Re: [PATCH v2 1/3] keys, trusted: select the hash algorithm

[Mimi Zohar <zohar@linux.vnet.ibm.com>]

On Tue, 2015-11-03 at 09:39 +0200, Jarkko Sakkinen wrote:
> On Mon, Nov 02, 2015 at 07:16:49AM -0500, Mimi Zohar wrote:
> > On Fri, 2015-10-30 at 13:35 +0200, Jarkko Sakkinen wrote:
> > 
> > > @@ -787,6 +791,20 @@ static int getoptions(char *c, struct trusted_key_payload *pay,
> > >  				return -EINVAL;
> > >  			opt->pcrlock = lock;
> > >  			break;
> > > +		case Opt_hash:
> > > +			for (i = 0; i < HASH_ALGO__LAST; i++) {
> > > +				if (!strcmp(args[0].from, hash_algo_name[i])) {
> > > +					opt->hash = i;
> > > +					break;
> > > +				}
> > > +			}
> > > +			res = tpm_is_tpm2(TPM_ANY_NUM);
> > 
> > While looking at this, I wanted to verify that chips are still added to
> > the tail of the tpm_chip_list.  Unfortunately, commit "afb5abc tpm:
> > two-phase chip management functions" reverted David Howell's commit
> > "770ab65 TPM: Add new TPMs to the tail of the list to prevent
> > inadvertent change of dev".
> > 
> > > +			if (res < 0)
> > > +				return res;
> > > +			if (i == HASH_ALGO__LAST ||
> > > +			    (!res && i != HASH_ALGO_SHA1))
> > > +				return -EINVAL;
> > > +			break;
> > 
> > If the first TPM registered is a TPM 1.2, then changing the default TPM
> > 2.0 hash algorithm will fail.
> 
> Now that we are going fix this issue in 4.3 and 4.4 do you find this
> patch otherwise acceptable?
> 
> PS. In other options that we don't support in TPM2 I'm planning to
> submit a fix that they will return -EINVAL (like pcrinfo).

I don't have a problem failing the request, but I do suggest adding some
sort of error message.  Different systems might behavior differently
without any explanation.

Mimi