On 10/04/2017 08:48, Djalal Harouni wrote:

On Wed, Mar 29, 2017 at 1:46 AM, Micka?l Sala?n [off-list ref] wrote:

quoted

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.

New in v6

Signed-off-by: Micka?l Sala?n <mic@digikod.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: David S. Miller <davem@davemloft.net>
Cc: James Morris <redacted>
Cc: Kees Cook <redacted>
Cc: Serge E. Hallyn <serge@hallyn.com>
---
 security/landlock/Makefile       |   2 +-
 security/landlock/hooks_ptrace.c | 126 +++++++++++++++++++++++++++++++++++++++
 security/landlock/hooks_ptrace.h |  11 ++++
 security/landlock/init.c         |   2 +
 4 files changed, 140 insertions(+), 1 deletion(-)
 create mode 100644 security/landlock/hooks_ptrace.c
 create mode 100644 security/landlock/hooks_ptrace.h

[...]

quoted

+/**
+ * landlock_ptrace_access_check - determine whether the current process may
+ *                               access another
+ *
+ * @child: the process to be accessed
+ * @mode: the mode of attachment
+ *
+ * If the current task has Landlock rules, then the child must have at least
+ * the same rules.  Else denied.
+ *
+ * Determine whether a process may access another, returning 0 if permission
+ * granted, -errno if denied.
+ */
+static int landlock_ptrace_access_check(struct task_struct *child,
+               unsigned int mode)
+{
+       if (!landlocked(current))
+               return 0;
+
+       if (!landlocked(child))
+               return -EPERM;
+
+       if (landlock_task_has_subset_events(current, child))
+               return 0;
+
+       return -EPERM;
+}
+

Maybe you want to check the mode argument here if it is a
PTRACE_ATTACH which may translate to read/writes ? PTRACE_READ are
normally for reads only. Or also which creds were used if this was a
direct syscall or a filesystem call through procfs.

The idea is to mimic the behavior of UID/GID checks, namespaces and so
on. A hierarchy of Landlock rules has a similar semantic as namespaces,
at least for now with the FS event.

I'm bringing this, since you may want to make some room for landlock
ptrace events and what others may want to do with it. Also I'm

I don't see any no problem to add a ptrace event in the future as long
as the composition with this default rule is a logical AND to allow a
ptrace action.

It would be possible to relax this default policy for rules with a
dedicated option flag, but the current behavior is a sane one from a
security point of view.

planning to send another v2 RFC for procfs separate instances [1], the
aim is to give LSMs a security_ptrace_access_check hook path when
dealing with /proc/<pids>/ [2]  . Right now LSMs don't really have a
security path there, and the implementation does not guarantee that.
With this Yama ptrace scope or other LSMs may take advantage of it and
check the 'PTRACE_MODE_READ_FSCRED' mode for filesystem accesses.

Interesting, feel free to CC me.

That's why I think it would be better if the default landlock ptrace
semantics are not that wide.

Thanks!

[1] https://lkml.org/lkml/2017/3/30/670
[2] http://lxr.free-electrons.com/source/fs/proc/base.c#L719

If a task is allowed to ptrace/read the memory of another task with
different privileges, the tracer could also access sensitive data not
allowed otherwise.

Do you have an use case where this constraint would be an issue?

 Micka?l

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://kernsec.org/pipermail/linux-security-module-archive/attachments/20170411/89fd76ee/attachment.sig>