The SECCOMP_RET_KILL mode is documented as immediately killing the
process as if a SIGSYS had been sent and not caught (similar to a
SIGKILL).  However, a SIGSYS is documented as triggering a coredump
which does not happen today.

This has the advantage of being able to more easily debug a process
that fails a seccomp filter.  Today, most apps need to recompile and
change their filter in order to get detailed info out, or manually run
things through strace, or enable detailed kernel auditing.  Now we get
coredumps that fit into existing system-wide crash reporting setups.

From a security pov, this shouldn't be a problem.  Unhandled signals
can already be sent externally which trigger a coredump independent of
the status of the seccomp filter.  The act of dumping core itself does
not cause change in execution of the program.

URL: https://crbug.com/676357Signed-off-by: Mike Frysinger <vapier@chromium.org>
Acked-by: Jorge Lucangeli Obes <jorgelo@chromium.org>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>

I am still tired of having to find indirect ways to determine
what security modules are active on a system. I have added
/sys/kernel/security/lsm, which contains a comma separated
list of the active security modules. No more groping around
in /proc/filesystems or other clever hacks.

Unchanged from previous versions except for being updated
to the latest security next branch.
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
Acked-by: John Johansen <john.johansen@canonical.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>

The kernel build bot turned up a bad config combination when
CONFIG_SECURITY_APPARMOR is y and CONFIG_SECURITY_APPARMOR_HASH is n,
resulting in the build error
   security/built-in.o: In function `aa_unpack':
   (.text+0x841e2): undefined reference to `aa_g_hash_policy'
Signed-off-by: John Johansen <john.johansen@canonical.com>

AA_BUG() uses WARN and won't break the kernel like BUG_ON().
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

If this sysctl is set to non-zero and a process with CAP_MAC_ADMIN in
the root namespace has created an AppArmor policy namespace,
unprivileged processes will be able to change to a profile in the
newly created AppArmor policy namespace and, if the profile allows
CAP_MAC_ADMIN and appropriate file permissions, will be able to load
policy in the respective policy namespace.
Signed-off-by: Tyler Hicks <tyhicks@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>

Allow a profile to carry extra data that can be queried via userspace.
This provides a means to store extra data in a profile that a trusted
helper can extract and use from live policy.
Signed-off-by: William Hua <william.hua@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>

apparmor should be checking the SECURITY_CAP_NOAUDIT constant. Also
in complain mode make it so apparmor can elect to log a message,
informing of the check.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Allow turning off the computation of the policy hashes via the
apparmor.hash_policy kernel parameter.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Moving the use of fqname to later allows learning profiles to be based
on the fqname request instead of just the hname. It also allows cleaning
up some of the name parsing and lookup by allowing the use of
the fqlookupn_profile() lib fn.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

The aad macro can replace aad strings when it is not intended to. Switch
to a fn macro so it is only applied when intended.

Also at the same time cleanup audit_data initialization by putting
common boiler plate behind a macro, and dropping the gfp_t parameter
which will become useless.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Having ops be an integer that is an index into an op name table is
awkward and brittle. Every op change requires an edit for both the
op constant and a string in the table. Instead switch to using const
strings directly, eliminating the need for the table that needs to
be kept in sync.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Trying to update the task cred while the task current cred is not the
real cred will result in an error at the cred layer. Avoid this by
failing early and delaying the update.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Having per policy ns interface files helps with containers restoring
policy.
Signed-off-by: John Johansen <john.johansen@canonical.com>

This is just setup for new ns specific .load, .replace, .remove interface
files.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Verify that profiles in a load set specify the same policy ns and
audit the name of the policy ns that policy is being loaded for.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Store loaded policy and allow introspecting it through apparmorfs. This
has several uses from debugging, policy validation, and policy checkpoint
and restore for containers.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Policy management will be expanded beyond traditional unconfined root.
This will require knowning the profile of the task doing the management
and the ns view.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Prepare for a tighter pairing of user namespaces and apparmor policy
namespaces, by making the ns to be viewed available.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Prepare for a tighter pairing of user namespaces and apparmor policy
namespaces, by making the ns to be viewed available and checking
that the user namespace level is the same as the policy ns level.

This strict pairing will be relaxed once true support of user namespaces
lands.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

This is prep work for fs operations being able to remove namespaces.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Borrow the special null device file from selinux to "close" fds that
don't have sufficient permissions at exec time.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Commit 9f834ec1 ("binfmt_elf: switch to new creds when switching to new mm")
changed when the creds are installed by the binfmt_elf handler. This
affects which creds are used to mmap the executable into the address
space. Which can have an affect on apparmor policy.

Add a flag to apparmor at
/sys/kernel/security/apparmor/features/domain/fix_binfmt_elf_mmap

to make it possible to detect this semantic change so that the userspace
tools and the regression test suite can correctly deal with the change.

BugLink: http://bugs.launchpad.net/bugs/1630069Signed-off-by: John Johansen <john.johansen@canonical.com>

Instead of testing whether a given dfa exists in every code path, have
a default null dfa that is used when loaded policy doesn't provide a
dfa.

This will let us get rid of special casing and avoid dereference bugs
when special casing is missed.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Newer policy will combine the file and policydb dfas, allowing for
better optimizations. However to support older policy we need to
keep the ability to address the "file" dfa separately. So dup
the policydb as if it is the file dfa and set the appropriate start
state.
Signed-off-by: John Johansen <john.johansen@canonical.com>

The dfa is currently setup to be shared (has the basis of refcounting)
but currently can't be because the count can't be increased.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Newer policy encodes more than just version in the version tag,
so add masking to make sure the comparison remains correct.

Note: this is fully compatible with older policy as it will never set
the bits being masked out.
Signed-off-by: John Johansen <john.johansen@canonical.com>

Signed-off-by: John Johansen <john.johansen@canonical.com>

Policy should always under go a full paranoid verification.
Signed-off-by: John Johansen <john.johansen@canonical.com>

When possible its better to name a learning profile after the missing
profile in question. This allows for both more informative names and
for profile reuse.
Signed-off-by: John Johansen <john.johansen@canonical.com>