Before these patches, the Userspace Path Manager would allow the creation of subflows with wrong families: taking the one of the MPTCP socket instead of the provided ones and resulting in the creation of subflows with likely not the right source and/or destination IPs. It would also allow the creation of subflows between different families or not respecting v4/v6-only socket attributes.
Patch 1 lets the userspace PM select the proper family to avoid creating subflows with the wrong source and/or destination addresses because the family is not the expected one.
Patch 2 makes sure the userspace PM doesn't allow the userspace to create subflows for a family that is not allowed.
Patch 3 validates scenarios with a mix of v4 and v6 subflows for the same MPTCP connection.
These patches fix issues introduced in v5.19 when the userspace path manager has been introduced.
To: "David S. Miller" davem@davemloft.net To: Eric Dumazet edumazet@google.com To: Jakub Kicinski kuba@kernel.org To: Kishen Maloor kishen.maloor@intel.com To: Florian Westphal fw@strlen.de To: Shuah Khan shuah@kernel.org Cc: netdev@vger.kernel.org Cc: mptcp@lists.linux.dev Cc: linux-kernel@vger.kernel.org Cc: linux-kselftest@vger.kernel.org Cc: Paolo Abeni pabeni@redhat.com Cc: Mat Martineau mathew.j.martineau@linux.intel.com Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net
--- Matthieu Baerts (2): mptcp: netlink: respect v4/v6-only sockets selftests: mptcp: userspace: validate v4-v6 subflows mix
Paolo Abeni (1): mptcp: explicitly specify sock family at subflow creation time
net/mptcp/pm.c | 25 ++++++++++++ net/mptcp/pm_userspace.c | 7 ++++ net/mptcp/protocol.c | 2 +- net/mptcp/protocol.h | 6 ++- net/mptcp/subflow.c | 9 +++-- tools/testing/selftests/net/mptcp/userspace_pm.sh | 47 +++++++++++++++++++++++ 6 files changed, 90 insertions(+), 6 deletions(-) --- base-commit: be53771c87f4e322a9835d3faa9cd73a4ecdec5b change-id: 20230112-upstream-net-20230112-netlink-v4-v6-b6b958039ee0
Best regards,
From: Paolo Abeni pabeni@redhat.com
Let the caller specify the to-be-created subflow family.
For a given MPTCP socket created with the AF_INET6 family, the current userspace PM can already ask the kernel to create subflows in v4 and v6. If "plain" IPv4 addresses are passed to the kernel, they are automatically mapped in v6 addresses "by accident". This can be problematic because the userspace will need to pass different addresses, now the v4-mapped-v6 addresses to destroy this new subflow.
On the other hand, if the MPTCP socket has been created with the AF_INET family, the command to create a subflow in v6 will be accepted but the result will not be the one as expected as new subflow will be created in IPv4 using part of the v6 addresses passed to the kernel: not creating the expected subflow then.
No functional change intended for the in-kernel PM where an explicit enforcement is currently in place. This arbitrary enforcement will be leveraged by other patches in a future version.
Fixes: 702c2f646d42 ("mptcp: netlink: allow userspace-driven subflow establishment") Cc: stable@vger.kernel.org Co-developed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Mat Martineau mathew.j.martineau@linux.intel.com Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- net/mptcp/protocol.c | 2 +- net/mptcp/protocol.h | 3 ++- net/mptcp/subflow.c | 9 +++++---- 3 files changed, 8 insertions(+), 6 deletions(-)
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index b7ad030dfe89..8cd6cc67c2c5 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -98,7 +98,7 @@ static int __mptcp_socket_create(struct mptcp_sock *msk) struct socket *ssock; int err;
- err = mptcp_subflow_create_socket(sk, &ssock); + err = mptcp_subflow_create_socket(sk, sk->sk_family, &ssock); if (err) return err;
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h index a0d1658ce59e..a9e0355744b6 100644 --- a/net/mptcp/protocol.h +++ b/net/mptcp/protocol.h @@ -641,7 +641,8 @@ bool mptcp_addresses_equal(const struct mptcp_addr_info *a, /* called with sk socket lock held */ int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, const struct mptcp_addr_info *remote); -int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); +int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, + struct socket **new_sock); void mptcp_info2sockaddr(const struct mptcp_addr_info *info, struct sockaddr_storage *addr, unsigned short family); diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c index bd387d4b5a38..ec54413fb31f 100644 --- a/net/mptcp/subflow.c +++ b/net/mptcp/subflow.c @@ -1547,7 +1547,7 @@ int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, if (!mptcp_is_fully_established(sk)) goto err_out;
- err = mptcp_subflow_create_socket(sk, &sf); + err = mptcp_subflow_create_socket(sk, loc->family, &sf); if (err) goto err_out;
@@ -1660,7 +1660,9 @@ static void mptcp_subflow_ops_undo_override(struct sock *ssk) #endif ssk->sk_prot = &tcp_prot; } -int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock) + +int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, + struct socket **new_sock) { struct mptcp_subflow_context *subflow; struct net *net = sock_net(sk); @@ -1673,8 +1675,7 @@ int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock) if (unlikely(!sk->sk_socket)) return -EINVAL;
- err = sock_create_kern(net, sk->sk_family, SOCK_STREAM, IPPROTO_TCP, - &sf); + err = sock_create_kern(net, family, SOCK_STREAM, IPPROTO_TCP, &sf); if (err) return err;
If an MPTCP socket has been created with AF_INET6 and the IPV6_V6ONLY option has been set, the userspace PM would allow creating subflows using IPv4 addresses, e.g. mapped in v6.
The kernel side of userspace PM will also accept creating subflows with local and remote addresses having different families. Depending on the subflow socket's family, different behaviours are expected: - If AF_INET is forced with a v6 address, the kernel will take the last byte of the IP and try to connect to that: a new subflow is created but to a non expected address. - If AF_INET6 is forced with a v4 address, the kernel will try to connect to a v4 address (v4-mapped-v6). A -EBADF error from the connect() part is then expected.
It is then required to check the given families can be accepted. This is done by using a new helper for addresses family matching, taking care of IPv4 vs IPv4-mapped-IPv6 addresses. This helper will be re-used later by the in-kernel path-manager to use mixed IPv4 and IPv6 addresses.
While at it, a clear error message is now reported if there are some conflicts with the families that have been passed by the userspace.
Fixes: 702c2f646d42 ("mptcp: netlink: allow userspace-driven subflow establishment") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau mathew.j.martineau@linux.intel.com Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- net/mptcp/pm.c | 25 +++++++++++++++++++++++++ net/mptcp/pm_userspace.c | 7 +++++++ net/mptcp/protocol.h | 3 +++ 3 files changed, 35 insertions(+)
diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c index 45e2a48397b9..70f0ced3ca86 100644 --- a/net/mptcp/pm.c +++ b/net/mptcp/pm.c @@ -420,6 +420,31 @@ void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk) } }
+/* if sk is ipv4 or ipv6_only allows only same-family local and remote addresses, + * otherwise allow any matching local/remote pair + */ +bool mptcp_pm_addr_families_match(const struct sock *sk, + const struct mptcp_addr_info *loc, + const struct mptcp_addr_info *rem) +{ + bool mptcp_is_v4 = sk->sk_family == AF_INET; + +#if IS_ENABLED(CONFIG_MPTCP_IPV6) + bool loc_is_v4 = loc->family == AF_INET || ipv6_addr_v4mapped(&loc->addr6); + bool rem_is_v4 = rem->family == AF_INET || ipv6_addr_v4mapped(&rem->addr6); + + if (mptcp_is_v4) + return loc_is_v4 && rem_is_v4; + + if (ipv6_only_sock(sk)) + return !loc_is_v4 && !rem_is_v4; + + return loc_is_v4 == rem_is_v4; +#else + return mptcp_is_v4 && loc->family == AF_INET && rem->family == AF_INET; +#endif +} + void mptcp_pm_data_reset(struct mptcp_sock *msk) { u8 pm_type = mptcp_get_pm_type(sock_net((struct sock *)msk)); diff --git a/net/mptcp/pm_userspace.c b/net/mptcp/pm_userspace.c index 65dcc55a8ad8..ea6ad9da7493 100644 --- a/net/mptcp/pm_userspace.c +++ b/net/mptcp/pm_userspace.c @@ -294,6 +294,13 @@ int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info) }
sk = (struct sock *)msk; + + if (!mptcp_pm_addr_families_match(sk, &addr_l, &addr_r)) { + GENL_SET_ERR_MSG(info, "families mismatch"); + err = -EINVAL; + goto create_err; + } + lock_sock(sk);
err = __mptcp_subflow_connect(sk, &addr_l, &addr_r); diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h index a9e0355744b6..601469249da8 100644 --- a/net/mptcp/protocol.h +++ b/net/mptcp/protocol.h @@ -777,6 +777,9 @@ int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, bool require_family, struct mptcp_pm_addr_entry *entry); +bool mptcp_pm_addr_families_match(const struct sock *sk, + const struct mptcp_addr_info *loc, + const struct mptcp_addr_info *rem); void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
MPTCP protocol supports having subflows in both IPv4 and IPv6. In Linux, it is possible to have that if the MPTCP socket has been created with AF_INET6 family without the IPV6_V6ONLY option.
Here, a new IPv4 subflow is being added to the initial IPv6 connection, then being removed using Netlink commands.
Cc: stable@vger.kernel.org # v5.19+ Reviewed-by: Mat Martineau mathew.j.martineau@linux.intel.com Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- tools/testing/selftests/net/mptcp/userspace_pm.sh | 47 +++++++++++++++++++++++ 1 file changed, 47 insertions(+)
diff --git a/tools/testing/selftests/net/mptcp/userspace_pm.sh b/tools/testing/selftests/net/mptcp/userspace_pm.sh index a29deb9fa024..ab2d581f28a1 100755 --- a/tools/testing/selftests/net/mptcp/userspace_pm.sh +++ b/tools/testing/selftests/net/mptcp/userspace_pm.sh @@ -752,6 +752,52 @@ test_subflows() "$server4_token" > /dev/null 2>&1 }
+test_subflows_v4_v6_mix() +{ + # Attempt to add a listener at 10.0.2.1:<subflow-port> + ip netns exec "$ns1" ./pm_nl_ctl listen 10.0.2.1\ + $app6_port > /dev/null 2>&1 & + local listener_pid=$! + + # ADD_ADDR4 from server to client machine reusing the subflow port on + # the established v6 connection + :>"$client_evts" + ip netns exec "$ns1" ./pm_nl_ctl ann 10.0.2.1 token "$server6_token" id\ + $server_addr_id dev ns1eth2 > /dev/null 2>&1 + stdbuf -o0 -e0 printf "ADD_ADDR4 id:%d 10.0.2.1 (ns1) => ns2, reuse port\t\t" $server_addr_id + sleep 0.5 + verify_announce_event "$client_evts" "$ANNOUNCED" "$client6_token" "10.0.2.1"\ + "$server_addr_id" "$app6_port" + + # CREATE_SUBFLOW from client to server machine + :>"$client_evts" + ip netns exec "$ns2" ./pm_nl_ctl csf lip 10.0.2.2 lid 23 rip 10.0.2.1 rport\ + $app6_port token "$client6_token" > /dev/null 2>&1 + sleep 0.5 + verify_subflow_events "$client_evts" "$SUB_ESTABLISHED" "$client6_token"\ + "$AF_INET" "10.0.2.2" "10.0.2.1" "$app6_port" "23"\ + "$server_addr_id" "ns2" "ns1" + + # Delete the listener from the server ns, if one was created + kill_wait $listener_pid + + sport=$(sed --unbuffered -n 's/.*(sport:)([[:digit:]]*).*$/\2/p;q' "$client_evts") + + # DESTROY_SUBFLOW from client to server machine + :>"$client_evts" + ip netns exec "$ns2" ./pm_nl_ctl dsf lip 10.0.2.2 lport "$sport" rip 10.0.2.1 rport\ + $app6_port token "$client6_token" > /dev/null 2>&1 + sleep 0.5 + verify_subflow_events "$client_evts" "$SUB_CLOSED" "$client6_token" \ + "$AF_INET" "10.0.2.2" "10.0.2.1" "$app6_port" "23"\ + "$server_addr_id" "ns2" "ns1" + + # RM_ADDR from server to client machine + ip netns exec "$ns1" ./pm_nl_ctl rem id $server_addr_id token\ + "$server6_token" > /dev/null 2>&1 + sleep 0.5 +} + test_prio() { local count @@ -861,6 +907,7 @@ make_connection "v6" test_announce test_remove test_subflows +test_subflows_v4_v6_mix test_prio test_listener
Hello:
This series was applied to netdev/net.git (master) by Jakub Kicinski kuba@kernel.org:
On Thu, 12 Jan 2023 18:42:51 +0100 you wrote:
Before these patches, the Userspace Path Manager would allow the creation of subflows with wrong families: taking the one of the MPTCP socket instead of the provided ones and resulting in the creation of subflows with likely not the right source and/or destination IPs. It would also allow the creation of subflows between different families or not respecting v4/v6-only socket attributes.
[...]
Here is the summary with links: - [net,1/3] mptcp: explicitly specify sock family at subflow creation time https://git.kernel.org/netdev/net/c/6bc1fe7dd748 - [net,2/3] mptcp: netlink: respect v4/v6-only sockets https://git.kernel.org/netdev/net/c/fb00ee4f3343 - [net,3/3] selftests: mptcp: userspace: validate v4-v6 subflows mix https://git.kernel.org/netdev/net/c/4656d72c1efa
You are awesome, thank you!
linux-kselftest-mirror@lists.linaro.org
-
Matthieu Baerts -
patchwork-bot+netdevbpf＠kernel.org