tpm2: Sign with hash ticket about go-tpm

[salrashid123]

+1
i ran into this as well.

I needed the AK to Sign something and would need the ticket provided by Hash to do that

it looks like even Hash takes the first handle bytes alone prematurely:
https://github.com/google/go-tpm/blob/master/tpm2/tpm2.go#L1385

when it should be able to get the full struct:
https://github.com/tpm2-software/tpm2-tools/blob/master/lib/tpm2.c#L966

• with tpm2_tools:

tpm2_createek -c ek.ctx -G rsa -u ek.pub
tpm2_createak -C ek.ctx -c ak.ctx -n ak.name -u ak.pub


echo "meet me at.." > message.txt

tpm2_hash -C e -g sha256 -o hash.bin -t ticket.bin message.txt

w/o ticket
tpm2_sign -c ak.ctx -g sha256 -o sig.rssa message.txt
  ERROR: Eys_Sign(0x3E0) - tpm:parameter(3):invalid ticket
  ERROR: Unable to run tpm2_sign

w/ ticket

tpm2_sign -c ak.ctx -g sha256 -o sig.rssa -t ticket.bin message.txt

tpm2_verifysignature -c ak.ctx -g sha256 -s sig.rssa -m message.txt

TPMT_TK_HASHCHECK – This ticket is used to indicate that a digest of external data is safe to sign
using a restricted signing key. A restricted signing key may only sign a digest that was produced by the
TPM. If the digest was produced from externally provided data, there needs to be an indication that the
data did not start with the same first octets as are used for data that is generated within the TPM. This
prevents “forgeries” of attestation data. This ticket is used to provide the evidence that the data used in
the digest was checked by the TPM and is safe to sign. Assuming that the external data is "safe", this
type of ticket is produced by TPM2_Hash() or TPM2_SequenceComplete() and used by TPM2_Sign().

[twitchy-jsonp]

It should be straightforward to allow you to sign while providing a hash ticket, as ATM tpm2.Sign just passes nothing for that parameter.

Would that change be enough, or would you also need changes to tpm2.Hash as well?

[salrashid123]

i think tpm2.Hash needs to return the rest of the part of data struct it gets from the tpm as the ticket
...that ticket needs to get stuffed into the Sign() function

take a look at the tpm2_tools command sequence: tpm2_hash outputs a ticket which must be provided to tpm2_sign (i.,e if you ask the ak to sign arbitrary stuff, its not going to do so unless it knows the tpm hashed it...and its not going to hash any arbitrary bit of data on behalf of an AK (i.,e magic number prefix)...thats my read atleast).

[chrisfenner]

I started PR #178, went to create the new issue and found it already existed. 😊

Rather than commit a breaking change to tpm2.Hash (to add a new return parameter for the ticket) as well as to tpm2.Sign (to accept a new validation parameter, again for the ticket), I added new overloads of these functions: HashVerified and SignVerifiedHash.

It should be noted (so I did, in the function documentation) that the maximum amount of data you can hash with the one-shot TPM2_Hash command is implementation-dependent, depending on MAX_DIGEST_BUFFER which is only guaranteed by the spec to be up to 1,024 bytes. Some TPMs may support larger buffers, leading to code which works on some TPMs (like the simulator) but not others (like a real discrete TPM with limited RAM). For more than 1024 bytes of data, the sequence commands should be used, but these aren't yet implemented by go-tpm.

I filed issue #179 to track possibly extending the API for this.