Re: [PATCH bpf-next v5 4/7] bpf: lsm: Implement attach, detach and execution

From: Casey Schaufler
Date: Tue Mar 24 2020 - 12:25:18 EST


On 3/24/2020 7:58 AM, Stephen Smalley wrote:
> On Tue, Mar 24, 2020 at 10:50 AM KP Singh <kpsingh@xxxxxxxxxxxx> wrote:
>> On 24-MÃr 10:35, Stephen Smalley wrote:
>>> On Mon, Mar 23, 2020 at 12:46 PM KP Singh <kpsingh@xxxxxxxxxxxx> wrote:
>>>> From: KP Singh <kpsingh@xxxxxxxxxx>
>>>> diff --git a/kernel/bpf/bpf_lsm.c b/kernel/bpf/bpf_lsm.c
>>>> index 530d137f7a84..2a8131b640b8 100644
>>>> --- a/kernel/bpf/bpf_lsm.c
>>>> +++ b/kernel/bpf/bpf_lsm.c
>>>> @@ -9,6 +9,9 @@
>>>> #include <linux/btf.h>
>>>> #include <linux/lsm_hooks.h>
>>>> #include <linux/bpf_lsm.h>
>>>> +#include <linux/jump_label.h>
>>>> +#include <linux/kallsyms.h>
>>>> +#include <linux/bpf_verifier.h>
>>>>
>>>> /* For every LSM hook that allows attachment of BPF programs, declare a NOP
>>>> * function where a BPF program can be attached as an fexit trampoline.
>>>> @@ -27,6 +30,32 @@ noinline __weak void bpf_lsm_##NAME(__VA_ARGS__) {}
>>>> #include <linux/lsm_hook_names.h>
>>>> #undef LSM_HOOK
>>>>
>>>> +#define BPF_LSM_SYM_PREFX "bpf_lsm_"
>>>> +
>>>> +int bpf_lsm_verify_prog(struct bpf_verifier_log *vlog,
>>>> + const struct bpf_prog *prog)
>>>> +{
>>>> + /* Only CAP_MAC_ADMIN users are allowed to make changes to LSM hooks
>>>> + */
>>>> + if (!capable(CAP_MAC_ADMIN))
>>>> + return -EPERM;
>>> I had asked before, and will ask again: please provide an explicit LSM
>>> hook for mediating whether one can make changes to the LSM hooks.
>>> Neither CAP_MAC_ADMIN nor CAP_SYS_ADMIN suffices to check this for SELinux.
>> What do you think about:
>>
>> int security_check_mutable_hooks(void)
>>
>> Do you have any suggestions on the signature of this hook? Does this
>> hook need to be BPF specific?
> I'd do something like int security_bpf_prog_attach_security(const
> struct bpf_prog *prog) or similar.
> Then the security module can do a check based on the current task
> and/or the prog. We already have some bpf-specific hooks.

I *strongly* disagree with Stephen on this. KRSI and SELinux are peers.
Just as Yama policy is independent of SELinux policy so KRSI policy should
be independent of SELinux policy. I understand the argument that BDF programs
ought to be constrained by SELinux, but I don't think it's right. Further,
we've got unholy layering when security modules call security_ functions.
I'm not saying there is no case where it would be appropriate, but this is not
one of them.