[Xense-devel] RE: vTPM data seal issue

["Scarlata, Vincent R" <vincent.r.scarlata@xxxxxxxxx>]

I've CCed Mario Strasser. He is the author of the TPM emulator that is used for the vTPMs.

 

Thanks Justin, this I believe is a bug in the TPM emulator. If pcrInfoSize is zero, then the TPM protocols say to completely ignore the PCR_INFO structure and the next byte in this case should be inDataSize. The marshalling code always unpacks the PCR_INFO regardless of is pcrInfoSize == 0.

 

Mario, I have to think about it for a second, but I think everywhere you unmarshal PCR_INFO it should look something like

 

   if (tpm_unmarshal_TPM_KEY_HANDLE(&ptr, &len, &keyHandle)
           || tpm_unmarshal_TPM_ENCAUTH(&ptr, &len, &encAuth)
           || tpm_unmarshal_UINT32(&ptr, &len, &pcrInfoSize)
           || (  pcrInfoSize > 0  ) && tpm_unmarshal_TPM_PCR_INFO(&ptr, &len, &pcrInfo) )
           || tpm_unmarshal_UINT32(&ptr, &len, &inDataSize)
           || tpm_unmarshal_BLOB(&ptr, &len, &inData, inDataSize)
           || len != 0) return TPM_BAD_PARAMETER;

 

-Vinnie Scarlata

 Trusted Platform Lab

 Corporate Technology Group

 Intel Corporation

---

From: Osborn, Justin D. [mailto:Justin.Osborn@xxxxxxxxxx]
Sent: Wednesday, October 18, 2006 3:16 PM
To: xense-devel@xxxxxxxxxxxxxxxxxxx; trousers-users@xxxxxxxxxxxxxxxxxxxxx; Scarlata, Vincent R
Subject: vTPM data seal issue

Hi everybody,

 

    For a while now I've been working on a project to do data sealing to TPM PCRs in a Xen Dom U using the vtpm.  I finally got everything set up, I'm using xen-unstable, the tpm_tis driver in Dom 0, and trousers 0.2.8 in Dom U (having patched Trousers with the TPM_PCR_INFO_LONG patch from Trusted Java that allows data sealing on Winbond/Infineon TPMs).

 

I booted my Dom U image, modprobe tpm_xenu, ran tcsd, and tpm_takeownership.  For the test I ran:

 

    ps auwx > foo

    tpm_sealdata < foo

 

This returns a TPM_BAD_PARAMETER.  However, executing:

 

    tpm_sealdata -p 1 < foo

 

works completely.

 

Further investigation in the vtpmd code revealed a block in tpm_cmd_handler.c that looks like:

 

     if (tpm_unmarshal_TPM_KEY_HANDLE(&ptr, &len, &keyHandle)
           || tpm_unmarshal_TPM_ENCAUTH(&ptr, &len, &encAuth)
           || tpm_unmarshal_UINT32(&ptr, &len, &pcrInfoSize)
           || tpm_unmarshal_TPM_PCR_INFO(&ptr, &len, &pcrInfo)
           || tpm_unmarshal_UINT32(&ptr, &len, &inDataSize)
           || tpm_unmarshal_BLOB(&ptr, &len, &inData, inDataSize)
           || len != 0) return TPM_BAD_PARAMETER;

 

vtpmd tries to unpackage everything regardless of pcrInfoSize.  So essentially what happens is, if not sealing to PCRs, trousers sends pcrInfoSize as 0, and then doesn't include the pcrInfo.  However, the vtpm is looking for the pcrInfo no matter what, and it tries to read past the end of the data.  It would appear that trousers is not breaking spec to do this because without vtpm data sealing works.  Perhaps the vtpm should unmarshal everything up to the pcrInfo and then check pcrInfoSize before trying to unmarshal pcrInfo?

 

For now I'm just always sealing to PCRs. :-)

 

Let me know if that makes sense.

 

Thanks,

Justin

 

P.S.  I'm completely new to TCG/TPM, so please excuse any naivete. 

_______________________________________________
Xense-devel mailing list
Xense-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xense-devel