Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20211013165616.19846-1-pbonzini@redhat.com> <20211013165616.19846-2-pbonzini@redhat.com>
In-Reply-To: <20211013165616.19846-2-pbonzini@redhat.com>
From:   Marc Orr <marcorr@google.com>
Date:   Sun, 24 Oct 2021 18:31:01 -0700
Message-ID: <CAA03e5F8qvkbnPNvDHjrnM1hLs2fu5L_Mxtuhi3T5Y7u+_ydrw@mail.gmail.com>
Subject: Re: [PATCH 1/8] KVM: SEV-ES: fix length of string I/O
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        fwilhelm@google.com, seanjc@google.com, oupton@google.com,
        stable@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Oct 13, 2021 at 9:56 AM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> The size of the data in the scratch buffer is not divided by the size of
> each port I/O operation, so vcpu->arch.pio.count ends up being larger
> than it should be by a factor of size.
>
> Cc: stable@vger.kernel.org
> Fixes: 7ed9abfe8e9f ("KVM: SVM: Support string IO operations for an SEV-ES guest")
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  arch/x86/kvm/svm/sev.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index c36b5fe4c27c..e672493b5d8d 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -2583,7 +2583,7 @@ int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in)
>                 return -EINVAL;
>
>         return kvm_sev_es_string_io(&svm->vcpu, size, port,
> -                                   svm->ghcb_sa, svm->ghcb_sa_len, in);
> +                                   svm->ghcb_sa, svm->ghcb_sa_len / size, in);
>  }
>
>  void sev_es_init_vmcb(struct vcpu_svm *svm)
> --
> 2.27.0
>
>

I could be missing something, but I'm pretty sure that this is wrong.
The GHCB spec says that `exit_info_2` is the `rep` count. Not the
string length.

For example, given a `rep outsw` instruction, with `ECX` set to `8`,
the rep count written into `SW_EXITINFO2` should be eight x86 words
(i.e., 16 bytes) and the IO size should be one x86 word (i.e., 2
bytes). In other words, the code was correct before this patch. This
patch is incorrectly dividing the rep count by the IO size, causing
the string IO to be truncated.