Re: [PATCH] [RFC] xsm: add a default policy to .init.data

[Daniel De Graaf <dgdegra@tycho.nsa.gov>]

On 05/23/2016 11:25 AM, Andrew Cooper wrote:
> On 23/05/16 15:51, Daniel De Graaf wrote:
>> diff --git a/xen/xsm/xsm_core.c b/xen/xsm/xsm_core.c
>> index 634ec98..af1d86f 100644
>> --- a/xen/xsm/xsm_core.c
>> +++ b/xen/xsm/xsm_core.c
>> @@ -47,6 +47,17 @@ static void __init do_xsm_initcalls(void)
>>       }
>>   }
>>
>> +extern char __xsm_init_policy_start[], __xsm_init_policy_end[];
>> +
>> +static void __init xsm_policy_init(void)
>> +{
>> +    if ( policy_size == 0 )
>> +    {
>> +        policy_buffer = __xsm_init_policy_start;
>> +        policy_size = __xsm_init_policy_end - __xsm_init_policy_start;
>> +    }
>
> Logic like this is slightly problematic if there is no policy.
>
> With these changes, we presumably always expect to have an embedded policy.

People who already have the policy specified in the bootloader may want
to omit the built-in policy.  I'm not sure that this should be excluded
completely, although this patch doesn't support it (it would require the
Kconfig option I mentioned).

> It would be cleaner to have a linker ASSERT(__xsm_init_policy_start !=
> __xsm_init_policy_end) to guarentee that something is present, at which
> point policy_buffer can unilaterally point at __xsm_init_policy_start,
> and size can be initialised to __xsm_init_policy_end -
> __xsm_init_policy_start.

No, because this would break the ability to specify a policy module in
GRUB.  If there is no built-in policy present, this code will work
correctly: policy_size will remain set to zero, and that is the
condition checked (in flask_init) when the policy itself is used.

-- 
Daniel De Graaf
National Security Agency

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel