Date:   Tue, 7 Dec 2021 21:24:54 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Peter Gonda <pgonda@google.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Marc Orr <marcorr@google.com>
Subject: Re: [PATCH] selftests: sev_migrate_tests: Fix sev_ioctl()
Message-ID: <Ya/RJiTOQjJ+fj73@google.com>
References: <20211207201034.1392660-1-pgonda@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20211207201034.1392660-1-pgonda@google.com>
Precedence: bulk

On Tue, Dec 07, 2021, Peter Gonda wrote:
> TEST_ASSERT in SEV ioctl was allowing errors because it checked return
> value was good OR the FW error code was OK. This TEST_ASSERT should
> require both (aka. AND) values are OK. Removes the LAUNCH_START from the
> mirror VM because this call correctly fails because mirror VMs cannot
> call this command.

This probably should be two separate patches.  First remove the bogus LAUNCH_START
call, then fix the assert.
 
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Sean Christopherson <seanjc@google.com>
> Cc: Marc Orr <marcorr@google.com>
> Signed-off-by: Peter Gonda <pgonda@google.com>
> ---
>  tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c | 7 ++-----
>  1 file changed, 2 insertions(+), 5 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c b/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c
> index 29b18d565cf4..8e1b1e737cb1 100644
> --- a/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c
> +++ b/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c
> @@ -31,7 +31,7 @@ static void sev_ioctl(int vm_fd, int cmd_id, void *data)
>  	int ret;
>  
>  	ret = ioctl(vm_fd, KVM_MEMORY_ENCRYPT_OP, &cmd);
> -	TEST_ASSERT((ret == 0 || cmd.error == SEV_RET_SUCCESS),
> +	TEST_ASSERT(ret == 0 && cmd.error == SEV_RET_SUCCESS,
>  		    "%d failed: return code: %d, errno: %d, fw error: %d",
>  		    cmd_id, ret, errno, cmd.error);

Hmm, reading cmd.error could also consume uninitialized data, e.g. if the ioctl()
fails before getting into the PSP command, the error message will dump garbage.

And theoretically this could get a false negative if the test stack happens to have
'0' for cmd.error and KVM neglects to fill cmd.error when the ioctl() succeeds.

So in additional to fixing the assert itself, I vote we also do:

diff --git a/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c b/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c
index 29b18d565cf4..50132e165a8d 100644
--- a/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c
+++ b/tools/testing/selftests/kvm/x86_64/sev_migrate_tests.c
@@ -26,6 +26,7 @@ static void sev_ioctl(int vm_fd, int cmd_id, void *data)
        struct kvm_sev_cmd cmd = {
                .id = cmd_id,
                .data = (uint64_t)data,
+               .error = -1u,
                .sev_fd = open_sev_dev_path_or_exit(),
        };
        int ret;