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

[David Safford]

On Wed, 2011-11-02 at 18:37 +0100, Roberto Sassu wrote:
> On 11/02/2011 05:58 PM, David Safford wrote:
> > 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@xxxxxxxxx>
> >
> > 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?
> >
> 
> Hi Dave
> 
> i think this line should be removed because it overwrites
> the assignment made in datablob_parse() -> getoptions().
> 
> You can see this behaviour by executing the following commands:
> 
> 1) keyctl add trusted kmk "new 32" @u
> 2) keyctl update $(keyctl search @u trusted kmk) "update migratable=0"
> 3) keyctl update $(keyctl search @u trusted kmk) "update migratable=0"
> 
> The third operation should fail because of the lines at the
> begin of trusted_update():
> 
> ---------
> 	if (!p->migratable)
> 		return -EPERM;
> ---------
> 
> but instead, without the patch, it is performed successfully.
> 
> Thanks
> 
> Roberto Sassu

Sorry - you are absolutely correct. The patch looks good,
and it is safe, as it allows only migratable -> non-migratable updates.
Give me a day to test both...
dave

> 
> > dave
> >
> >>   	new_p->key_len = p->key_len;
> >>   	memcpy(new_p->key, p->key, p->key_len);
> >>   	dump_payload(p);
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-security-module" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/