[PATCH v2 3/3] security: Add LSMs dependencies to CONFIG_LSM

[Mickaël Salaün]

On 21/02/2021 09:50, Masahiro Yamada wrote:
> On Tue, Feb 16, 2021 at 4:03 AM Ondrej Mosnacek <omosnace at redhat.com> wrote:
>>
>> On Mon, Feb 15, 2021 at 7:17 PM Mickaël Salaün <mic at digikod.net> wrote:
>>> From: Mickaël Salaün <mic at linux.microsoft.com>
>>>
>>> Thanks to the previous commit, this gives the opportunity to users, when
>>> running make oldconfig, to update the list of enabled LSMs at boot time
>>> if an LSM has just been enabled or disabled in the build.  Moreover,
>>> this list only makes sense if at least one LSM is enabled.
>>>
>>> Cc: Casey Schaufler <casey at schaufler-ca.com>
>>> Cc: James Morris <jmorris at namei.org>
>>> Cc: Masahiro Yamada <masahiroy at kernel.org>
>>> Cc: Serge E. Hallyn <serge at hallyn.com>
>>> Signed-off-by: Mickaël Salaün <mic at linux.microsoft.com>
>>> Link: https://lore.kernel.org/r/20210215181511.2840674-4-mic@digikod.net
>>> ---
>>>
>>> Changes since v1:
>>> * Add CONFIG_SECURITY as a dependency of CONFIG_LSM.  This prevent an
>>>   error when building without any LSMs.
>>> ---
>>>  security/Kconfig | 4 ++++
>>>  1 file changed, 4 insertions(+)
>>>
>>> diff --git a/security/Kconfig b/security/Kconfig
>>> index 7561f6f99f1d..addcc1c04701 100644
>>> --- a/security/Kconfig
>>> +++ b/security/Kconfig
>>> @@ -277,6 +277,10 @@ endchoice
>>>
>>>  config LSM
>>>         string "Ordered list of enabled LSMs"
>>> +       depends on SECURITY || SECURITY_LOCKDOWN_LSM || SECURITY_YAMA || \
>>> +               SECURITY_LOADPIN || SECURITY_SAFESETID || INTEGRITY || \
>>> +               SECURITY_SELINUX || SECURITY_SMACK || SECURITY_TOMOYO || \
>>> +               SECURITY_APPARMOR || BPF_LSM
>>
>> This looks really awkward, since all of these already depend on
>> SECURITY (if not, it's a bug)... I guarantee you that after some time
>> someone will come, see that the weird boolean expression is equivalent
>> to just SECURITY, and simplify it.
> 
> 
> Currently, LSM does not depend on SECURITY.
> So you can always define LSM irrespective of SECURITY,
> which seems a bug.
> 
> So, I agree with adding 'depends on SECURITY'.
> 
> What he is trying to achieve in this series
> seems wrong, of course.

This may be wrong in the general case, but not for CONFIG_LSM.

> 
> 
>> I assume the new mechanism wouldn't work as intended if there is just
>> SECURITY? If not, then maybe you should rather specify this value
>> dependency via some new  field rather than abusing "depends on" (say,
>> "value depends on"?). The fact that a seemingly innocent change to the
>> config definition breaks your mechanism suggests that the design is
>> flawed.

Masahiro, what do you think about this suggested "value depends on"?


>>
>> I do think this would be a useful feature, but IMHO shouldn't be
>> implemented like this.
>>
>>>         default "lockdown,yama,loadpin,safesetid,integrity,smack,selinux,tomoyo,apparmor,bpf" if DEFAULT_SECURITY_SMACK
>>>         default "lockdown,yama,loadpin,safesetid,integrity,apparmor,selinux,smack,tomoyo,bpf" if DEFAULT_SECURITY_APPARMOR
>>>         default "lockdown,yama,loadpin,safesetid,integrity,tomoyo,bpf" if DEFAULT_SECURITY_TOMOYO
>>> --
>>> 2.30.0
>>>
>>
>> --
>> Ondrej Mosnacek
>> Software Engineer, Linux Security - SELinux kernel
>> Red Hat, Inc.
>>
> 
>