Thanks @vunnyso for the review. Especially item 1) is a nice catch. I'll push some changes to address your comments.

On every boot system is asking for PIN which can be prevented.

This is on purpose to prevent the partition to be decrypted automatically, in the event of a stolen device for example. Do you mean that the PIN should be remembered for a small period of time after the first input?
I agree that this is a tradeoff with usability to ask the user to enter a password twice (although the two passwords can be identical), but it gives the best assurance security-wise.

Thanks @vunnyso for the review. Especially item 1) is a nice catch. I'll push some changes to address your comments.

On every boot system is asking for PIN which can be prevented.

This is on purpose to prevent the partition to be decrypted automatically, in the event of a stolen device for example. Do you mean that the PIN should be remembered for a small period of time after the first input? I agree that this is a tradeoff with usability to ask the user to enter a password twice (although the two passwords can be identical), but it gives the best assurance security-wise.

Thanks @hros-tii. I mean entering a PIN at every boot is not user-friendly. It would be beneficial if we could make PIN as persistent by trying different TPM configuration registers or another security mechanism. For an additional layer of security, we can enroll mandatory Yubikey presence if needed.

Hi all,
Thank you for all the comments, I tried to include your feedback in the recently pushed commits. I just want to warn that I haven't had time to thoroughly test all the changes - in particular the Yubikey/FIDO related functionality - and plan to do so during the week. Still, comments on the new code are welcome.

On configuring the TPM without a PIN: an attacker having access to a stolen device that unlocks automatically can use other physical channels to access the disk contents (memory dumps for example). Binding to different PCRs cannot prevent these side-channel attacks as they don't modify any code running on the device. If this setup has to be used, I believe the user should be warned that their data could still be recovered in case their laptop is stolen. Even a weak PIN with a few digits can help mitigate this when coupled with rate limiting from the TPM to prevent brute-force attacks.
With that said I have added the option to encrypt the disk using a Yubikey instead of the TPM, and in that case no input is required from the user.

The new code is ready for review. A couple things were added:

• build time option to enroll a Yubikey instead of the hardware TPM
• encrypted swap partition
• persistence partition is resized on first boot

The setup steps are the same as in the PR description, but in order to verify the new features some more things can be checked. On ghaf-host:

1. Verify that swap is enabled: free -h
2. Check that persist and swap are encrypted (dmsetup status or mount).
3. persist should be the size of the whole disk
4. The partitions should decrypt successfully on reboot with either TPM or Yubikey.

There is an open upstream bug when using FIDO2 devices to unlock the disk, specifically the models without a PIN input: systemd-cryptsetup + FIDO2 + --fido2-with-client-pin=false: fails to prompt for unlock at boot.

There's a race between the kernel detecting the plugged in USB devices and when the cryptsetup service tries to decrypt the partition. In some cases the latter runs first and finds the security token missing, falling back to the passphrase prompt. I could replicate it on the Lenovo laptop, here is how it looks in the logs:

Jun 25 14:53:19 ghaf-host systemd-cryptsetup[277]: Set cipher aes, mode xts-plain64, key size 512 bits for device /dev/disk/by-partuuid/20936304-3d57-49c2-8762-bbba07edbe75.
Jun 25 14:53:19 ghaf-host systemd-cryptsetup[277]: Security token not present for unlocking volume persist, please plug it in.
[...]
Jun 25 14:53:20 ghaf-host kernel: usb 3-7: New USB device found, idVendor=1050, idProduct=0402, bcdDevice= 5.71
Jun 25 14:53:20 ghaf-host kernel: usb 3-7: New USB device strings: Mfr=1, Product=2, SerialNumber=0
Jun 25 14:53:20 ghaf-host kernel: usb 3-7: Product: YubiKey FIDO
Jun 25 14:53:20 ghaf-host kernel: usb 3-7: Manufacturer: Yubico
Jun 25 14:53:20 ghaf-host kernel: hid-generic 0003:1050:0402.0001: hiddev96,hidraw0: USB HID v1.10 Device [Yubico YubiKey FIDO] on usb-0000:00:14.0-7/input0

Tested on Lenovo-X1 Gen11, filesystem snapshot remains intact. Basic testing looks fine.

Branch	File system snapshot
mainline
This PR