When the server has MPTCP enabled but receives a non-MP-capable request from a client, it calls mptcp_fallback_tcp_ops().
Since non-MPTCP connections are allowed to use sockmap, which replaces sk->sk_prot, using sk->sk_prot to determine the IP version in mptcp_fallback_tcp_ops() becomes unreliable. This can lead to assigning incorrect ops to sk->sk_socket->ops.
Additionally, when BPF Sockmap modifies the protocol handlers, the original WARN_ON_ONCE(sk->sk_prot != &tcp_prot) check would falsely trigger warnings.
Fix this by using the more stable sk_family to distinguish between IPv4 and IPv6 connections, ensuring correct fallback protocol operations are selected even when BPF Sockmap has modified the socket protocol handlers.
Fixes: 0b4f33def7bb ("mptcp: fix tcp fallback crash") Cc: stable@vger.kernel.org Signed-off-by: Jiayuan Chen jiayuan.chen@linux.dev Reviewed-by: Jakub Sitnicki jakub@cloudflare.com --- net/mptcp/protocol.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 0292162a14ee..2393741bc310 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -61,11 +61,16 @@ static u64 mptcp_wnd_end(const struct mptcp_sock *msk)
static const struct proto_ops *mptcp_fallback_tcp_ops(const struct sock *sk) { + /* When BPF sockmap is used, it may replace sk->sk_prot. + * Using sk_family is a reliable way to determine the IP version. + */ + unsigned short family = READ_ONCE(sk->sk_family); + #if IS_ENABLED(CONFIG_MPTCP_IPV6) - if (sk->sk_prot == &tcpv6_prot) + if (family == AF_INET6) return &inet6_stream_ops; #endif - WARN_ON_ONCE(sk->sk_prot != &tcp_prot); + WARN_ON_ONCE(family != AF_INET); return &inet_stream_ops; }
Hi Jiayuan,
On 23/10/2025 14:54, Jiayuan Chen wrote:
When the server has MPTCP enabled but receives a non-MP-capable request from a client, it calls mptcp_fallback_tcp_ops().
Since non-MPTCP connections are allowed to use sockmap, which replaces sk->sk_prot, using sk->sk_prot to determine the IP version in mptcp_fallback_tcp_ops() becomes unreliable. This can lead to assigning incorrect ops to sk->sk_socket->ops.
Additionally, when BPF Sockmap modifies the protocol handlers, the original WARN_ON_ONCE(sk->sk_prot != &tcp_prot) check would falsely trigger warnings.
Fix this by using the more stable sk_family to distinguish between IPv4 and IPv6 connections, ensuring correct fallback protocol operations are selected even when BPF Sockmap has modified the socket protocol handlers.
Fixes: 0b4f33def7bb ("mptcp: fix tcp fallback crash") Cc: stable@vger.kernel.org Signed-off-by: Jiayuan Chen jiayuan.chen@linux.dev Reviewed-by: Jakub Sitnicki jakub@cloudflare.com
net/mptcp/protocol.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 0292162a14ee..2393741bc310 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -61,11 +61,16 @@ static u64 mptcp_wnd_end(const struct mptcp_sock *msk) static const struct proto_ops *mptcp_fallback_tcp_ops(const struct sock *sk) {
- /* When BPF sockmap is used, it may replace sk->sk_prot.
* Using sk_family is a reliable way to determine the IP version.*/- unsigned short family = READ_ONCE(sk->sk_family);
#if IS_ENABLED(CONFIG_MPTCP_IPV6)
- if (sk->sk_prot == &tcpv6_prot)
- if (family == AF_INET6) return &inet6_stream_ops;
#endif
- WARN_ON_ONCE(sk->sk_prot != &tcp_prot);
- WARN_ON_ONCE(family != AF_INET); return &inet_stream_ops;
Just to be sure: is there anything in BPF modifying sk->sk_socket->ops? Because that's what mptcp_fallback_tcp_ops() will do somehow.
In other words, is it always fine to set inet(6)_stream_ops? (I guess yes, but better to be sure while we are looking at that :) )
}
Cheers, Matt
October 23, 2025 at 22:10, "Matthieu Baerts" <matttbe@kernel.org mailto:matttbe@kernel.org?to=%22Matthieu%20Baerts%22%20%3Cmatttbe%40kernel.org%3E > wrote:
Hi Jiayuan,
On 23/10/2025 14:54, Jiayuan Chen wrote:
When the server has MPTCP enabled but receives a non-MP-capable request from a client, it calls mptcp_fallback_tcp_ops(). Since non-MPTCP connections are allowed to use sockmap, which replaces sk->sk_prot, using sk->sk_prot to determine the IP version in mptcp_fallback_tcp_ops() becomes unreliable. This can lead to assigning incorrect ops to sk->sk_socket->ops. Additionally, when BPF Sockmap modifies the protocol handlers, the original WARN_ON_ONCE(sk->sk_prot != &tcp_prot) check would falsely trigger warnings. Fix this by using the more stable sk_family to distinguish between IPv4 and IPv6 connections, ensuring correct fallback protocol operations are selected even when BPF Sockmap has modified the socket protocol handlers. Fixes: 0b4f33def7bb ("mptcp: fix tcp fallback crash") Cc: stable@vger.kernel.org Signed-off-by: Jiayuan Chen jiayuan.chen@linux.dev Reviewed-by: Jakub Sitnicki jakub@cloudflare.com
net/mptcp/protocol.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 0292162a14ee..2393741bc310 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -61,11 +61,16 @@ static u64 mptcp_wnd_end(const struct mptcp_sock *msk) static const struct proto_ops *mptcp_fallback_tcp_ops(const struct sock *sk) {
- /* When BPF sockmap is used, it may replace sk->sk_prot.
- Using sk_family is a reliable way to determine the IP version.
- */
- unsigned short family = READ_ONCE(sk->sk_family);
#if IS_ENABLED(CONFIG_MPTCP_IPV6)
- if (sk->sk_prot == &tcpv6_prot)
- if (family == AF_INET6)
return &inet6_stream_ops; #endif
- WARN_ON_ONCE(sk->sk_prot != &tcp_prot);
- WARN_ON_ONCE(family != AF_INET);
return &inet_stream_ops;
Just to be sure: is there anything in BPF modifying sk->sk_socket->ops? Because that's what mptcp_fallback_tcp_ops() will do somehow.
In other words, is it always fine to set inet(6)_stream_ops? (I guess yes, but better to be sure while we are looking at that :) )
Hi Matt,
I can confirm that on the BPF side, the only special operations targeting sockets currently are sockmap/sockhash. Their implementations do not modify sk->sk_socket->ops. Currently, they only modify sk->prot, because the BPF side typically operates on 'struct sock' and does not concern itself with 'struct socket'.
Therefore, setting inet(6)_stream_ops is fine.
Thanks, Jiayuan
}
Cheers, Matt -- Sponsored by the NGI0 Core fund.
On 10/23/25 2:54 PM, Jiayuan Chen wrote:
When the server has MPTCP enabled but receives a non-MP-capable request from a client, it calls mptcp_fallback_tcp_ops().
Since non-MPTCP connections are allowed to use sockmap, which replaces sk->sk_prot, using sk->sk_prot to determine the IP version in mptcp_fallback_tcp_ops() becomes unreliable. This can lead to assigning incorrect ops to sk->sk_socket->ops.
I don't see how sockmap could modify the to-be-accepted socket sk_prot before mptcp_fallback_tcp_ops(), as such call happens before the fd is installed, and AFAICS sockmap can only fetch sockets via fds.
Is this patch needed?
/P
On 10/28/25 12:30 PM, Paolo Abeni wrote:
On 10/23/25 2:54 PM, Jiayuan Chen wrote:
When the server has MPTCP enabled but receives a non-MP-capable request from a client, it calls mptcp_fallback_tcp_ops().
Since non-MPTCP connections are allowed to use sockmap, which replaces sk->sk_prot, using sk->sk_prot to determine the IP version in mptcp_fallback_tcp_ops() becomes unreliable. This can lead to assigning incorrect ops to sk->sk_socket->ops.
I don't see how sockmap could modify the to-be-accepted socket sk_prot before mptcp_fallback_tcp_ops(), as such call happens before the fd is installed, and AFAICS sockmap can only fetch sockets via fds.
Is this patch needed?
Matttbe explained off-list the details of how that could happen. I think the commit message here must be more verbose to explain clearly the whys, even to those non proficient in sockmap like me.
Thanks,
Paolo
linux-stable-mirror@lists.linaro.org
-
Jiayuan Chen -
Matthieu Baerts -
Paolo Abeni