[FlorianFranzen]

The Yubico IMPORT ASYMMETRIC KEY extension only accepted RSA keys, and only into slot 9A. This adds a P-256/P-384 arm that reads the private scalar from DO 0x06 (zero-padded to the curve's scalar size, as ykman and yubikey.rs send it), injects it through the existing unsafe_inject_key path, and accepts every asymmetric key reference including the retired slots. The RSA path is left as it is. The PIN and touch policy DOs those clients append are ignored, as GENERATE ASYMMETRIC ignores them today.

The zero scalar is rejected in the app because the software fallback behind the virt platform accepts it while real backends reject it, which keeps behaviour identical across backends; an out-of-range scalar is rejected by the backend.

Tests import into 9A, 9C, 9E and retired slots, exercise the per-slot PIN conditions, verify the resulting ECDSA signatures against verifying keys recomputed from the imported scalars (ECDSA is randomized, so signature bytes cannot be compared directly), and reject truncated, wrong-curve, all-zero and out-of-range scalars.

Refs #16

[mirko]

Hijacking this thread: could anybody tell my why there is an artificial restriction into only allowing importing keys into slot 9D and not e.g. into the retired ones? I couldn't find anything reasoning. Is this maybe, because retired slots got introduced only later into the firmware and this restriction is a remnant?