In other words, DRM from top to bottom ...

This is a very common misunderstanding, and one that really hurts every discussion of TPM. Benefits such as hard-disk encryption can be achieved without problems even if all keys in the system are known to the user (at least in principle). Nobody is against that as far as I know. In this scenario, the TPM is nothing but a glorified smart card with some tamper-resistant storage and a crypto OS.

The point of contention has been the remote attestion feature, which relies on a secret key in the TPM chip that is not known (and must not be known) to the owner of the hardware, but only to the manufacturer. This remote attestion feature in principle allows third parties to verify the content of the system, and implementation of other supsicious features. It is not needed for local disk encryption.

This is why the GPLv3 allows TPM features in software, but only if all keys are provided to the user. This effectively disables all features based on remote attestion (DRM etc), but does allow local disk encryption etc.