| 16 Jul 2021 |
andi- | It is nice that we have a well documented user of all of the TPM infrastructure. | 12:41:49 |
| hexa joined the room. | 12:41:58 |
andi- | I now wish that I could use the TPM for wireguard key derivation. | 12:41:58 |
@grahamc:nixos.org | is that openconnect? | 12:42:04 |
andi- | Yeah | 12:42:10 |
@grahamc:nixos.org | :) | 12:42:14 |
| spacesbot - keeps a log of public NixOS channels | 13:00:04 |
andi- | So yesterday I was able to wipe my state without th ecorrect password IIRC. All I did was call tpm2_clear. | 13:16:47 |
andi- | How do you protect against that? | 13:17:04 |
andi- | IIRC I did set two passwords when I first setup secrets. | 13:17:24 |
@grahamc:nixos.org | interesting | 13:21:19 |
@grahamc:nixos.org | not sure you can actually | 13:21:38 |
@grahamc:nixos.org | maybe you can | 13:21:44 |
@grahamc:nixos.org | but I'm thinking about how the bios can wipe it too | 13:24:08 |
andi- | That would mean that I must lock the tpm device away and only let root / a special user interact with it. | 13:24:25 |
andi- | I read some text that said that there are some hardware keys to adjust it | 13:24:38 |
@grahamc:nixos.org | you sort of need to do that anyway | 13:25:12 |
@grahamc:nixos.org | because the nvram isn't partitioned or anything, it has no fs, you just have offsets in to the memory you write to | 13:25:35 |
andi- | So you need to coordinate offsets across all your tools? e.g. OpenConnect and my kerberos daemon must each know where they can write? | 13:26:50 |