On 19/11/2026 01:37, Jarkko Sakkinen wrote:
[snip]
The exitential problem here is that KEYCTL_CRYPT_OPEN creates a circular relationship between the keyrings API and a specific key type.
Agreed: it should be a key_type operation, dispatched the way KEYCTL_PKEY_* dispatches to asym_eds_op, so that keyctl knows no key type.
At hardware level I did not see any support for TPM, which narrows down the potential user base.
On a TPM system the sw backend can take its master from a TPM-sealed trusted key, but there keycrypt adds little over a trusted key. The users are SoCs with a TEE and often no TPM, OP-TEE on i.MX for one.
The remaining potential user base is further reduced by those that use trusted keys; leaving out those who need to solve the assumed problem that only this patch set can presumebly sort out.
And we don't really know at this point that cut. This is pretty much the break down of what is wrong here.
If there is/was a problem to sort out, it should be so constrained that it should be as quick to describe as stepping over a nail really.
The nail: a device must encrypt and decrypt its own files, also from a minimal initramfs, with a key that root or a kernel exploit can use on the device but never copy off it. Trusted keys hand their users the key's bytes (dm-crypt, encrypted keys and AF_ALG read them). CAAM's "pk" option is the one exception, but even there a new key is drawn in kernel memory before the CAAM wraps it, and user space reaches it only through AF_ALG, which the kernel calls insecure and deprecated (Documentation/crypto/userspace-if.rst, and "Userspace interface (deprecated)" in crypto/Kconfig), privileged by default, with no purpose bound to the key. The i.MX93 has no CAAM at all: there the key can live in OP-TEE, and keycrypt lets the keyring use it without it ever leaving the TA.
The stream of patch 9 has no user yet and can go.
Ciao,
Rodolfo