Re: [PATCH v6] KEYS: Protect pkey method checks against type removal

[Yuansheng Yang <0xbeac@gmail.com>]

On Tue, Nov 10, 2026 at 03:49:28AM +0200, Jarkko Sakkinen wrote:
> > -	if (!params->key->type->asym_query)
> > -		return -EOPNOTSUPP;
> > -
>
> I think I asked to remove this. I need to revoke my reviewed-by from
> this version.
>
> It's not relevant for the bug fix.

You did, in the v4 review, and v6 dropped the check again by mistake.
v7 restores it: keyctl_pkey_params_get() is identical to 7.4-rc1 again,
and the diff does not touch it at all.

The two early checks in keyctl_pkey_e_d_s() and keyctl_pkey_verify()
stay where v6 moved them, in the key->sem section of
keyctl_pkey_params_get_2() and after the buffer size validation, so the
error precedence is unchanged.  Only the check you named is restored.

Sashiko's remaining point on v5 was that this check reads key->type
without key->sem.  That is how 7.4-rc1 has it and you asked to keep the
hunk out of the patch, so v7 leaves it alone; the call sites recheck
the method under key->sem before they use it.

> Also if possible try to make a bit more compact description of the
> fix. I did not ask this for reviewed patch because did not see it
> relevant enough but now that this needs a new round, you could
> address this part too.

Done: the race diagram and the paragraph describing the early-check
move are gone; the problem, the action, why it is safe and the coverage
list stay.

v7 follows right after this mail.

Thanks,
Yuansheng Yang