <dave.hansen@intel.com>
To: Sean Christopherson <sean.j.christopherson@intel.com>,
	Paolo Bonzini <pbonzini@redhat.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>,
	Wanpeng Li <wanpengli@tencent.com>,
	Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
	kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
	Derek Yerger <derek@djy.llc>,
	kernel@najdan.com, Thomas Lambertz <mail@thomaslambertz.de>,
	Rik van Riel <riel@surriel.com>,
	Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
	Borislav Petkov <bp@suse.de>,
	Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH 1/4] KVM: x86: Handle TIF_NEED_FPU_LOAD in kvm_{load,put}_guest_fpu()
Date: Fri, 17 Jan 2020 10:31:53 -0800	[thread overview]
Message-ID: <4d5dca91-8dbc-9ff3-b67a-2fa963da29cf@intel.com> (raw)
In-Reply-To: <20200117062628.6233-2-sean.j.christopherson@intel.com>

On 1/16/20 10:26 PM, Sean Christopherson wrote:
> Handle TIF_NEED_FPU_LOAD similar to how fpu__copy() handles the flag
> when duplicating FPU state to a new task struct.  TIF_NEED_FPU_LOAD can
> be set any time control is transferred out of KVM, be it voluntarily,
> e.g. if I/O is triggered during a KVM call to get_user_pages, or
> involuntarily, e.g. if softirq runs after an IRQ occurs.  Therefore,
> KVM must account for TIF_NEED_FPU_LOAD whenever it is (potentially)
> accessing CPU FPU state.
> 
> Fixes: 5f409e20b7945 ("x86/fpu: Defer FPU state load until return to userspace")
> Cc: stable@vger.kernel.org
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> ---
>  arch/x86/kvm/x86.c | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index cf917139de6b..0c7211491f98 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -8476,8 +8476,20 @@ static void kvm_load_guest_fpu(struct kvm_vcpu *vcpu)
>  {
>  	fpregs_lock();
>  
> -	copy_fpregs_to_fpstate(vcpu->arch.user_fpu);
> -	/* PKRU is separately restored in kvm_x86_ops->run.  */
> +	/*
> +	 * If userspace's FPU state is not resident in the CPU registers, just
> +	 * memcpy() from current, else save CPU state directly to user_fpu.
> +	 */
> +	if (test_thread_flag(TIF_NEED_FPU_LOAD))
> +		memcpy(&vcpu->arch.user_fpu->state, &current->thread.fpu.state,
> +		       fpu_kernel_xstate_size);
> +	else
> +		copy_fpregs_to_fpstate(vcpu->arch.user_fpu);
> +
> +	/*
> +	 * Load guest's FPU state to the CPU registers.  PKRU is separately
> +	 * loaded in kvm_x86_ops->run.
> +	 */
>  	__copy_kernel_to_fpregs(&vcpu->arch.guest_fpu->state,
>  				~XFEATURE_MASK_PKRU);

Nit: it took me a minute to realize that there is both:

	vcpu->arch.user_fpu
and
	vcpu->arch.guest_fpu

It might help readability to have local variables for those, or at least
a comment to help differentiate the two.

> @@ -8492,7 +8504,16 @@ static void kvm_put_guest_fpu(struct kvm_vcpu *vcpu)
>  {
>  	fpregs_lock();
>  
> -	copy_fpregs_to_fpstate(vcpu->arch.guest_fpu);
> +	/*
> +	 * If guest's FPU state is not resident in the CPU registers, just
> +	 * memcpy() from current, else save CPU state directly to guest_fpu.
> +	 */
> +	if (test_thread_flag(TIF_NEED_FPU_LOAD))
> +		memcpy(&vcpu->arch.guest_fpu->state, &current->thread.fpu.state,
> +		       fpu_kernel_xstate_size);
> +	else
> +		copy_fpregs_to_fpstate(vcpu->arch.guest_fpu);
> +
>  	copy_kernel_to_fpregs(&vcpu->arch.user_fpu->state);
>  
>  	fpregs_mark_activate();

This also makes me wonder if we want to have copy_fpregs_to_fpstate()
check for TIF_NEED_FPU_LOAD and complain if it's set.