Re: [PATCH bpf-next v8 08/11] landlock: Add ptrace restrictions

[MickaÃl SalaÃn]

On 28/02/2018 00:23, Andy Lutomirski wrote:
> On Tue, Feb 27, 2018 at 11:02 PM, Andy Lutomirski <luto@xxxxxxxxxx> wrote:
>> On Tue, Feb 27, 2018 at 10:14 PM, MickaÃl SalaÃn <mic@xxxxxxxxxxx> wrote:
>>>
>>> On 27/02/2018 06:01, Andy Lutomirski wrote:
>>>>
>>>>
>>>>> On Feb 26, 2018, at 8:17 PM, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
>>>>>
>>>>>> On Tue, Feb 27, 2018 at 12:41 AM, MickaÃl SalaÃn <mic@xxxxxxxxxxx> wrote:
>>>>>> A landlocked process has less privileges than a non-landlocked process
>>>>>> and must then be subject to additional restrictions when manipulating
>>>>>> processes. To be allowed to use ptrace(2) and related syscalls on a
>>>>>> target process, a landlocked process must have a subset of the target
>>>>>> process' rules.
>>>>>>
>>>>>> Signed-off-by: MickaÃl SalaÃn <mic@xxxxxxxxxxx>
>>>>>> Cc: Alexei Starovoitov <ast@xxxxxxxxxx>
>>>>>> Cc: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
>>>>>> Cc: Daniel Borkmann <daniel@xxxxxxxxxxxxx>
>>>>>> Cc: David S. Miller <davem@xxxxxxxxxxxxx>
>>>>>> Cc: James Morris <james.l.morris@xxxxxxxxxx>
>>>>>> Cc: Kees Cook <keescook@xxxxxxxxxxxx>
>>>>>> Cc: Serge E. Hallyn <serge@xxxxxxxxxx>
>>>>>> ---
>>>>>>
>>>>>> Changes since v6:
>>>>>> * factor out ptrace check
>>>>>> * constify pointers
>>>>>> * cleanup headers
>>>>>> * use the new security_add_hooks()
>>>>>> ---
>>>>>> security/landlock/Makefile       |   2 +-
>>>>>> security/landlock/hooks_ptrace.c | 124 +++++++++++++++++++++++++++++++++++++++
>>>>>> security/landlock/hooks_ptrace.h |  11 ++++
>>>>>> security/landlock/init.c         |   2 +
>>>>>> 4 files changed, 138 insertions(+), 1 deletion(-)
>>>>>> create mode 100644 security/landlock/hooks_ptrace.c
>>>>>> create mode 100644 security/landlock/hooks_ptrace.h
>>>>>>
>>>>>> diff --git a/security/landlock/Makefile b/security/landlock/Makefile
>>>>>> index d0f532a93b4e..605504d852d3 100644
>>>>>> --- a/security/landlock/Makefile
>>>>>> +++ b/security/landlock/Makefile
>>>>>> @@ -3,4 +3,4 @@ obj-$(CONFIG_SECURITY_LANDLOCK) := landlock.o
>>>>>> landlock-y := init.o chain.o task.o \
>>>>>>        tag.o tag_fs.o \
>>>>>>        enforce.o enforce_seccomp.o \
>>>>>> -       hooks.o hooks_cred.o hooks_fs.o
>>>>>> +       hooks.o hooks_cred.o hooks_fs.o hooks_ptrace.o
>>>>>> diff --git a/security/landlock/hooks_ptrace.c b/security/landlock/hooks_ptrace.c
>>>>>> new file mode 100644
>>>>>> index 000000000000..f1b977b9c808
>>>>>> --- /dev/null
>>>>>> +++ b/security/landlock/hooks_ptrace.c
>>>>>> @@ -0,0 +1,124 @@
>>>>>> +/*
>>>>>> + * Landlock LSM - ptrace hooks
>>>>>> + *
>>>>>> + * Copyright  2017 MickaÃl SalaÃn <mic@xxxxxxxxxxx>
>>>>>> + *
>>>>>> + * This program is free software; you can redistribute it and/or modify
>>>>>> + * it under the terms of the GNU General Public License version 2, as
>>>>>> + * published by the Free Software Foundation.
>>>>>> + */
>>>>>> +
>>>>>> +#include <asm/current.h>
>>>>>> +#include <linux/errno.h>
>>>>>> +#include <linux/kernel.h> /* ARRAY_SIZE */
>>>>>> +#include <linux/lsm_hooks.h>
>>>>>> +#include <linux/sched.h> /* struct task_struct */
>>>>>> +#include <linux/seccomp.h>
>>>>>> +
>>>>>> +#include "common.h" /* struct landlock_prog_set */
>>>>>> +#include "hooks.h" /* landlocked() */
>>>>>> +#include "hooks_ptrace.h"
>>>>>> +
>>>>>> +static bool progs_are_subset(const struct landlock_prog_set *parent,
>>>>>> +               const struct landlock_prog_set *child)
>>>>>> +{
>>>>>> +       size_t i;
>>>>>> +
>>>>>> +       if (!parent || !child)
>>>>>> +               return false;
>>>>>> +       if (parent == child)
>>>>>> +               return true;
>>>>>> +
>>>>>> +       for (i = 0; i < ARRAY_SIZE(child->programs); i++) {
>>>>>
>>>>> ARRAY_SIZE(child->programs) seems misleading.  Is there no define
>>>>> NUM_LANDLOCK_PROG_TYPES or similar?
>>>>>
>>>>>> +               struct landlock_prog_list *walker;
>>>>>> +               bool found_parent = false;
>>>>>> +
>>>>>> +               if (!parent->programs[i])
>>>>>> +                       continue;
>>>>>> +               for (walker = child->programs[i]; walker;
>>>>>> +                               walker = walker->prev) {
>>>>>> +                       if (walker == parent->programs[i]) {
>>>>>> +                               found_parent = true;
>>>>>> +                               break;
>>>>>> +                       }
>>>>>> +               }
>>>>>> +               if (!found_parent)
>>>>>> +                       return false;
>>>>>> +       }
>>>>>> +       return true;
>>>>>> +}
>>>>>
>>>>> If you used seccomp, you'd get this type of check for free, and it
>>>>> would be a lot easier to comprehend.  AFAICT the only extra leniency
>>>>> you're granting is that you're agnostic to the order in which the
>>>>> rules associated with different program types were applied, which
>>>>> could easily be added to seccomp.
>>>>
>>>> On second thought, this is all way too complicated.  I think the correct logic is either "if you are filtered by Landlock, you cannot ptrace anything" or to delete this patch entirely.
>>>
>>> This does not fit a lot of use cases like running a container
>>> constrained with some Landlock programs. We should not deny users the
>>> ability to debug their stuff.
>>>
>>> This patch add the minimal protection which are needed to have
>>> meaningful Landlock security policy. Without it, they may be easily
>>> bypassable, hence useless.
>>>
>>
>> I think you're wrong here.  Any sane container trying to use Landlock
>> like this would also create a PID namespace.  Problem solved.  I still
>> think you should drop this patch.

Containers is one use case, another is build-in sandboxing (e.g. for web
browserâ) and another one is for sandbox managers (e.g. Firejail,
Bubblewrap, Flatpackâ). In some of these use cases, especially from a
developer point of view, you may want/need to debug your applications
(without requiring to be root). For nested Landlock access-controls
(e.g. container + user session + web browser), it may not be allowed to
create a PID namespace, but you still want to have a meaningful
access-control.

>>
>>>
>>>> If something like Tycho's notifiers goes in, then it's not obvious that, just because you have the same set of filters, you have the same privilege.  Similarly, if a feature that lets a filter query its cgroup goes in (and you proposed this once!) then the logic you implemented here is wrong.
>>>
>>> I don't get your point. Please take a look at the tests (patch 10).
>>
>> I don't know what you want me to look at.
>>
>> What I'm saying is: suppose I write a filter like this:
>>
>> bool allow_some_action(void)
>> {
>>   int value_from_container_manager = call_out_to_user_notifier();
>>   return value_from_container_manager == 42;
>> }
>>
>> or
>>
>> bool allow_some_action(void)
>> {
>>   return my_cgroup_is("/foo/bar");
>> }
>>
>> In both of these cases, your code will do the wrong thing.

You are right about the fact that the same filters/programs may not be
equivalent if they use external data (other than from the eBPF context)
to take a decision. This is why using a function
my_cgroup_is("/foo/bar") should not be allowed. If we want to enforce a
security policy according to a cgroup, the Landlock programs should be
pinned on this cgroup. This way, the kernel knows if this programs
should be called or not. It is the same argument I used in the thread
[PATCH bpf-next v8 05/11] about the cache.

The only way a Landlock program may change its behavior is because of an
eBPF map. However, in this case the map is common to all the instances
of this program.

To say it another way, the Landlock's enforce API (currently only
seccomp) is in charge of defining what is a subject. By using seccomp to
enforce a policy, the subject is a hierarchy of processes. By pinning a
Landlock program to a cgroup, the subject is the set of processes under
this cgroup. This is much more efficient than letting a program define
its one subjects. This also allows to audit which processes are
restricted by a set of Landlock programs. Because of that, calls to
functions like bpf_get_current_pid_tgid() should not be allowed (or
limited) for a Landlock program. Let's make this programs as pure as
possible. :)


>>
>>>
>>>>
>>>> Or you could just say that it's the responsibility of a Landlock user to properly filter ptrace() just like it's the responsibility of seccomp users to filter ptrace if needed.
>>>
>>> A user should be able to enforce a security policy on ptrace as well,
>>> but this patch enforce a minimal set of security boundaries. It will be
>>> easy to add a new Landlock program type to get this kind of access control.
>>
>> It sounds like you want Landlock to be a complete container system all
>> by itself.  I disagree with that design goal.
> 
> Having actually read your series more correctly now (oops!), I still
> think that this patch should be dropped.  I can see an argument for
> having a flag that one can set when adding a seccomp filter that says
> "prevent ptrace of any child that doesn't have this exact stack
> installed", but I think that could be added later and should not be
> part of an initial submission.  For now, Landlock users can block
> ptrace() entirely or use PID namespaces.
> 

I also though about using a flag, but we should encourage sane/safe
default behavior, which means at least to not have trivially bypassable
access-control rules, to not shoot yourself in the foot. However, a flag
could be added to disable this safe behavior.

Attachment: signature.asc
Description: OpenPGP digital signature