Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220627161123.1386853-1-pgonda@google.com> <Yrnync27TAhgSRUq@google.com>
In-Reply-To: <Yrnync27TAhgSRUq@google.com>
From:   Peter Gonda <pgonda@google.com>
Date:   Mon, 27 Jun 2022 12:39:41 -0600
Message-ID: <CAMkAt6q9iTsJO=UY_6588Zqa_rUjr5c01H5NsCj-4FiuTGnncw@mail.gmail.com>
Subject: Re: [PATCH] KVM: SEV: Clear the pages pointer in sev_unpin_memory
To:     Sean Christopherson <seanjc@google.com>
Cc:     kvm list <kvm@vger.kernel.org>, Greg Thelen <gthelen@google.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Tom Lendacky <thomas.lendacky@amd.com>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Jun 27, 2022 at 12:10 PM Sean Christopherson <seanjc@google.com> wrote:
>
> On Mon, Jun 27, 2022, Peter Gonda wrote:
> > Clear to the @pages array pointer in sev_unpin_memory to avoid leaving a
> > dangling pointer to invalid memory.
> >
> > Signed-off-by: Peter Gonda <pgonda@google.com>
> > Cc: Greg Thelen <gthelen@google.com>
> > Cc: Paolo Bonzini <pbonzini@redhat.com>
> > Cc: Sean Christopherson <seanjc@google.com>
> > Cc: Tom Lendacky <thomas.lendacky@amd.com>
> > Cc: kvm@vger.kernel.org
> > Cc: linux-kernel@vger.kernel.org
> > ---
> >  arch/x86/kvm/svm/sev.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> > index 309bcdb2f929..485ad86c01c6 100644
> > --- a/arch/x86/kvm/svm/sev.c
> > +++ b/arch/x86/kvm/svm/sev.c
> > @@ -452,6 +452,7 @@ static void sev_unpin_memory(struct kvm *kvm, struct page **pages,
> >       unpin_user_pages(pages, npages);
> >       kvfree(pages);
> >       sev->pages_locked -= npages;
> > +     *pages = NULL;
>
> Would this have helped detect a real bug?  I generally like cleaning up, but this
> leaves things in a somewhat inconsistent state, e.g. when unpinning a kvm_enc_region,
> pages will be NULL but npages will be non-zero.  It's somewhat moot because the
> region is immediately freed in that case, but that begs the question of what real
> benefit this provides.  sev_dbg_crypt() is the only flow where there's much danger
> of a use-after-free.
>

No strong opinion here, I just thought since this is a helper that
takes a 'struct page **pages" we may as well clear this. While there
are no bugs caught now if someone were to introduce something wrong
this would make it more clear.

We could update sev_unpin_memory() to take a int *npages so it can be
cleared as well. Since kvm_enc_region describes a region with u64
instead of pointers that seemed "safer" give you'd have to cast them
to dereference.

Totally fine with the NACK of course. =]