Re: [PATCH v6 1/7] tpm: dynamically allocate the allocated_banks array

From: Roberto Sassu <roberto.sassu@huawei.com>
To: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Cc: <zohar@linux.ibm.com>, <david.safford@ge.com>,
	<monty.wiseman@ge.com>, <linux-integrity@vger.kernel.org>,
	<linux-security-module@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>, <silviu.vlasceanu@huawei.com>,
	Nayna Jain <nayna@linux.ibm.com>
Subject: Re: [PATCH v6 1/7] tpm: dynamically allocate the allocated_banks array
Date: Thu, 6 Dec 2018 18:56:33 +0100	[thread overview]
Message-ID: <9193bc05-8222-f0d6-9ad8-a2a7eaf1a34c@huawei.com> (raw)
In-Reply-To: <20181204231848.GA1233@linux.intel.com>

On 12/5/2018 12:18 AM, Jarkko Sakkinen wrote:
> On Tue, Dec 04, 2018 at 09:21:32AM +0100, Roberto Sassu wrote:
>> tpm2_get_pcr_allocation() determines if a PCR bank is allocated by checking
>> the mask in the TPML_PCR_SELECTION structure returned by the TPM for
>> TPM2_Get_Capability(). One PCR bank with algorithm set to SHA1 is always
>> allocated for TPM 1.x.
> 
> ...
> 
>> +		for (j = 0; j < pcr_selection.size_of_select; j++)
>> +			if (pcr_selection.pcr_select[j])
>> +				break;
>> +
>> +		if (j < pcr_selection.size_of_select) {
>> +			chip->allocated_banks[nr_alloc_banks] = hash_alg;
>> +			nr_alloc_banks++;
>> +		}
>> +
> 
> Why was this needed? Can CAP_PCRS return completely unallocated banks?

This was discussed for patch v4 1/6:
---

Nayna wrote:

# /usr/local/bin/tssgetcapability -cap 5
2 PCR selections
     hash TPM_ALG_SHA1
     TPMS_PCR_SELECTION length 3
     ff ff ff
     hash TPM_ALG_SHA256
     TPMS_PCR_SELECTION length 3
     00 00 00

The pcr_select fields - "ff ff ff" and "00 00 00" - are bit masks for
the enabled PCRs. The SHA1 bank is enabled for all PCRs (0-23), while
the SHA256 bank is not enabled.
---

> Kind of out-of-context for the rest of the changes.
> 
> Should this be a bug fix of its own because it looks like as this is a
> bug fix for existing code, and not a new feature? Just asking because
> I don't yet fully understand this change.

If we store in tpm_chip the possible banks, IMA would calculate more
digests unnecessarily. But this problem does not happen without my patch
set, because tpm_pcr_extend() only accepts a SHA1 digest.

> Anyway, I believe that you can streamline this by:
> 
> /* Check that at least some of the PCRs have been allocated. This is
>   * required because CAP_PCRS ...
>   */
> if (memchr_inv(pcr_selection.pcr_select, 0, pcr_selection.size_of_select))
> 	nr_allocated_banks++;
> 
> [yeah, comment would be awesome about CAP_PCRS. Did not finish up the
> comment because I don't know the answer]
> 
> In addition, it would be consistent to call the local variable also
> nr_allocated_banks (not nr_alloc_banks).

Unfortunately, I exceed the limit of characters per line.

Roberto

> /Jarkko
> 

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