Add a new AUDIT_LANDLOCK_ACCESS record type dedicated to an access
request denied by a Landlock domain. AUDIT_LANDLOCK_ACCESS indicates
that something unexpected happened.
For now, only denied access are logged, which means that any
AUDIT_LANDLOCK_ACCESS record is always followed by a SYSCALL record with
"success=no". However, log parsers should check this syscall property
because this is the only sign that a request was denied. Indeed, we
could have "success=yes" if Landlock would support a "permissive" mode.
We could also add a new field to AUDIT_LANDLOCK_DOMAIN for this mode
(see following commit).
By default, the only logged access requests are those coming from the
same executed program that enforced the Landlock restriction on itself.
In other words, no audit record are created for a task after it called
execve(2). This is required to avoid log spam because programs may only
be aware of their own restrictions, but not the inherited ones.
Following commits will allow to conditionally generate
AUDIT_LANDLOCK_ACCESS records according to dedicated
landlock_restrict_self(2)'s flags.
The AUDIT_LANDLOCK_ACCESS message contains:
- the "domain" ID restricting the action on an object,
- the "blockers" that are missing to allow the requested access,
- a set of fields identifying the related object (e.g. task identified
with "opid" and "ocomm").
The blockers are implicit restrictions (e.g. ptrace), or explicit access
rights (e.g. filesystem), or explicit scopes (e.g. signal). This field
contains a list of at least one element, each separated with a comma.
The initial blocker is "ptrace", which describe all implicit Landlock
restrictions related to ptrace (e.g. deny tracing of tasks outside a
sandbox).
Add audit support to ptrace_access_check and ptrace_traceme hooks. For
the ptrace_access_check case, we log the current/parent domain and the
child task. For the ptrace_traceme case, we log the parent domain and
the current/child task. Indeed, the requester and the target are the
current task, but the action would be performed by the parent task.
Audit event sample:
type=LANDLOCK_ACCESS msg=audit(1729738800.349:44): domain=195ba459b blockers=ptrace opid=1 ocomm="systemd"
type=SYSCALL msg=audit(1729738800.349:44): arch=c000003e syscall=101 success=no [...] pid=300 auid=0
A following commit adds user documentation.
Add KUnit tests to check reading of domain ID relative to layer level.
The quick return for non-landlocked tasks is moved from task_ptrace() to
each LSM hooks.
It is not useful to inline the audit_enabled check because other
computation are performed by landlock_log_denial().
Use scoped guards for RCU read-side critical sections.
Cc: Günther Noack <gnoack@google.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Link: https://lore.kernel.org/r/20250320190717.2287696-10-mic@digikod.net
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Landlock IDs can be generated to uniquely identify Landlock objects.
For now, only Landlock domains get an ID at creation time. These IDs
map to immutable domain hierarchies.
Landlock IDs have important properties:
- They are unique during the lifetime of the running system thanks to
the 64-bit values: at worse, 2^60 - 2*2^32 useful IDs.
- They are always greater than 2^32 and must then be stored in 64-bit
integer types.
- The initial ID (at boot time) is randomly picked between 2^32 and
2^33, which limits collisions in logs across different boots.
- IDs are sequential, which enables users to order them.
- IDs may not be consecutive but increase with a random 2^4 step, which
limits side channels.
Such IDs can be exposed to unprivileged processes, even if it is not the
case with this audit patch series. The domain IDs will be useful for
user space to identify sandboxes and get their properties.
These Landlock IDs are more secure that other absolute kernel IDs such
as pipe's inodes which rely on a shared global counter.
For checkpoint/restore features (i.e. CRIU), we could easily implement a
privileged interface (e.g. sysfs) to set the next ID counter.
IDR/IDA are not used because we only need a bijection from Landlock
objects to Landlock IDs, and we must not recycle IDs. This enables us
to identify all Landlock objects during the lifetime of the system (e.g.
in logs), but not to access an object from an ID nor know if an ID is
assigned. Using a counter is simpler, it scales (i.e. avoids growing
memory footprint), and it does not require locking. We'll use proper
file descriptors (with IDs used as inode numbers) to access Landlock
objects.
Cc: Günther Noack <gnoack@google.com>
Cc: Paul Moore <paul@paul-moore.com>
Link: https://lore.kernel.org/r/20250320190717.2287696-3-mic@digikod.net
Signed-off-by: Mickaël Salaün <mic@digikod.net>
ptrace.[ch] are currently only used for the ptrace LSM hooks but their
scope will expand with IPCs and audit support. Rename ptrace.[ch] to
task.[ch], which better reflect their content. Similarly, rename
landlock_add_ptrace_hooks() to landlock_add_task_hooks(). Keep header
files for now.
Cc: Günther Noack <gnoack@google.com>
Cc: Paul Moore <paul@paul-moore.com>
Link: https://lore.kernel.org/r/20240307093923.1466071-2-mic@digikod.net
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Add network rules support in the ruleset management helpers and the
landlock_create_ruleset() syscall. Extend user space API to support
network actions:
* Add new network access rights: LANDLOCK_ACCESS_NET_BIND_TCP and
LANDLOCK_ACCESS_NET_CONNECT_TCP.
* Add a new network rule type: LANDLOCK_RULE_NET_PORT tied to struct
landlock_net_port_attr. The allowed_access field contains the network
access rights, and the port field contains the port value according to
the controlled protocol. This field can take up to a 64-bit value
but the maximum value depends on the related protocol (e.g. 16-bit
value for TCP). Network port is in host endianness [1].
* Add a new handled_access_net field to struct landlock_ruleset_attr
that contains network access rights.
* Increment the Landlock ABI version to 4.
Implement socket_bind() and socket_connect() LSM hooks, which enable
to control TCP socket binding and connection to specific ports.
Expand access_masks_t from u16 to u32 to be able to store network access
rights alongside filesystem access rights for rulesets' handled access
rights.
Access rights are not tied to socket file descriptors but checked at
bind() or connect() call time against the caller's Landlock domain. For
the filesystem, a file descriptor is a direct access to a file/data.
However, for network sockets, we cannot identify for which data or peer
a newly created socket will give access to. Indeed, we need to wait for
a connect or bind request to identify the use case for this socket.
Likewise a directory file descriptor may enable to open another file
(i.e. a new data item), but this opening is also restricted by the
caller's domain, not the file descriptor's access rights [2].
[1] https://lore.kernel.org/r/278ab07f-7583-a4e0-3d37-1bacd091531d@digikod.net
[2] https://lore.kernel.org/r/263c1eb3-602f-57fe-8450-3f138581bee7@digikod.net
Signed-off-by: Konstantin Meskhidze <konstantin.meskhidze@huawei.com>
Link: https://lore.kernel.org/r/20231026014751.414649-9-konstantin.meskhidze@huawei.com
[mic: Extend commit message, fix typo in comments, and specify
endianness in the documentation]
Co-developed-by: Mickaël Salaün <mic@digikod.net>
Signed-off-by: Mickaël Salaün <mic@digikod.net>
These 3 system calls are designed to be used by unprivileged processes
to sandbox themselves:
* landlock_create_ruleset(2): Creates a ruleset and returns its file
descriptor.
* landlock_add_rule(2): Adds a rule (e.g. file hierarchy access) to a
ruleset, identified by the dedicated file descriptor.
* landlock_restrict_self(2): Enforces a ruleset on the calling thread
and its future children (similar to seccomp). This syscall has the
same usage restrictions as seccomp(2): the caller must have the
no_new_privs attribute set or have CAP_SYS_ADMIN in the current user
namespace.
All these syscalls have a "flags" argument (not currently used) to
enable extensibility.
Here are the motivations for these new syscalls:
* A sandboxed process may not have access to file systems, including
/dev, /sys or /proc, but it should still be able to add more
restrictions to itself.
* Neither prctl(2) nor seccomp(2) (which was used in a previous version)
fit well with the current definition of a Landlock security policy.
All passed structs (attributes) are checked at build time to ensure that
they don't contain holes and that they are aligned the same way for each
architecture.
See the user and kernel documentation for more details (provided by a
following commit):
* Documentation/userspace-api/landlock.rst
* Documentation/security/landlock.rst
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: James Morris <jmorris@namei.org>
Cc: Jann Horn <jannh@google.com>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Acked-by: Serge Hallyn <serge@hallyn.com>
Link: https://lore.kernel.org/r/20210422154123.13086-9-mic@digikod.net
Signed-off-by: James Morris <jamorris@linux.microsoft.com>
Using Landlock objects and ruleset, it is possible to tag inodes
according to a process's domain. To enable an unprivileged process to
express a file hierarchy, it first needs to open a directory (or a file)
and pass this file descriptor to the kernel through
landlock_add_rule(2). When checking if a file access request is
allowed, we walk from the requested dentry to the real root, following
the different mount layers. The access to each "tagged" inodes are
collected according to their rule layer level, and ANDed to create
access to the requested file hierarchy. This makes possible to identify
a lot of files without tagging every inodes nor modifying the
filesystem, while still following the view and understanding the user
has from the filesystem.
Add a new ARCH_EPHEMERAL_INODES for UML because it currently does not
keep the same struct inodes for the same inodes whereas these inodes are
in use.
This commit adds a minimal set of supported filesystem access-control
which doesn't enable to restrict all file-related actions. This is the
result of multiple discussions to minimize the code of Landlock to ease
review. Thanks to the Landlock design, extending this access-control
without breaking user space will not be a problem. Moreover, seccomp
filters can be used to restrict the use of syscall families which may
not be currently handled by Landlock.
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Cc: James Morris <jmorris@namei.org>
Cc: Jann Horn <jannh@google.com>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Serge E. Hallyn <serge@hallyn.com>
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Link: https://lore.kernel.org/r/20210422154123.13086-8-mic@digikod.net
Signed-off-by: James Morris <jamorris@linux.microsoft.com>
Using ptrace(2) and related debug features on a target process can lead
to a privilege escalation. Indeed, ptrace(2) can be used by an attacker
to impersonate another task and to remain undetected while performing
malicious activities. Thanks to ptrace_may_access(), various part of
the kernel can check if a tracer is more privileged than a tracee.
A landlocked process has fewer 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's rules (i.e. the tracee must be in a sub-domain of the tracer).
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Reviewed-by: Jann Horn <jannh@google.com>
Acked-by: Serge Hallyn <serge@hallyn.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20210422154123.13086-5-mic@digikod.net
Signed-off-by: James Morris <jamorris@linux.microsoft.com>
Process's credentials point to a Landlock domain, which is underneath
implemented with a ruleset. In the following commits, this domain is
used to check and enforce the ptrace and filesystem security policies.
A domain is inherited from a parent to its child the same way a thread
inherits a seccomp policy.
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Reviewed-by: Jann Horn <jannh@google.com>
Acked-by: Serge Hallyn <serge@hallyn.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20210422154123.13086-4-mic@digikod.net
Signed-off-by: James Morris <jamorris@linux.microsoft.com>
A Landlock ruleset is mainly a red-black tree with Landlock rules as
nodes. This enables quick update and lookup to match a requested
access, e.g. to a file. A ruleset is usable through a dedicated file
descriptor (cf. following commit implementing syscalls) which enables a
process to create and populate a ruleset with new rules.
A domain is a ruleset tied to a set of processes. This group of rules
defines the security policy enforced on these processes and their future
children. A domain can transition to a new domain which is the
intersection of all its constraints and those of a ruleset provided by
the current process. This modification only impact the current process.
This means that a process can only gain more constraints (i.e. lose
accesses) over time.
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Acked-by: Serge Hallyn <serge@hallyn.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Jann Horn <jannh@google.com>
Link: https://lore.kernel.org/r/20210422154123.13086-3-mic@digikod.net
Signed-off-by: James Morris <jamorris@linux.microsoft.com>
A Landlock object enables to identify a kernel object (e.g. an inode).
A Landlock rule is a set of access rights allowed on an object. Rules
are grouped in rulesets that may be tied to a set of processes (i.e.
subjects) to enforce a scoped access-control (i.e. a domain).
Because Landlock's goal is to empower any process (especially
unprivileged ones) to sandbox themselves, we cannot rely on a
system-wide object identification such as file extended attributes.
Indeed, we need innocuous, composable and modular access-controls.
The main challenge with these constraints is to identify kernel objects
while this identification is useful (i.e. when a security policy makes
use of this object). But this identification data should be freed once
no policy is using it. This ephemeral tagging should not and may not be
written in the filesystem. We then need to manage the lifetime of a
rule according to the lifetime of its objects. To avoid a global lock,
this implementation make use of RCU and counters to safely reference
objects.
A following commit uses this generic object management for inodes.
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Reviewed-by: Jann Horn <jannh@google.com>
Acked-by: Serge Hallyn <serge@hallyn.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20210422154123.13086-2-mic@digikod.net
Signed-off-by: James Morris <jamorris@linux.microsoft.com>
