Secure boot for BMC

[Andrew Jeffery]

On Thu, 14 Feb 2019, at 11:04, Joseph Reynolds wrote:
> On 2019-02-12 17:13, Andrew Jeffery wrote:
> > On Tue, 12 Feb 2019, at 11:00, Nancy Yuen wrote:
> >> We are working on secure boot, but we have a requirement for a Google 
> >> HW
> >> root of trust so I'm not sure if that fits in with these discussions.
> > 
> > I think it would help to have some idea of Google's requirements so the 
> > project
> > can accommodate them where we can, if you can reveal any details. It 
> > may also
> > help inform others (me?) on strategies to secure firmware.
> 
> The OpenBMC security working group has discussed various "root of trust" 
> ideas.  The way I understand it, OpenBMC community members are looking 
> into different solutions including
> "Secure Boot" and "Trusted Platform Module" (TPM) solutions, including 
> Google's OpenTitan chip.  See the meeting minutes for details:
> https://docs.google.com/document/d/1b7x9BaxsfcukQDqbvZsU2ehMq4xoJRQvLxxsDUWmAOI
> 
> My understanding of the "Secure Boot" concept is that some chip 
> validates the boot loader's digital signature after loading it and 
> before jumping into it.  Then the boot loader would validate the code it 
> loads before jumping into it.  Etc.  A validation failure could either 
> (a) cause the BMC to fail to boot, or (b) boot the BMC in failsafe mode 
> where it could not write to its flash or talk to its host.  OpenBMC may 
> also need some way to talk to the chip.
> 
> My understanding of TPMs is much more limited.  So we are waiting for 
> proposals.

On OpenPOWER systems I think we need the TPM approach, as we can't
restrict our customers by burning e.g. IBM keys into the ASPEED OTP key
slots (... in the 2600).

Andrew