Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 21 Jul 2022 00:38:24 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Maxim Levitsky <mlevitsk@redhat.com>
Cc:     kvm@vger.kernel.org, x86@kernel.org,
        Kees Cook <keescook@chromium.org>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        linux-kernel@vger.kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
        Borislav Petkov <bp@alien8.de>, Joerg Roedel <joro@8bytes.org>,
        Ingo Molnar <mingo@redhat.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>
Subject: Re: [PATCH v2 09/11] KVM: x86: emulator/smm: use smram struct for 64
 bit smram load/restore
Message-ID: <YtigALu4ccTi4/v0@google.com>
References: <20220621150902.46126-1-mlevitsk@redhat.com>
 <20220621150902.46126-10-mlevitsk@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220621150902.46126-10-mlevitsk@redhat.com>
Precedence: bulk

On Tue, Jun 21, 2022, Maxim Levitsky wrote:
> Use kvm_smram_state_64 struct to save/restore the 64 bit SMM state
> (used when X86_FEATURE_LM is present in the guest CPUID,
> regardless of 32-bitness of the guest).
> 
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> ---
> @@ -9814,7 +9805,7 @@ static void enter_smm(struct kvm_vcpu *vcpu)
>  	memset(buf, 0, 512);
>  #ifdef CONFIG_X86_64
>  	if (guest_cpuid_has(vcpu, X86_FEATURE_LM))
> -		enter_smm_save_state_64(vcpu, buf);
> +		enter_smm_save_state_64(vcpu, (struct kvm_smram_state_64 *)buf);
>  	else
>  #endif
>  		enter_smm_save_state_32(vcpu, (struct kvm_smram_state_32 *)buf);

Hrm, I _love_ the approach overall, but I really dislike having to cast an
arbitrary buffer, especially in the SVM code.

Aha!  Rather than keeping a buffer and casting, create a union to hold everything:

	union kvm_smram {
		struct kvm_smram_state_64 smram64;
		struct kvm_smram_state_32 smram32;
		u8 bytes[512];
	};

and then enter_smm() becomes:

  static void enter_smm(struct kvm_vcpu *vcpu)
  {
	struct kvm_segment cs, ds;
	struct desc_ptr dt;
	unsigned long cr0;

	union kvm_smram smram;

	BUILD_BUG_ON(sizeof(smram) != 512);

	memset(smram.bytes, 0, sizeof(smram));
#ifdef CONFIG_X86_64
	if (guest_cpuid_has(vcpu, X86_FEATURE_LM))
		enter_smm_save_state_64(vcpu, &smram.smram64);
	else
#endif
		enter_smm_save_state_32(vcpu, &smram.smram32);

	/*
	 * Give enter_smm() a chance to make ISA-specific changes to the vCPU
	 * state (e.g. leave guest mode) after we've saved the state into the
	 * SMM state-save area.
	 */
	static_call(kvm_x86_enter_smm)(vcpu, &smram);

	kvm_smm_changed(vcpu, true);
	kvm_vcpu_write_guest(vcpu, vcpu->arch.smbase + 0xfe00, smram.bytes, sizeof(smram));

and em_rsm() gets similar treatment.  Then the vendor code doesn't have to cast,
e.g. SVM can do:

	smram->smram64.svm_guest_flag = 1;
	smram->smram64.svm_guest_vmcb_gpa = svm->nested.vmcb12_gpa;

That way we don't have to refactor this all again if we want to use SMRAM to save
something on Intel for VMX (though I agree with Jim that that's probably a bad idea).