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: <20211025162517.2152628-1-pbonzini@redhat.com>
In-Reply-To: <20211025162517.2152628-1-pbonzini@redhat.com>
From:   Marc Orr <marcorr@google.com>
Date:   Mon, 25 Oct 2021 13:55:58 -0700
Message-ID: <CAA03e5E+tP30B1DOgwZQGRTMmXXTwKUFHJ6HNFk-jMRgxKMF+A@mail.gmail.com>
Subject: Re: [PATCH] KVM: SEV-ES: fix another issue with string I/O VMGEXITs
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        stable@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Oct 25, 2021 at 9:25 AM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> If the guest requests string I/O from the hypervisor via VMGEXIT,
> SW_EXITINFO2 will contain the REP count.  However, sev_es_string_io
> was incorrectly treating it as the size of the GHCB buffer in
> bytes.
>
> This fixes the "outsw" test in the experimental SEV tests of
> kvm-unit-tests.
>
> Cc: stable@vger.kernel.org
> Fixes: 7ed9abfe8e9f ("KVM: SVM: Support string IO operations for an SEV-ES guest")
> Reported-by: Marc Orr <marcorr@google.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  arch/x86/kvm/svm/sev.c | 11 ++++++++---
>  1 file changed, 8 insertions(+), 3 deletions(-)
>
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index e672493b5d8d..12d29d669cbc 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -2579,11 +2579,16 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu)
>
>  int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in)
>  {
> -       if (!setup_vmgexit_scratch(svm, in, svm->vmcb->control.exit_info_2))
> +       u32 bytes;
> +
> +       if (unlikely(check_mul_overflow(svm->vmcb->control.exit_info_2, size, &bytes)))
> +               return -EINVAL;

Maybe this is only an issue on our internal setup, but when I tested
this I found that `check_mul_overflow()` is very particular that all
three args having the same type. Therefore, to get it to compile, I
changed all of the arguments to match the type of `exit_info_2`, like
so:

u64 bytes;

if (unlikely(check_mul_overflow(svm->vmcb->control.exit_info_2,
(u64)size, &bytes)))

> +
> +       if (!setup_vmgexit_scratch(svm, in, bytes))
>                 return -EINVAL;
>
> -       return kvm_sev_es_string_io(&svm->vcpu, size, port,
> -                                   svm->ghcb_sa, svm->ghcb_sa_len / size, in);
> +       return kvm_sev_es_string_io(&svm->vcpu, size, port, svm->ghcb_sa,
> +                                   svm->vmcb->control.exit_info_2, in);
>  }
>
>  void sev_es_init_vmcb(struct vcpu_svm *svm)
> --
> 2.27.0
>

I've tested that this works. To test it, I (temporarily) modified
`setup_vmgexit_scratch()` to always treat the scratch area as if it
resides outside of the GHCB page (i.e., always go down the `else` code
path). Doing this, I was able to see the stringio test case fail
before this patch and pass with it.

Reviewed-by: Marc Orr <marcorr@google.com>
Tested-by: Marc Orr <marcorr@google.com>