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;
Message-ID: <b56dd2e9f3934e24f08005b9c5588c54b4837ff6.camel@HansenPartnership.com>
Subject: Re: [PATCH v3 3/3] KEYS: trusted: Reserve TPM for seal and unseal
 operations
From:   James Bottomley <James.Bottomley@HansenPartnership.com>
To:     Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>,
        linux-integrity@vger.kernel.org
Cc:     David Howells <dhowells@redhat.com>,
        Mimi Zohar <zohar@linux.ibm.com>, stable@vger.kernel.org,
        Sumit Garg <sumit.garg@linaro.org>,
        kernel test robot <lkp@intel.com>,
        Peter Huewe <peterhuewe@gmx.de>,
        Jason Gunthorpe <jgg@ziepe.ca>, Arnd Bergmann <arnd@arndb.de>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        James Morris <jmorris@namei.org>,
        "Serge E. Hallyn" <serge@hallyn.com>,
        Jerry Snitselaar <jsnitsel@redhat.com>,
        Alexey Klimov <aklimov@redhat.com>,
        open list <linux-kernel@vger.kernel.org>,
        "open list:KEYS-TRUSTED" <keyrings@vger.kernel.org>,
        "open list:SECURITY SUBSYSTEM" 
        <linux-security-module@vger.kernel.org>
Date:   Mon, 12 Oct 2020 17:58:04 -0700
In-Reply-To: <20201013002815.40256-4-jarkko.sakkinen@linux.intel.com>
References: <20201013002815.40256-1-jarkko.sakkinen@linux.intel.com>
         <20201013002815.40256-4-jarkko.sakkinen@linux.intel.com>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.34.4 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Tue, 2020-10-13 at 03:28 +0300, Jarkko Sakkinen wrote:
[...]
> diff --git a/include/linux/tpm.h b/include/linux/tpm.h
> index 8f4ff39f51e7..f0ebce14d2f8 100644
> --- a/include/linux/tpm.h
> +++ b/include/linux/tpm.h
> @@ -397,6 +397,10 @@ static inline u32 tpm2_rc_value(u32 rc)
>  #if defined(CONFIG_TCG_TPM) || defined(CONFIG_TCG_TPM_MODULE)
>  
>  extern int tpm_is_tpm2(struct tpm_chip *chip);
> +extern __must_check int tpm_try_get_ops(struct tpm_chip *chip);
> +extern void tpm_put_ops(struct tpm_chip *chip);
> +extern ssize_t tpm_transmit_cmd(struct tpm_chip *chip, struct
> tpm_buf *buf,
> +				size_t min_rsp_body_length, const char
> *desc);
>  extern int tpm_pcr_read(struct tpm_chip *chip, u32 pcr_idx,
>  			struct tpm_digest *digest);
>  extern int tpm_pcr_extend(struct tpm_chip *chip, u32 pcr_idx,
> @@ -410,7 +414,18 @@ static inline int tpm_is_tpm2(struct tpm_chip
> *chip)
>  {
>  	return -ENODEV;
>  }
> -
> +static inline int tpm_try_get_ops(struct tpm_chip *chip)
> +{
> +	return -ENODEV;
> +}
> +static inline void tpm_put_ops(struct tpm_chip *chip)
> +{
> +}
> +static inline ssize_t tpm_transmit_cmd(struct tpm_chip *chip, struct
> tpm_buf *buf,
> +				       size_t min_rsp_body_length,
> const char *desc)
> +{
> +	return -ENODEV;
> +}
>  static inline int tpm_pcr_read(struct tpm_chip *chip, int pcr_idx,

I don't think we want this, do we?  That's only for API access which
should be available when the TPM isn't selected.  Given that get/put
are TPM critical operations, they should only appear when inside code
where the TPM has already been selected.  If they appear outside TPM
selected code, I think we want the compile to fail, which is why we
don't want these backup definitions.

James