After subflow lock is dropped, more wmem might have been made available. This fixes a deadlock in mptcp_connect.sh 'mmap' mode: wmem is exhausted. But as the mptcp socket holds on to already-acked data (for retransmit) no wakeup will occur. Using 'goto restart' calls mptcp_clean_una(sk) which will free pages that have been acked completely in the mean time. Fixes: fb529e62d3f3 ("mptcp: break and restart in case mptcp sndbuf is full") Signed-off-by: Florian Westphal <fw(a)strlen.de&gt; --- I suspect Paolos recent sndbuf tracking changes made this bug trigger more frequently. Afaics net-next is just more lucky so it might make sense to pass this via net-next. Let me know. net/mptcp/protocol.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index bd02c568c4b9..ad91f4588216 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -1178,7 +1178,6 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) goto out; } -wait_for_sndbuf: __mptcp_flush_join_list(msk); ssk = mptcp_subflow_get_send(msk, &sndbuf); while (!sk_stream_memory_free(sk) || @@ -1282,7 +1281,7 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) */ mptcp_set_timeout(sk, ssk); release_sock(ssk); - goto wait_for_sndbuf; + goto restart; } } } -- 2.26.2