Re: [PATCH] Smack: replace capable() with ns_capable()

[Casey Schaufler <casey@schaufler-ca.com>]

On 7/26/2015 6:27 PM, Sungbae Yoo wrote:
> So, Do you agree to allow the process to change its own labels?

No. This requires CAP_MAC_ADMIN. Smack is mandatory access control.
Being in a namespace (as they are implemented today) is not sufficient.

>
> Now, init process(eg. systemd) can't be running in user namespace properly
> because it can't be assign smack label to service.
>
> If you agree, I'll upload another patch limited to this.
>
>
> -----Original Message-----
> From: Lukasz Pawelczyk [mailto:l.pawelczyk@samsung.com] 
> Sent: Friday, July 24, 2015 8:41 PM
> To: Sungbae Yoo; Casey Schaufler
> Cc: James Morris; Serge E. Hallyn; linux-security-module@vger.kernel.org; linux-kernel@vger.kernel.org
> Subject: Re: [PATCH] Smack: replace capable() with ns_capable()
>
> On pią, 2015-07-24 at 20:26 +0900, Sungbae Yoo wrote:
>> If current task has capabilities, Smack operations (eg. Changing own 
>> smack
>> label) should be available even inside of namespace.
>>
>> Signed-off-by: Sungbae Yoo <sungbae.yoo@samsung.com>
>>
>> diff --git a/security/smack/smack_access.c 
>> b/security/smack/smack_access.c index 00f6b38..f6b2c35 100644
>> --- a/security/smack/smack_access.c
>> +++ b/security/smack/smack_access.c
>> @@ -639,7 +639,7 @@ int smack_privileged(int cap)
>>  	struct smack_known *skp = smk_of_current();
>>  	struct smack_onlycap *sop;
>>  
>> -	if (!capable(cap))
>> +	if (!ns_capable(current_user_ns(), cap))
>>  		return 0;
> It's not that easy.
>
> With this change Smack becomes completely insecure. You can change rules as an unprivileged user without any problems now.
> What you want is Smack namespace that was made to remedy exactly this issue (e.g. changing own labels inside a namespace).
>
>>  
>>  	rcu_read_lock();
>> diff --git a/security/smack/smack_lsm.c b/security/smack/smack_lsm.c 
>> index a143328..7fdc3dd 100644
>> --- a/security/smack/smack_lsm.c
>> +++ b/security/smack/smack_lsm.c
>> @@ -403,7 +403,8 @@ static int smk_ptrace_rule_check(struct 
>> task_struct *tracer,
>>  			rc = 0;
>>  		else if (smack_ptrace_rule ==
>> SMACK_PTRACE_DRACONIAN)
>>  			rc = -EACCES;
>> -		else if (capable(CAP_SYS_PTRACE))
>> +		else if (ns_capable(__task_cred(tracer)->user_ns,
>> +				    CAP_SYS_PTRACE))
>>  			rc = 0;
>>  		else
>>  			rc = -EACCES;
> --
> Lukasz Pawelczyk
> Samsung R&D Institute Poland
> Samsung Electronics
>
>
>
>