Re: [PATCH] LSM: Pass comm name via get_task_comm() [was: Re: [PATCH] Change task_struct->comm to use RCU.]

[Stephen Smalley <sds@tycho.nsa.gov>]

On 03/27/2014 01:20 PM, Richard Guy Briggs wrote:
> On 14/03/12, James Morris wrote:
>> On Tue, 11 Mar 2014, Tetsuo Handa wrote:
>>
>>> And the same phrase goes to James Morris...
>>>
>>> If you are sure that it is safe to use get_task_comm() from
>>> dump_common_audit_data() and you prefer locked version, please pick up below
>>> patch via your git tree.
>>>
>>> If you are unsure or prefer lockless version, I'll make a lockless version
>>> using do_get_task_comm() proposed in this thread.
>>
>> If you can't understand whether your patch is correct or not, don't ask me 
>> to apply it to my tree.
>>
>> If you're unsure, get it reviewed first.
> 
> Steve (see https://lkml.org/lkml/2014/3/11/218 ) and James,
> 
> Are the labels on data output in LSM_AUDIT_DATA_TASK even right?  The
> general case gives pid and comm of current.  Then the
> LSM_AUDIT_DATA_TASK case gives pid and comm from the task handed in in
> the struct common_audit_data pointer.  They are a duplicate of the
> general case without generating a new message.  I expect this will cause
> ausearch to ignore those latter two fields.  Should the latter two be
> renamed to something like ad_pid= and ad_comm= ?

Hmmm..only seems to be used by Smack.
SELinux had a tsk field in common_audit_data that was removed by
b466066.  This other tsk field seems to have been added for Smack by
6e837fb.

That said, it would be nice to have pid/comm info for the target of a
signal check as well as current.