Re: [PATCH v5 11/11] PM: hibernate: seal the encryption key with a PCR policy

From: Eric Biggers <ebiggers@kernel.org>
To: Evan Green <evgreen@chromium.org>
Cc: linux-kernel@vger.kernel.org, corbet@lwn.net,
	linux-integrity@vger.kernel.org, gwendal@chromium.org,
	dianders@chromium.org, apronin@chromium.org,
	Pavel Machek <pavel@ucw.cz>, Ben Boeckel <me@benboeckel.net>,
	rjw@rjwysocki.net, jejb@linux.ibm.com,
	Kees Cook <keescook@chromium.org>,
	dlunev@google.com, zohar@linux.ibm.com,
	Matthew Garrett <mgarrett@aurora.tech>,
	jarkko@kernel.org, linux-pm@vger.kernel.org,
	Matthew Garrett <mjg59@google.com>,
	Len Brown <len.brown@intel.com>,
	"Rafael J. Wysocki" <rafael@kernel.org>, axelj <axelj@axis.com>
Subject: Re: [PATCH v5 11/11] PM: hibernate: seal the encryption key with a PCR policy
Date: Sun, 13 Nov 2022 15:51:04 -0800	[thread overview]
Message-ID: <Y3GC6M6umF+MOu1f@sol.localdomain> (raw)
In-Reply-To: <20221111151451.v5.11.Ifce072ae1ef1ce39bd681fff55af13a054045d9f@changeid>

On Fri, Nov 11, 2022 at 03:16:36PM -0800, Evan Green wrote:
> +static int tpm_setup_policy(struct tpm_chip *chip, int *session_handle)
> +{
> +	struct tpm_header *head;
> +	struct tpm_buf buf;
> +	char nonce[32] = {0x00};
> +	int rc;
> +
> +	rc = tpm_buf_init(&buf, TPM2_ST_NO_SESSIONS,
> +			  TPM2_CC_START_AUTH_SESSION);
> +	if (rc)
> +		return rc;
> +
> +	/* Decrypt key */
> +	tpm_buf_append_u32(&buf, TPM2_RH_NULL);
> +
> +	/* Auth entity */
> +	tpm_buf_append_u32(&buf, TPM2_RH_NULL);
> +
> +	/* Nonce - blank is fine here */
> +	tpm_buf_append_u16(&buf, sizeof(nonce));
> +	tpm_buf_append(&buf, nonce, sizeof(nonce));

In general, hardcoded nonces are a huge red flag.  If it's fine here, it would
be helpful to leave a comment explaining why that is.

> +	rc = tpm_send(chip, buf.data, tpm_buf_length(&buf));
> +	if (rc)
> +		goto out;

This is another instance of the bug where TPM2_RC_* codes are being returned
from a function that is expected to return -errno values.

> +	*session_handle = be32_to_cpu(*(__be32 *)&buf.data[10]);

get_unaligned_be32, to avoid an unaligned memory access.

> @@ -497,11 +602,16 @@ static int snapshot_setup_encryption_common(struct snapshot_data *data)
>  static int snapshot_create_kernel_key(struct snapshot_data *data)
>  {
>  	/* Create a key sealed by the SRK. */
> -	char *keyinfo = "new\t32\tkeyhandle=0x81000000\tcreationpcrs=0x00800000";
> +	const char *keytemplate =
> +		"new\t32\tkeyhandle=0x81000000\tcreationpcrs=0x00800000\tpolicydigest=%s";
>  	const struct cred *cred = current_cred();
>  	struct tpm_digest *digests = NULL;
> +	char policy[SHA256_DIGEST_SIZE];
> +	char *policydigest = NULL;
> +	int session_handle = -1;
>  	struct key *key = NULL;
>  	struct tpm_chip *chip;
> +	char *keyinfo = NULL;
>  	int ret, i;
>  
>  	chip = tpm_default_chip();
> @@ -534,6 +644,28 @@ static int snapshot_create_kernel_key(struct snapshot_data *data)
>  	if (ret != 0)
>  		goto out;
>  
> +	policydigest = kmalloc(SHA256_DIGEST_SIZE * 2 + 1, GFP_KERNEL);
> +	if (!policydigest) {
> +		ret = -ENOMEM;
> +		goto out;
> +	}
> +
> +	ret = tpm_setup_policy(chip, &session_handle);
> +	if (ret != 0)
> +		goto out;
> +
> +	ret = tpm_policy_get_digest(chip, session_handle, policy);
> +	if (ret != 0)
> +		goto out;
> +
> +	bin2hex(policydigest, policy, SHA256_DIGEST_SIZE);
> +	policydigest[SHA256_DIGEST_SIZE * 2] = '\0';
> +	keyinfo = kasprintf(GFP_KERNEL, keytemplate, policydigest);
> +	if (!keyinfo) {
> +		ret = -ENOMEM;
> +		goto out;
> +	}

With the %*phN format specifier, there would be no need for bin2hex().

- Eric