linux-kernel - Re: [PATCH 2/6] KVM: SEV: Explicitly document that there are no TOCTOU races in copy ASID

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 16 Nov 2021 12:38:28 +0100
From:   Paolo Bonzini <pbonzini@...hat.com>
To:     Sean Christopherson <seanjc@...gle.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>,
        Marc Orr <marcorr@...gle.com>,
        Nathan Tempelman <natet@...gle.com>,
        Brijesh Singh <brijesh.singh@....com>,
        Tom Lendacky <thomas.lendacky@....com>
Subject: Re: [PATCH 2/6] KVM: SEV: Explicitly document that there are no
 TOCTOU races in copy ASID

On 11/9/21 22:50, Sean Christopherson wrote:
> Deliberately grab the source's SEV info for COPY_ENC_CONTEXT_FROM outside
> of kvm->lock and document that doing so is safe due to SEV/SEV-ES info,
> e.g. ASID, active, etc... being "write-once" and set atomically with
> respect to kvm->lock.
> 
> No functional change intended.

This isn't true anymore with the move-enc-context-from patches, though! 
  I will send a follow-up shortly.

Paolo

> Signed-off-by: Sean Christopherson <seanjc@...gle.com>
> ---
>   arch/x86/kvm/svm/sev.c | 24 ++++++++++++++++--------
>   1 file changed, 16 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index eeec499e4372..6d14e2595c96 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -1737,9 +1737,9 @@ int svm_unregister_enc_region(struct kvm *kvm,
>   
>   int svm_vm_copy_asid_from(struct kvm *kvm, unsigned int source_fd)
>   {
> +	struct kvm_sev_info *mirror_sev, *source_sev;
>   	struct file *source_kvm_file;
>   	struct kvm *source_kvm;
> -	struct kvm_sev_info source_sev, *mirror_sev;
>   	int ret;
>   
>   	source_kvm_file = fget(source_fd);
> @@ -1762,9 +1762,6 @@ int svm_vm_copy_asid_from(struct kvm *kvm, unsigned int source_fd)
>   		goto e_source_unlock;
>   	}
>   
> -	memcpy(&source_sev, &to_kvm_svm(source_kvm)->sev_info,
> -	       sizeof(source_sev));
> -
>   	/*
>   	 * The mirror kvm holds an enc_context_owner ref so its asid can't
>   	 * disappear until we're done with it
> @@ -1785,14 +1782,25 @@ int svm_vm_copy_asid_from(struct kvm *kvm, unsigned int source_fd)
>   		goto e_mirror_unlock;
>   	}
>   
> +	/*
> +	 * Referencing the source's sev_info without holding the source's lock
> +	 * is safe as SEV/SEV-ES activation is a one-way, "atomic" operation.
> +	 * SEV state, e.g. the ASID, is modified under kvm->lock, and cannot be
> +	 * changed after SEV is marked active (here or in normal activation).
> +	 * That same atomicity also prevents TOC-TOU issues with respect to
> +	 * related sanity checks on source_kvm.
> +	 */
> +	source_sev = &to_kvm_svm(source_kvm)->sev_info;
> +
>   	/* Set enc_context_owner and copy its encryption context over */
>   	mirror_sev = &to_kvm_svm(kvm)->sev_info;
>   	mirror_sev->enc_context_owner = source_kvm;
> +	mirror_sev->asid = source_sev->asid;
>   	mirror_sev->active = true;
> -	mirror_sev->asid = source_sev.asid;
> -	mirror_sev->fd = source_sev.fd;
> -	mirror_sev->es_active = source_sev.es_active;
> -	mirror_sev->handle = source_sev.handle;
> +	mirror_sev->asid = source_sev->asid;
> +	mirror_sev->fd = source_sev->fd;
> +	mirror_sev->es_active = source_sev->es_active;
> +	mirror_sev->handle = source_sev->handle;
>   	/*
>   	 * Do not copy ap_jump_table. Since the mirror does not share the same
>   	 * KVM contexts as the original, and they may have different
>