Re: Bug confirmed...?

[Taylor R Campbell <riastradh%NetBSD.org@localhost>]

> From: Ellie <el%horse64.org@localhost>
> Date: Fri, 30 Oct 2026 23:30:00 +0100
> 
> Thanks to some super helpful people in the #netbsd chat I found a
> minimal broken config triggering the ioctl error now:
> 
> algorithm aes-cbc;
> iv-method encblkno1;
> keylength 512;
> [...]
> 
> This one works:
> 
> algorithm aes-xts;
> iv-method encblkno1;
> keylength 512;
> [...]

The aes-cbc configuration is, in fact, invalid.

For aes-cbc, the supported key lengths are 128, 192, and 256,
corresponding to the lengths of keys for AES-128, AES-192, and
AES-256, respectively.

But for aes-xts, the supported key lengths are 256, 384, and 512,
corresponding to the lengths of _pairs_ of keys for AES-128, AES-192,
and AES-256, respectively.  (This is because XTS is defined in terms
of two independent block cipher instances, even though it doesn't
improve security meaningfully over using one as proposed in Rogaway's
original XEX paper.)

This is noted in the cgd(4) man page.

If sysinst(8) is producing `keylength 512;' for aes-cbc, that would be
a bug in sysinst(8).