| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 7 May 2021 18:21:37 -0500
From: Tom Lendacky <thomas.lendacky@....com>
To: Sean Christopherson <seanjc@...gle.com>,
Paolo Bonzini <pbonzini@...hat.com>
Cc: Vitaly Kuznetsov <vkuznets@...hat.com>,
Wanpeng Li <wanpengli@...cent.com>,
Jim Mattson <jmattson@...gle.com>,
Joerg Roedel <joro@...tes.org>, kvm@...r.kernel.org,
linux-kernel@...r.kernel.org, Peter Gonda <pgonda@...gle.com>,
Maxim Levitsky <mlevitsk@...hat.com>
Subject: Re: [PATCH 2/2] KVM: x86: Allow userspace to update tracked sregs for
protected guests
On 5/7/21 11:59 AM, Sean Christopherson wrote:
> Allow userspace to set CR0, CR4, CR8, and EFER via KVM_SET_SREGS for
> protected guests, e.g. for SEV-ES guests with an encrypted VMSA. KVM
> tracks the aforementioned registers by trapping guest writes, and also
> exposes the values to userspace via KVM_GET_SREGS. Skipping the regs
> in KVM_SET_SREGS prevents userspace from updating KVM's CPU model to
> match the known hardware state.
This is very similar to the original patch I had proposed that you were
against :)
I'm assuming it's meant to make live migration a bit easier?
>
> Fixes: 5265713a0737 ("KVM: x86: Update __get_sregs() / __set_sregs() to support SEV-ES")
> Reported-by: Peter Gonda <pgonda@...gle.com>
> Cc: stable@...r.kernel.org
> Cc: Maxim Levitsky <mlevitsk@...hat.com>
> Signed-off-by: Sean Christopherson <seanjc@...gle.com>
Acked-by: Tom Lendacky <thomas.lendacky@....com>
> ---
> arch/x86/kvm/x86.c | 73 ++++++++++++++++++++++++++--------------------
> 1 file changed, 42 insertions(+), 31 deletions(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 3bf52ba5f2bb..1b7d0e97c82b 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -9963,21 +9963,25 @@ static int __set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
> if (kvm_set_apic_base(vcpu, &apic_base_msr))
> goto out;
>
> - if (vcpu->arch.guest_state_protected)
> - goto skip_protected_regs;
> + if (!vcpu->arch.guest_state_protected) {
> + dt.size = sregs->idt.limit;
> + dt.address = sregs->idt.base;
> + static_call(kvm_x86_set_idt)(vcpu, &dt);
> + dt.size = sregs->gdt.limit;
> + dt.address = sregs->gdt.base;
> + static_call(kvm_x86_set_gdt)(vcpu, &dt);
>
> - dt.size = sregs->idt.limit;
> - dt.address = sregs->idt.base;
> - static_call(kvm_x86_set_idt)(vcpu, &dt);
> - dt.size = sregs->gdt.limit;
> - dt.address = sregs->gdt.base;
> - static_call(kvm_x86_set_gdt)(vcpu, &dt);
> -
> - vcpu->arch.cr2 = sregs->cr2;
> - mmu_reset_needed |= kvm_read_cr3(vcpu) != sregs->cr3;
> - vcpu->arch.cr3 = sregs->cr3;
> - kvm_register_mark_available(vcpu, VCPU_EXREG_CR3);
> + vcpu->arch.cr2 = sregs->cr2;
> + mmu_reset_needed |= kvm_read_cr3(vcpu) != sregs->cr3;
> + vcpu->arch.cr3 = sregs->cr3;
> + kvm_register_mark_available(vcpu, VCPU_EXREG_CR3);
> + }
>
> + /*
> + * Writes to CR0, CR4, CR8, and EFER are trapped (after the instruction
> + * completes) for SEV-EV guests, thus userspace is allowed to set them
> + * so that KVM's model can be updated to mirror hardware state.
> + */
> kvm_set_cr8(vcpu, sregs->cr8);
>
> mmu_reset_needed |= vcpu->arch.efer != sregs->efer;
> @@ -9990,35 +9994,42 @@ static int __set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
> mmu_reset_needed |= kvm_read_cr4(vcpu) != sregs->cr4;
> static_call(kvm_x86_set_cr4)(vcpu, sregs->cr4);
>
> - idx = srcu_read_lock(&vcpu->kvm->srcu);
> - if (is_pae_paging(vcpu)) {
> + /*
> + * PDPTEs, like regular PTEs, are always encrypted, thus reading them
> + * will return garbage. Shadow paging, including nested NPT, isn't
> + * compatible with protected guests, so ignoring the PDPTEs is a-ok.
> + */
> + if (!vcpu->arch.guest_state_protected && is_pae_paging(vcpu)) {
> + idx = srcu_read_lock(&vcpu->kvm->srcu);
> load_pdptrs(vcpu, vcpu->arch.walk_mmu, kvm_read_cr3(vcpu));
> + srcu_read_unlock(&vcpu->kvm->srcu, idx);
> +
> mmu_reset_needed = 1;
> }
> - srcu_read_unlock(&vcpu->kvm->srcu, idx);
>
> if (mmu_reset_needed)
> kvm_mmu_reset_context(vcpu);
>
> - kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
> - kvm_set_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
> - kvm_set_segment(vcpu, &sregs->es, VCPU_SREG_ES);
> - kvm_set_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
> - kvm_set_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
> - kvm_set_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
> + if (!vcpu->arch.guest_state_protected) {
> + kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
> + kvm_set_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
> + kvm_set_segment(vcpu, &sregs->es, VCPU_SREG_ES);
> + kvm_set_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
> + kvm_set_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
> + kvm_set_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
>
> - kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
> - kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
> + kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
> + kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
>
> - update_cr8_intercept(vcpu);
> + update_cr8_intercept(vcpu);
>
> - /* Older userspace won't unhalt the vcpu on reset. */
> - if (kvm_vcpu_is_bsp(vcpu) && kvm_rip_read(vcpu) == 0xfff0 &&
> - sregs->cs.selector == 0xf000 && sregs->cs.base == 0xffff0000 &&
> - !is_protmode(vcpu))
> - vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
> + /* Older userspace won't unhalt the vcpu on reset. */
> + if (kvm_vcpu_is_bsp(vcpu) && kvm_rip_read(vcpu) == 0xfff0 &&
> + sregs->cs.selector == 0xf000 &&
> + sregs->cs.base == 0xffff0000 && !is_protmode(vcpu))
> + vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
> + }
>
> -skip_protected_regs:
> max_bits = KVM_NR_INTERRUPTS;
> pending_vec = find_first_bit(
> (const unsigned long *)sregs->interrupt_bitmap, max_bits);
>
Powered by blists - more mailing lists