Message-ID: <c6a3f3e4-bd9d-d060-9ee1-f84310857ba0@huawei.com>
Date:   Thu, 17 Jan 2019 08:59:00 +0100
From:   Roberto Sassu <roberto.sassu@...wei.com>
To:     Jarkko Sakkinen <jarkko.sakkinen@...ux.intel.com>
CC:     <zohar@...ux.ibm.com>, <david.safford@...com>,
        <monty.wiseman@...com>, <linux-integrity@...r.kernel.org>,
        <linux-security-module@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <silviu.vlasceanu@...wei.com>
Subject: Re: [PATCH v7 5/5] tpm: pass an array of tpm_extend_digest structures
 to tpm_pcr_extend()

On 12/20/2018 4:21 PM, Jarkko Sakkinen wrote:
> On Thu, Dec 13, 2018 at 11:29:45AM +0100, Roberto Sassu wrote:
>> Currently, tpm_pcr_extend() accepts as an input only a SHA1 digest.
>>
>> This patch modifies the definition of tpm_pcr_extend() to allow other
>> kernel subsystems to pass a digest for each algorithm supported by the TPM.
>> All digests are processed by the TPM in one operation.
>>
>> If a tpm_pcr_extend() caller provides a subset of the supported algorithms,
>> the TPM driver extends the remaining PCR banks with the first digest
>> passed as an argument to the function.
>>
>> The new tpm_extend digest structure has been preferred to the tpm_digest
>> structure, to let the caller specify the size of the digest (which may be
>> unknown to the TPM driver).
>>
>> Due to the API change, ima_pcr_extend() and pcrlock() have been modified.
>>
>> Signed-off-by: Roberto Sassu <roberto.sassu@...wei.com>
>> ---
>>   drivers/char/tpm/tpm-interface.c   | 24 +++++---------------
>>   drivers/char/tpm/tpm.h             |  5 +++--
>>   drivers/char/tpm/tpm1-cmd.c        | 13 ++++++++---
>>   drivers/char/tpm/tpm2-cmd.c        | 35 +++++++++++++++++++++---------
>>   include/linux/tpm.h                | 13 ++++++++---
>>   security/integrity/ima/ima_queue.c |  5 ++++-
>>   security/keys/trusted.c            |  5 ++++-
>>   7 files changed, 62 insertions(+), 38 deletions(-)
>>
>> diff --git a/drivers/char/tpm/tpm-interface.c b/drivers/char/tpm/tpm-interface.c
>> index eb7c79ca8a94..911fea19e408 100644
>> --- a/drivers/char/tpm/tpm-interface.c
>> +++ b/drivers/char/tpm/tpm-interface.c
>> @@ -478,42 +478,30 @@ EXPORT_SYMBOL_GPL(tpm_pcr_read);
>>    * tpm_pcr_extend - extend a PCR value in SHA1 bank.
>>    * @chip:	a &struct tpm_chip instance, %NULL for the default chip
>>    * @pcr_idx:	the PCR to be retrieved
>> - * @hash:	the hash value used to extend the PCR value
>> + * @count:	number of tpm_extend_digest structures
>> + * @digests:	array of tpm_extend_digest structures used to extend PCRs
>>    *
>>    * Note: with TPM 2.0 extends also those banks for which no digest was
>>    * specified in order to prevent malicious use of those PCR banks.
>>    *
>>    * Return: same as with tpm_transmit_cmd()
>>    */
>> -int tpm_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, const u8 *hash)
>> +int tpm_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, u32 count,
>> +		   const struct tpm_extend_digest *digests)
> 
> Remove const. Document how @digests is used  like the special meaning
> of the first index. I faintly remember asking this last time.
> 
>>   {
>>   	int rc;
>> -	struct tpm_digest *digest_list;
>> -	int i;
>>   
>>   	chip = tpm_find_get_ops(chip);
>>   	if (!chip)
>>   		return -ENODEV;
>>   
>>   	if (chip->flags & TPM_CHIP_FLAG_TPM2) {
>> -		digest_list = kcalloc(chip->nr_allocated_banks,
>> -				      sizeof(*digest_list), GFP_KERNEL);
>> -		if (!digest_list)
>> -			return -ENOMEM;
>> -
>> -		for (i = 0; i < chip->nr_allocated_banks; i++) {
>> -			digest_list[i].alg_id = chip->allocated_banks[i].alg_id;
>> -			memcpy(digest_list[i].digest, hash, TPM_DIGEST_SIZE);
>> -		}
>> -
>> -		rc = tpm2_pcr_extend(chip, pcr_idx, chip->nr_allocated_banks,
>> -				     digest_list);
>> -		kfree(digest_list);
>> +		rc = tpm2_pcr_extend(chip, pcr_idx, count, digests);
>>   		tpm_put_ops(chip);
>>   		return rc;
>>   	}
>>   
>> -	rc = tpm1_pcr_extend(chip, pcr_idx, hash,
>> +	rc = tpm1_pcr_extend(chip, pcr_idx, count, digests,
>>   			     "attempting extend a PCR value");
> 
> The validation is missing that the provided array has only one element
> and the algorithm is SHA1. Could be done also inside tpm1_pcr_extend()
> but what you are doing to that function does not make any sense so
> better to do it here.
> 
>>   	tpm_put_ops(chip);
>>   	return rc;
>> diff --git a/drivers/char/tpm/tpm.h b/drivers/char/tpm/tpm.h
>> index 64d93d26087f..6b446504d2fe 100644
>> --- a/drivers/char/tpm/tpm.h
>> +++ b/drivers/char/tpm/tpm.h
>> @@ -504,7 +504,8 @@ int tpm1_auto_startup(struct tpm_chip *chip);
>>   int tpm1_do_selftest(struct tpm_chip *chip);
>>   int tpm1_get_timeouts(struct tpm_chip *chip);
>>   unsigned long tpm1_calc_ordinal_duration(struct tpm_chip *chip, u32 ordinal);
>> -int tpm1_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, const u8 *hash,
>> +int tpm1_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, u32 count,
>> +		    const struct tpm_extend_digest *digests,
>>   		    const char *log_msg);
>>   int tpm1_pcr_read(struct tpm_chip *chip, u32 pcr_idx, u8 *res_buf);
>>   ssize_t tpm1_getcap(struct tpm_chip *chip, u32 subcap_id, cap_t *cap,
>> @@ -551,7 +552,7 @@ int tpm2_get_timeouts(struct tpm_chip *chip);
>>   int tpm2_pcr_read(struct tpm_chip *chip, u32 pcr_idx,
>>   		  struct tpm_digest *digest, u16 *digest_size_ptr);
>>   int tpm2_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, u32 count,
>> -		    struct tpm_digest *digests);
>> +		    const struct tpm_extend_digest *digests);
>>   int tpm2_get_random(struct tpm_chip *chip, u8 *dest, size_t max);
>>   void tpm2_flush_context_cmd(struct tpm_chip *chip, u32 handle,
>>   			    unsigned int flags);
>> diff --git a/drivers/char/tpm/tpm1-cmd.c b/drivers/char/tpm/tpm1-cmd.c
>> index 8b70a7f884a7..04ee10284b8c 100644
>> --- a/drivers/char/tpm/tpm1-cmd.c
>> +++ b/drivers/char/tpm/tpm1-cmd.c
>> @@ -449,12 +449,20 @@ int tpm1_get_timeouts(struct tpm_chip *chip)
>>   }
>>   
>>   #define TPM_ORD_PCR_EXTEND 20
>> -int tpm1_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, const u8 *hash,
>> +int tpm1_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, u32 count,
>> +		    const struct tpm_extend_digest *digests,
>>   		    const char *log_msg)
>>   {
>>   	struct tpm_buf buf;
>> +	u8 dummy_hash[TPM_DIGEST_SIZE] = { 0 };
>> +	const u8 *hash;
>>   	int rc;
>>   
>> +	hash = dummy_hash;
>> +	if (count)
>> +		memcpy(dummy_hash, digests[0].data,
>> +		       min(digests[0].size, (u16)sizeof(dummy_hash)));
>> +
> 
> You copy memory from one place to another without any good reason to do
> so. My suggestion is just not to change tpm1_pcr_extend() at all.
> 
>>   	rc = tpm_buf_init(&buf, TPM_TAG_RQU_COMMAND, TPM_ORD_PCR_EXTEND);
>>   	if (rc)
>>   		return rc;
>> @@ -743,7 +751,6 @@ int tpm1_auto_startup(struct tpm_chip *chip)
>>    */
>>   int tpm1_pm_suspend(struct tpm_chip *chip, u32 tpm_suspend_pcr)
>>   {
>> -	u8 dummy_hash[TPM_DIGEST_SIZE] = { 0 };
>>   	struct tpm_buf buf;
>>   	unsigned int try;
>>   	int rc;
>> @@ -751,7 +758,7 @@ int tpm1_pm_suspend(struct tpm_chip *chip, u32 tpm_suspend_pcr)
>>   
>>   	/* for buggy tpm, flush pcrs with extend to selected dummy */
>>   	if (tpm_suspend_pcr)
>> -		rc = tpm1_pcr_extend(chip, tpm_suspend_pcr, dummy_hash,
>> +		rc = tpm1_pcr_extend(chip, tpm_suspend_pcr, 0, NULL,
>>   				     "extending dummy pcr before suspend");
>>   
>>   	rc = tpm_buf_init(&buf, TPM_TAG_RQU_COMMAND, TPM_ORD_SAVESTATE);
>> diff --git a/drivers/char/tpm/tpm2-cmd.c b/drivers/char/tpm/tpm2-cmd.c
>> index 6ce5173cf0e5..77b5808270c6 100644
>> --- a/drivers/char/tpm/tpm2-cmd.c
>> +++ b/drivers/char/tpm/tpm2-cmd.c
>> @@ -247,21 +247,22 @@ struct tpm2_null_auth_area {
>>    *
>>    * @chip:	TPM chip to use.
>>    * @pcr_idx:	index of the PCR.
>> - * @count:	number of digests passed.
>> - * @digests:	list of pcr banks and corresponding digest values to extend.
>> + * @count:	number of tpm_extend_digest passed.
>> + * @digests:	array of tpm_extend_digest with digest values to extend.
>>    *
>>    * Return: Same as with tpm_transmit_cmd.
>>    */
> 
> The documentation about @digests.
> 
>>   int tpm2_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, u32 count,
>> -		    struct tpm_digest *digests)
>> +		    const struct tpm_extend_digest *digests)
>>   {
>>   	struct tpm_buf buf;
>>   	struct tpm2_null_auth_area auth_area;
>> +	const struct tpm_extend_digest *digest;
>> +	u8 dummy_hash[SHA512_DIGEST_SIZE] = { 0 };
>> +	const u8 *hash;
>>   	int rc;
>>   	int i;
>> -
>> -	if (count > chip->nr_allocated_banks)
>> -		return -EINVAL;
>> +	int j;
>>   
>>   	rc = tpm_buf_init(&buf, TPM2_ST_SESSIONS, TPM2_CC_PCR_EXTEND);
>>   	if (rc)
>> @@ -277,11 +278,25 @@ int tpm2_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, u32 count,
>>   	tpm_buf_append_u32(&buf, sizeof(struct tpm2_null_auth_area));
>>   	tpm_buf_append(&buf, (const unsigned char *)&auth_area,
>>   		       sizeof(auth_area));
>> -	tpm_buf_append_u32(&buf, count);
>> +	tpm_buf_append_u32(&buf, chip->nr_allocated_banks);
>> +
>> +	if (count)
>> +		memcpy(dummy_hash, digests[0].data, digests[0].size);
>> +
>> +	for (i = 0; i < chip->nr_allocated_banks; i++) {
>> +		tpm_buf_append_u16(&buf, chip->allocated_banks[i].alg_id);
>> +
>> +		hash = dummy_hash;
>> +		for (j = 0; j < count; j++) {
>> +			digest = digests + j;
>> +
>> +			if (digest->alg_id == chip->allocated_banks[i].alg_id) {
>> +				hash = digest->data;
>> +				break;
> 
> I think the whole design is just wrong. I did re-read your response to
> v6 again and I'm very sorry, but I just don't get this. Caller has all
> the information (from struct tpm_chip) to give the correct data. This
> function should validate that data (check algorithm ID and that's it).

The question is if checking tpm->allocated_banks is a strict
requirement, or we can allow callers to use the algorithm they are
currently using, without further modifications.


> Extending with the dummy hash should be done by the caller when
> preparing the array, not baked into this function. This kind of also
> makes obvious that we don't need this new struct. There should never be
> a local variable (whose size is BTW randomly chosen) called dummy_hash.

This means duplicating the code for each caller. Currently, this work is
done by the TPM driver.

Roberto


> /Jarkko
> 

-- 
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063
Managing Director: Bo PENG, Jian LI, Yanli SHI

Powered by blists - more mailing lists