Re: [PATCH 1/2] trusted-key: allow overwriting the migratable flag

[David Safford <safford@watson.ibm.com>]

On Wed, 2011-11-02 at 13:41 +0100, Roberto Sassu wrote:
> The migratable should be modifiable during the key update() method. This
> allows for example to update a migratable trusted key, wrapped by a TPM
> key, to a a non-migratable one sealed under the SRK with a PCR set.
> 
> Signed-off-by: Roberto Sassu <roberto.sassu@polito.it>

I can see a use case for updating a migratable key to a non-migratable
one - such as keeping a migratable master on a flash drive, and keeping
only the non-migratable copy on-line. I certainly don't want the 
ability to change a non-migratable to migratable, as that would defeat
the entire purpose of non-migratable.

I don't think this patch actually does either, though. 

> ---
>  security/keys/trusted.c |    1 -
>  1 files changed, 0 insertions(+), 1 deletions(-)
> 
> diff --git a/security/keys/trusted.c b/security/keys/trusted.c
> index 0c33e2e..8777015 100644
> --- a/security/keys/trusted.c
> +++ b/security/keys/trusted.c
> @@ -1036,7 +1036,6 @@ static int trusted_update(struct key *key, const void *data, size_t datalen)
>  		goto out;
>  	}
>  	/* copy old key values, and reseal with new pcrs */
> -	new_p->migratable = p->migratable;

Taking out this line appears only to remove a redundant assignment.
We can only get here if the old key is already migratable, and the
earlier trusted_payload_alloc() initializes the new copy to
migratable by default. I don't see how the flag can be changed
with this patch. Perhaps I'm missing something or this was just 
the start, and there is more to come?

dave

>  	new_p->key_len = p->key_len;
>  	memcpy(new_p->key, p->key, p->key_len);
>  	dump_payload(p);