Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
MIME-Version: 1.0
References: <20220414162325.1830014-1-jackyli@google.com> <fe209f25-6119-cc26-2608-8aaccf11ac36@amd.com>
In-Reply-To: <fe209f25-6119-cc26-2608-8aaccf11ac36@amd.com>
From:   Peter Gonda <pgonda@google.com>
Date:   Fri, 15 Apr 2022 10:20:40 -0600
Message-ID: <CAMkAt6pyDQaqLCRzJWow6hgtyMrAtarPri9yUWBgMPPuucm77A@mail.gmail.com>
Subject: Re: [PATCH v2] crypto: ccp - Fix the INIT_EX data file open failure
To:     Tom Lendacky <thomas.lendacky@amd.com>
Cc:     Jacky Li <jackyli@google.com>,
        Brijesh Singh <brijesh.singh@amd.com>,
        John Allen <john.allen@amd.com>,
        Herbert Xu <herbert@gondor.apana.org.au>,
        "David S. Miller" <davem@davemloft.net>,
        Marc Orr <marcorr@google.com>, Alper Gun <alpergun@google.com>,
        Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Apr 15, 2022 at 7:49 AM Tom Lendacky <thomas.lendacky@amd.com> wrote:
>
> On 4/14/22 11:23, Jacky Li wrote:
> > There are 2 common cases when INIT_EX data file might not be
> > opened successfully and fail the sev initialization:
> >
> > 1. In user namespaces, normal user tasks (e.g. VMM) can change their
> >     current->fs->root to point to arbitrary directories. While
> >     init_ex_path is provided as a module param related to root file
> >     system. Solution: use the root directory of init_task to avoid
> >     accessing the wrong file.
> >
> > 2. Normal user tasks (e.g. VMM) don't have the privilege to access
> >     the INIT_EX data file. Solution: open the file as root and
> >     restore permissions immediately.
> >
> > Fixes: 3d725965f836 ("crypto: ccp - Add SEV_INIT_EX support")
> > Signed-off-by: Jacky Li <jackyli@google.com>
> > Reviewed-by: Peter Gonda <pgonda@google.com>
>
> Looks good, just a quick question. Should there be any type of access
> checks before switching credentials? Should we check access to /dev/sev or
> such? Or is the capability to load the module enough?

I thought this was fine because regardless of if an admin sets
psp_init_on_probe=true or false, their intention is that people who
have rw access to /dev/sev can use the commands which require the PSP
to be init. In the case of psp_init_on_probe=false only rw users can
cause the file to be created. The case of psp_init_on_probe=true seems
a little less clear to me but if a user can modprobe ccp that seems
like sufficient privilege to create the file. What do you think, Tom?