Re: [PATCH v7 1/5] tpm: dynamically allocate the allocated_banks array

[Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>]

On Mon, Jan 07, 2019 at 11:06:33AM +0100, Roberto Sassu wrote:
> On 12/22/2018 1:03 AM, Jarkko Sakkinen wrote:
> > On Fri, Dec 21, 2018 at 10:40:09AM +0100, Roberto Sassu wrote:
> > > On 12/20/2018 3:55 PM, Jarkko Sakkinen wrote:
> > > > On Thu, Dec 13, 2018 at 11:29:41AM +0100, Roberto Sassu wrote:
> > > > > This patch renames active_banks (member of tpm_chip) to allocated_banks,
> > > > > stores the number of allocated PCR banks in nr_allocated_banks (new member
> > > > > of tpm_chip), and replaces the static array with a pointer to a dynamically
> > > > > allocated array.
> > > > > 
> > > > > 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(). If a bank is not allocated, the TPM returns that
> > > > > bank in TPML_PCR_SELECTION, with all bits in the mask set to zero. In this
> > > > > case, the bank is not included in chip->allocated_banks, to avoid that TPM
> > > > > driver users unnecessarily calculate a digest for that bank.
> > > > > 
> > > > > One PCR bank with algorithm set to SHA1 is always allocated for TPM 1.x.
> > > > > 
> > > > > As a consequence of the introduction of nr_allocated_banks,
> > > > > tpm_pcr_extend() does not check anymore if the algorithm stored in tpm_chip
> > > > > is equal to zero.
> > > > > 
> > > > > Signed-off-by: Roberto Sassu <roberto.sassu@huawei.com>
> > > > > Tested-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
> > > > > ---
> > > > >    drivers/char/tpm/tpm-chip.c      |  1 +
> > > > >    drivers/char/tpm/tpm-interface.c | 18 +++++++++--------
> > > > >    drivers/char/tpm/tpm.h           |  3 ++-
> > > > >    drivers/char/tpm/tpm1-cmd.c      | 10 ++++++++++
> > > > >    drivers/char/tpm/tpm2-cmd.c      | 34 ++++++++++++++++++++++----------
> > > > >    5 files changed, 47 insertions(+), 19 deletions(-)
> > > > > 
> > > > > diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
> > > > > index 32db84683c40..ce851c62bb68 100644
> > > > > --- a/drivers/char/tpm/tpm-chip.c
> > > > > +++ b/drivers/char/tpm/tpm-chip.c
> > > > > @@ -160,6 +160,7 @@ static void tpm_dev_release(struct device *dev)
> > > > >    	kfree(chip->log.bios_event_log);
> > > > >    	kfree(chip->work_space.context_buf);
> > > > >    	kfree(chip->work_space.session_buf);
> > > > > +	kfree(chip->allocated_banks);
> > > > >    	kfree(chip);
> > > > >    }
> > > > > diff --git a/drivers/char/tpm/tpm-interface.c b/drivers/char/tpm/tpm-interface.c
> > > > > index d9439f9abe78..7b80919228be 100644
> > > > > --- a/drivers/char/tpm/tpm-interface.c
> > > > > +++ b/drivers/char/tpm/tpm-interface.c
> > > > > @@ -488,8 +488,7 @@ EXPORT_SYMBOL_GPL(tpm_pcr_read);
> > > > >    int tpm_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, const u8 *hash)
> > > > >    {
> > > > >    	int rc;
> > > > > -	struct tpm2_digest digest_list[ARRAY_SIZE(chip->active_banks)];
> > > > > -	u32 count = 0;
> > > > > +	struct tpm2_digest *digest_list;
> > > > >    	int i;
> > > > >    	chip = tpm_find_get_ops(chip);
> > > > > @@ -497,16 +496,19 @@ int tpm_pcr_extend(struct tpm_chip *chip, u32 pcr_idx, const u8 *hash)
> > > > >    		return -ENODEV;
> > > > >    	if (chip->flags & TPM_CHIP_FLAG_TPM2) {
> > > > > -		memset(digest_list, 0, sizeof(digest_list));
> > > > > +		digest_list = kcalloc(chip->nr_allocated_banks,
> > > > > +				      sizeof(*digest_list), GFP_KERNEL);
> > > > > +		if (!digest_list)
> > > > > +			return -ENOMEM;
> > > > 
> > > > You could preallocate digest list and place it to struct tpm_chip
> > > > instead of doing it everytime tpm_pcr_extend() called.
> > > 
> > > This part will be removed with patch 5/5.
> > 
> > Even if it did, it does not make this patch unbroken.
> 
> Can two calls to tpm_pcr_extend() be executed at the same time?
> 
> If yes, the digest list should be protected by a mutex.

Good question: the answer is no. Mutex locking is done inside the
transmit flow ATM.

/Jarkko