Add certification to Seal #60
Conversation
|
Hi @josephlr PTAL |
josephlr
left a comment
There was a problem hiding this comment.
Good job on this, I have a bunch of lints, sorry.
tpm2tools/pcr.go
Outdated
| return &pl, nil | ||
| } | ||
|
|
||
| // SealCurrent represent current PCRs states. |
There was a problem hiding this comment.
Here and below, we need to rephrase these comments.
Idea:
// SealCurrent seals data to the specified PCR selection.
// SealTarget predicatively seals data to the specified PCR values.
// SealingOpt specifies the PCR values that should be used for Seal().
// CertifyCurrent certifies that a selection of PCRs have the same value they did when sealing the data.
// CertifyExpected certifies that the TPM had a specific set of PCR values when sealing the data.
// CertifyOpt determines which sets of PCR values pass certification in Unseal().There was a problem hiding this comment.
// SealCurrent seals data to the current specified PCR selection.
// SealTarget predicatively seals data to the given specified PCR values.
// SealingOpt specifies the PCR values that should be used for Seal().
// CertifyCurrent certifies that a selection of current PCRs have the same value when sealing.
// CertifyExpected certifies that the TPM had a specific set of PCR values when sealing.
// CertificationOpt determines if the given PCR value can pass certification in Unseal().
stevenrutherford
left a comment
There was a problem hiding this comment.
Looks like this was already reviewed pretty thoroughly. Comments are mostly nits.
|
This version will allow seal to empty PCRs selection.
|
|
Looks good. One last question: why disallow empty PCRSelections for CertifyOpt? Seems like the same story as Seal applies there: if you want to only assert which machine you are on and not that a particular machine is in a particular state, empty selections seem useful. |
We do support this (i.e. skipping certification and just unsealing) by providing a nil CertifyOpt. We then disallow empty selections as we don’t want to have/support two ways to do the same thing. I think for consistency, when sealing to an empty selection, we should use a nil SealingOpt and explicitly disallow empty selections or proto.Pcrs. That way, we can catch bugs if the programmer accidentally passes a nil selection. |
|
SealOpt now can be nil. Will panic if pass in empty selection or empty proto.Pcrs in SealOpt. I added a test for seal/reseal to nil SealOpt. |
stevenrutherford
left a comment
There was a problem hiding this comment.
Looks good, thanks
Changed Seal() Unseal() Reseal() APIs to accept SealingOpt and CertifyOpt.