Thanks Mimi for your comments.

On Fri, 14 Jun 2019 at 05:33, Mimi Zohar [off-list ref] wrote:

On Thu, 2019-06-13 at 09:40 -0700, Casey Schaufler wrote:

quoted

On 6/13/2019 3:30 AM, Sumit Garg wrote:

quoted

Add support for TEE based trusted keys where TEE provides the functionality
to seal and unseal trusted keys using hardware unique key. Also, this is
an alternative in case platform doesn't possess a TPM device.

This series also adds some TEE features like:

Please expand the acronym TEE on first use. That will
help people who don't work with it on a daily basis
understand what you're going on about.

Thanks, Casey.

"[6/7] doc: keys: Document usage of TEE based Trusted Keys" refers to
the kernel tee documentation, but that documentation is limited to
userspace interaction with the tee.

Thanks for pointing this out. I will update documentation to include
TEE bus approach and communication apis for kernel clients.

BTW, the interface is similar as with user-space. Only difference is
instead of IOCTL's from user-space, there are wrapper apis to
communicate with TEE.

Also, in case someone is interested to learn about TEE technology,
this webinar [1] could be one of starting points.

A trusted key is a random number generated and sealed(encrypted) by
the TPM, so that only the TPM may unseal it.  The sealing key never
leaves the TPM.  The sealed, trusted key may be exported to userspace.

Understood.

 In the tee case, can the "sealing" key ever leave the tee?

No, the "sealing" key never leaves TEE. Its basically a Hardware
Unique Key (HUK) tied to a particular SoC.

 Can the
sealed, trusted key, exported to userspace, be unsealed by the tee?

You mean using user-space interface to TEE? If yes, then answer is
"no" as user-space can't communicate with this TEE service as its
accessible to kernel clients only (see patch [2]).

In case you meant loading exported trusted key blob via "keyctl", then
"yes" this driver can unseal the trusted key. Have a look at examples
I have listed in documentation patch [3]. Also, this approach works
well across power cycles too.

 Are the tee security protections similar to those of the TPM?  How do
they compare?

Let me try to compare both environments. Regarding TEE, I will refer
to OP-TEE [4] as one of its implementation.

TPM:

1. External hardware.
2. Sealing key resides inside TPM.
3. Communicates via SPI, I2C etc.

OP-TEE:

1. On chip, trusted execution environment enforced via ARM TrustZone.
2. Sealing key is unique to a particular SoC provided by secure fuses,
secure crypto engine etc.
3. Communicates via Secure Monitor Calls (SMCs [5]).

[1] https://globalplatform.org/resource-publication/webinar-an-introduction-to-tee-technology/
[2] [RFC 3/7] tee: add private login method for kernel clients
[3] [RFC 6/7] doc: keys: Document usage of TEE based Trusted Keys
[4] https://optee.readthedocs.io/general/about.html
[5] http://infocenter.arm.com/help/topic/com.arm.doc.den0028b/ARM_DEN0028B_SMC_Calling_Convention.pdf


-Sumit

Mimi

quoted

quoted

Patch #1, #2 enables support for registered kernel shared memory with TEE.

Patch #3 enables support for private kernel login method required for
cases like trusted keys where we don't wan't user-space to directly access
TEE service to retrieve trusted key contents.

Rest of the patches from #4 to #7 adds support for TEE based trusted keys.

This patch-set has been tested with OP-TEE based pseudo TA which can be
found here [1].

Looking forward to your valuable feedback/suggestions.