Skip to contentSkip to conversation

Use socket's Landlock domain

RFC PATCH In review Mickaël Salaün 2 participants 6 messages Jul 19, 2024 – Jul 23, 2024

Subsystems (5)
lore
Share & tools
lore

Mickaël Salaün

author maintainer
#0 lore raw

cover Use socket's Landlock domain

Hi,

While the current approach works, I think we should change the way Landlock restricts network actions. Because this feature is relatively new, we can still fix this inconsistency. In a nutshell, let's follow a more capability-based model. Please let me know what you think.

Regards,

Mickaël Salaün (3):
landlock: Use socket's domain instead of current's domain selftests/landlock: Add test for socket's domain
landlock: Document network restrictions tied to sockets

 Documentation/userspace-api/landlock.rst    |  4 ++-
 security/landlock/net.c                     | 22 ++++++++--------
 tools/testing/selftests/landlock/net_test.c | 29 +++++++++++++++++++++
 3 files changed, 43 insertions(+), 12 deletions(-)

base-commit: f4b89d8ce5a835afa51404977ee7e3889c2b9722

Signature · 1 line
2.45.2
author maintainer
#1 lore raw

RFC PATCH 1/3 landlock: Use socket's domain instead of current's domain

Before this change, network restrictions were enforced according to the calling thread's Landlock domain, leading to potential inconsistent results when the same socket was used by different threads or processes (with different domains). This change fixes such access control inconsistency by enforcing the socket's Landlock domain instead of the caller's Landlock domain.

Socket's Landlock domain is inherited from the thread that created this socket. This means that a socket created without sandboxing will be free to connect and bind without limitation. This also means that a socket created by a sandboxed thread will inherit the thread's policy, which will be enforced on this socket even when used by another thread or passed to another process.

The initial rationale [1] was that a socket does not directly grants access to data, but it is an object used to define an access (e.g. connection to a peer). Contrary to my initial assumption, we can identify to which protocol/port a newly created socket can give access to with the socket's file->f_cred inherited from its creator. Moreover, from a kernel point of view, especially for shared objects, we need a more consistent access model. This means that the same action on the same socket performed by different threads will have the same effect. This follows the same approach as for file descriptors tied to the file system (e.g. LANDLOCK_ACCESS_FS_TRUNCATE).

One potential risk of this change is for unsandboxed processes to send socket file descriptors to sandboxed processes, which could give unrestricted network access to the sandboxed process (by reconfigure the socket). While it makes sense for processes to transfer (AF_UNIX) socketpairs, which is OK because they can only exchange data between themselves, it should be rare for processes to legitimately pass other kind of sockets (e.g. AF_INET).

Another potential risk of this approach is socket file descriptor leaks. This is the same risk as with regular file descriptor leaks giving access to the content of a file, which is well known and documented. This could be mitigated with a future complementary restriction on received or inherited file descriptors.

One interesting side effect of this new approach is that a process can create a socket that will only allow to connect to a set of ports. This can be done by creating a thread, sandboxing it, creating a socket, and using the related file descriptor (in the same process). Passing this restricted socket to a more sandboxed process makes it possible to have a more dynamic security policy.

This new approach aligns with SELinux and Smack instead of AppArmor and Tomoyo. It is also in line with capability-based security mechanisms such as Capsicum.

This slight semantic change is important for current and future Landlock's consistency, and it must be backported.

Current tests are still OK because this behavior wasn't covered. A following commit adds new tests.

Cc Günther Noack <[email protected]>
Cc Ivanov Mikhail <[email protected]>
Cc Konstantin Meskhidze <[email protected]>
Cc Paul Moore <[email protected]>
Cc Tahera Fahimi <[email protected]>
Cc <[email protected]> # 6.7.x: 088e2efaf3d2: landlock: Simplify current_check_access_socket()
Fixes fff69fb03dde ("landlock: Support network rules with TCP bind and connect")
---
 security/landlock/net.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)
security/landlock/net.c +11 -11
index c8bcd29bde09..78e027a74819 100644
@@ -50,10 +50,11 @@ get_raw_handled_net_accesses(const struct landlock_ruleset *const domain)
 	return access_dom;
 }
 
-static const struct landlock_ruleset *get_current_net_domain(void)
+static const struct landlock_ruleset *
+get_socket_net_domain(const struct socket *const sock)
 {
 	const struct landlock_ruleset *const dom =
-		landlock_get_current_domain();
+		landlock_cred(sock->file->f_cred)->domain;
 
 	if (!dom || !get_raw_handled_net_accesses(dom))
 		return NULL;
@@ -61,10 +62,9 @@ static const struct landlock_ruleset *get_current_net_domain(void)
 	return dom;
 }
 
-static int current_check_access_socket(struct socket *const sock,
-				       struct sockaddr *const address,
-				       const int addrlen,
-				       access_mask_t access_request)
+static int check_access_socket(struct socket *const sock,
+			       struct sockaddr *const address,
+			       const int addrlen, access_mask_t access_request)
 {
 	__be16 port;
 	layer_mask_t layer_masks[LANDLOCK_NUM_ACCESS_NET] = {};
@@ -72,7 +72,7 @@ static int current_check_access_socket(struct socket *const sock,
 	struct landlock_id id = {
 		.type = LANDLOCK_KEY_NET_PORT,
 	};
-	const struct landlock_ruleset *const dom = get_current_net_domain();
+	const struct landlock_ruleset *const dom = get_socket_net_domain(sock);
 
 	if (!dom)
 		return 0;
@@ -175,16 +175,16 @@ static int current_check_access_socket(struct socket *const sock,
 static int hook_socket_bind(struct socket *const sock,
 			    struct sockaddr *const address, const int addrlen)
 {
-	return current_check_access_socket(sock, address, addrlen,
-					   LANDLOCK_ACCESS_NET_BIND_TCP);
+	return check_access_socket(sock, address, addrlen,
+				   LANDLOCK_ACCESS_NET_BIND_TCP);
 }
 
 static int hook_socket_connect(struct socket *const sock,
 			       struct sockaddr *const address,
 			       const int addrlen)
 {
-	return current_check_access_socket(sock, address, addrlen,
-					   LANDLOCK_ACCESS_NET_CONNECT_TCP);
+	return check_access_socket(sock, address, addrlen,
+				   LANDLOCK_ACCESS_NET_CONNECT_TCP);
 }
 
 static struct security_hook_list landlock_hooks[] __ro_after_init = {
Signature · 1 line
2.45.2
author maintainer
#2 lore raw

RFC PATCH 2/3 selftests/landlock: Add test for socket's domain

In reply to Mickaël Salaün

This new ipv4_tcp.socket_domain test checks that the restrictions are tied to the socket at creation time, but not tied to the thread requesting a bind action.

Properly close file descriptor in ipv4.with_fs test.

Cc Günther Noack <[email protected]>
Cc Ivanov Mikhail <[email protected]>
Cc Konstantin Meskhidze <[email protected]>
Cc Paul Moore <[email protected]>
Cc Tahera Fahimi <[email protected]>
Fixes a549d055a22e ("selftests/landlock: Add network tests")
---
 tools/testing/selftests/landlock/net_test.c | 29 +++++++++++++++++++++
 1 file changed, 29 insertions(+)
tools/testing/selftests/landlock/net_test.c +29 -0
index f21cfbbc3638..79251e27d26d 100644
@@ -1579,6 +1579,35 @@ TEST_F(ipv4_tcp, with_fs)
 	bind_fd = socket(AF_INET, SOCK_STREAM | SOCK_CLOEXEC, 0);
 	ASSERT_LE(0, bind_fd);
 	EXPECT_EQ(-EACCES, bind_variant(bind_fd, &self->srv1));
+	EXPECT_EQ(0, close(bind_fd));
+}
+
+TEST_F(ipv4_tcp, socket_domain)
+{
+	const struct landlock_ruleset_attr ruleset_attr = {
+		.handled_access_net = LANDLOCK_ACCESS_NET_BIND_TCP,
+	};
+	int ruleset_fd, bind_fd;
+
+	/* Creates socket before sandboxing. */
+	bind_fd = socket(AF_INET, SOCK_STREAM | SOCK_CLOEXEC, 0);
+	ASSERT_LE(0, bind_fd);
+
+	ruleset_fd =
+		landlock_create_ruleset(&ruleset_attr, sizeof(ruleset_attr), 0);
+	ASSERT_LE(0, ruleset_fd);
+	enforce_ruleset(_metadata, ruleset_fd);
+	EXPECT_EQ(0, close(ruleset_fd));
+
+	/* Tests port binding with unsandboxed socket. */
+	EXPECT_EQ(0, bind_variant(bind_fd, &self->srv1));
+	EXPECT_EQ(0, close(bind_fd));
+
+	/* Tests port binding with new sandboxed socket. */
+	bind_fd = socket(AF_INET, SOCK_STREAM | SOCK_CLOEXEC, 0);
+	ASSERT_LE(0, bind_fd);
+	EXPECT_EQ(-EACCES, bind_variant(bind_fd, &self->srv1));
+	EXPECT_EQ(0, close(bind_fd));
 }
 
 FIXTURE(port_specific)
Signature · 1 line
2.45.2
author maintainer
#3 lore raw

RFC PATCH 3/3 landlock: Document network restrictions tied to sockets

In reply to Mickaël Salaün

The Landlock domain used to restrict operations on a socket is the domain from the thread that created this socket.

Cc Günther Noack <[email protected]>
Cc Ivanov Mikhail <[email protected]>
Cc Konstantin Meskhidze <[email protected]>
Cc Paul Moore <[email protected]>
Cc Tahera Fahimi <[email protected]>
---
 Documentation/userspace-api/landlock.rst | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
Documentation/userspace-api/landlock.rst +3 -1
index 37dafce8038b..4a9bfff575d5 100644
@@ -529,7 +529,9 @@ Network support (ABI < 4)
 Starting with the Landlock ABI version 4, it is now possible to restrict TCP
 bind and connect actions to only a set of allowed ports thanks to the new
 ``LANDLOCK_ACCESS_NET_BIND_TCP`` and ``LANDLOCK_ACCESS_NET_CONNECT_TCP``
-access rights.
+access rights.  These restrictions are tied to a socket and are inherited from
+the sandboxed thread that created this socket.  Hence, sockets created before
+sandboxing are not restricted.
 
 IOCTL (ABI < 5)
 ---------------
Signature · 1 line
2.45.2
#4 lore raw
In reply to Mickaël Salaün

Hello Mickaël!

Quoted 4 lines · Mickaël Salaün

While the current approach works, I think we should change the way Landlock restricts network actions. Because this feature is relatively new, we can still fix this inconsistency. In a nutshell, let's follow a more capability-based model. Please let me know what you think.

Thanks for sending the patch. The implementation with ->f_cred is much simpler than I had thought it would be. Some higher level questions:

* I assume that the plan is to backport this as a fix to older kernels that already have the feature? (Otherwise, we would potentially have backwards compatibility issues.)

* I believe it clashes a little bit with the TCP server example [1], which I found a useful use case for the TCP connect/bind and socket restriction features.

* accept(2) on a passive (listen-mode) socket will give you a new socket FD -- does that new socket inherit its f_cred from the server socket, or does it inherit its f_cred from the thread?

Regarding the TCP server example, the current implementation is *very* simple, and does the following steps:

1. create socket with socket(2)
2. bind(2) the socket to the desired port
3. enforce a Landlock ruleset that disables all networking features (TCP bind, TCP connect and socket creation with the new patch set)
4. listen(2) on the socket
5. go into the accept(2) loop

With the old behaviour, step 3 is going to affect the existing passive socket:
It will not be possible any more to bind(2) that passive socket to another port.

With the new behaviour (after your patch), step 3 does *not* affect the existing socket, and the server socket can be reused to bind(2) to other ports.

Or, in other words: If the relevant domain is tied to the socket at creation time, that means that a future client connection which takes over the process might be able to use that socket's Landlock domain, which potentially grants more permissions than the thread's domain

I think it would be nice if a use case like in the TCP server example would still be possible with Landlock; there are multiple ways to reach that:

- We could enforce two layers of Landlock rules, one before socket creation that restricts bind(2) to a given port, and one after socket creation that restricts other bind(2), create(2) and socket(2) operations.

Drawbacks:

- One Landlock layer more, and needs to add a Landlock rule:
This is a bit more complicated to implement.
- The bind(2) restriction on the socket is still only per port *number*, but it is still possible to bind with that port number on different IP addresses.

- Alternatively, I wish we could just lock the passive server socket in, so that it can't be made to reconnect anywhere else. After all, the socket disassociation with connect(2) with AF_UNSPEC seems to be a somewhat obscure and seldomly used feature - if we could just disallow that operation, we could ensure that this socket gets reused for such a nefarious purpose.

It would still require two nested Landlock rulesets to make the TCP server example work, but they would be a bit simpler than in the alternative above.

- There are probably more alternatives...?

What do you think?

—Günther

[1] https://wiki.gnoack.org/LandlockTcpServerExample

author maintainer
#5 lore raw
Quoted 8 lines · Günther Noack

Hello Mickaël!

Quoted 4 lines · Mickaël Salaün

While the current approach works, I think we should change the way Landlock restricts network actions. Because this feature is relatively new, we can still fix this inconsistency. In a nutshell, let's follow a more capability-based model. Please let me know what you think.

Quoted 6 lines

Thanks for sending the patch. The implementation with ->f_cred is much simpler than I had thought it would be. Some higher level questions:

* I assume that the plan is to backport this as a fix to older kernels that already have the feature? (Otherwise, we would potentially have backwards compatibility issues.)

Correct, if this patch is merged it must be backported too, but there might be better alternatives, or we might just stick to the initial approach.

Quoted 4 lines

* I believe it clashes a little bit with the TCP server example [1], which I found a useful use case for the TCP connect/bind and socket restriction features.

Indeed, because the socket is created before the sandboxing, the socket could be reused to bind (or connect) to other ports. This is a good example of why using current's instead of socket's credentials may be less surprising. From my point of view, the main issue is that a socket can be reconfigured.

Quoted 4 lines

* accept(2) on a passive (listen-mode) socket will give you a new socket FD -- does that new socket inherit its f_cred from the server socket, or does it inherit its f_cred from the thread?

According to sock_alloc_file(), the newly created socket inherits the caller's credentials, which is similar to a call to openat2(2) with a directory file descriptor.

Quoted 14 lines

Regarding the TCP server example, the current implementation is *very* simple, and does the following steps:

1. create socket with socket(2)
2. bind(2) the socket to the desired port
3. enforce a Landlock ruleset that disables all networking features (TCP bind, TCP connect and socket creation with the new patch set)
4. listen(2) on the socket
5. go into the accept(2) loop

With the old behaviour, step 3 is going to affect the existing passive socket:
It will not be possible any more to bind(2) that passive socket to another port.

Quoted 7 lines

With the new behaviour (after your patch), step 3 does *not* affect the existing socket, and the server socket can be reused to bind(2) to other ports.

Or, in other words: If the relevant domain is tied to the socket at creation time, that means that a future client connection which takes over the process might be able to use that socket's Landlock domain, which potentially grants more permissions than the thread's domain

Yes, that's why it might be more risky, but I wanted to have this discussion. Whatever the outcome, it should be explained in Documentation/security/landlock.rst

One thing to keep in mind and that contrary to other LSMs, Landlock, like seccomp, enables processes to (only) drop privileges, and this is done by the process itself (not at execve time). This was also one argument for the initial approach.

A thing that bothered me was related to the restrictions of sockets, especially with the WIP scoping feature. Datagram (unix) sockets could work before sandboxing, and suddenly become broken after sandboxing (because the security check would be done at send time instead of connect time). This kind of issue should be identified quite early and easily though.

However, there is still an inconsistency between connected stream sockets and datagram sockets. From a security point of view, this looks like a good thing though.

Quoted 4 lines

I think it would be nice if a use case like in the TCP server example would still be possible with Landlock; there are multiple ways to reach that:

Quoted 8 lines

- We could enforce two layers of Landlock rules, one before socket creation that restricts bind(2) to a given port, and one after socket creation that restricts other bind(2), create(2) and socket(2) operations.

Drawbacks:

- One Landlock layer more, and needs to add a Landlock rule:
This is a bit more complicated to implement.

Right, I think it's too complex for users.

Quoted 3 lines

- The bind(2) restriction on the socket is still only per port *number*, but it is still possible to bind with that port number on different IP addresses.

Good point. That's another argument for the initial approach and the way you sandboxed the example: dropping the *_TCP access rights, it is not possible to rebind or reconnect a socket (to another address).

Actually, I'm not sure if using the socket's credential would not confuse users to understand why an access is denied (or allowed).

Quoted 6 lines

- Alternatively, I wish we could just lock the passive server socket in, so that it can't be made to reconnect anywhere else. After all, the socket disassociation with connect(2) with AF_UNSPEC seems to be a somewhat obscure and seldomly used feature - if we could just disallow that operation, we could ensure that this socket gets reused for such a nefarious purpose.

We could also add a new "scope" for socket reconfiguration of sockets created by a parent or sibling domain, similar to the ptrace restrictions.

Quoted 7 lines

It would still require two nested Landlock rulesets to make the TCP server example work, but they would be a bit simpler than in the alternative above.

- There are probably more alternatives...?

What do you think?

I see other alternatives:

- We could have a new ruleset's attribute to specify if network restrictions should apply on the caller or the socket. That might be confusing for users though.

- We could just stick to the initial approach and add new access rights (denied by default, similar to FS_REFER) that will only apply to newly created sockets. This is close to the previous alternative but more explicit. Both use cases could then be used, with a default secure approach (i.e. the initial one). However, we need to have a clear rationale for the WIP scoping restrictions: should the caller or the socket be checked as the client?

- We could extend the current approach and check both the caller's credential and the socket's credential. This could be confusing to users though.

- We could have a new fcntl(2) command to (securely) transition a file descriptor's credential to the caller's one (e.g. approved by ptrace_may_access). That could be generic to all Linux access control systems.

- According to a new ruleset's attribute, we could revalidate (at use time) file descriptors not opened by the caller's Landlock domain, but users would have to be explicit (e.g. stdio issue). And how to handle partially allowed accesses?

Quoted 5 lines

—Günther

[1] https://wiki.gnoack.org/LandlockTcpServerExample