PSA: Recent AMD fTPM fixes introduced regressions in stable and LTS kernels that are breaking TPM, presumably for everyone

This shit again... I moved to LTS kernel exactly because of this bug and now it reached LTS kernel as well...

fixes introduced regressions

Software development in a nut shell

Good. TPM firmware is a high-value attack surface that, if penetrated, gives the attacker complete and irreversible control over the entire machine. It's also proprietary code and only receives security updates for a small fraction of its service life, so it's most likely riddled with vulnerabilities. It should be disabled, as much as possible, if you at all value the security of your system.

Huh, this might be one of the few examples of "don't break userspace" not being held to by Linus and co? I'm sure stuff like this has happened before, but "don't break userspace" has been a fairly strong guiding principle for the kernel for sometime. So seeing something like this happen is actually a bit surprising.

Though I guess it could be argued that if the removal of fTPM causes fewer bugs/issues than leaving it in place then userspace wasn't broken. But still, it's interesting to see regardless.

I've had fTPM issues with the bios because I changed CPUs and lost key attestation. Maybe it's my fault, but the reset process isn't documented anywhere clearly and it's not worth digging up my old CPU to reset the Mobo.

What's a good usecase for TPM in Linux?

Is it still broken even if you have TPM completely shut off in UEFI?