This is a note to let you know that I've just added the patch titled tcp: do not release socket ownership in tcp_close() to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary The filename of the patch is: tcp-do-not-release-socket-ownership-in-tcp_close.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable@xxxxxxxxxxxxxxx> know about it. >From 8873c064d1de579ea23412a6d3eee972593f142b Mon Sep 17 00:00:00 2001 From: Eric Dumazet <edumazet@xxxxxxxxxx> Date: Mon, 1 Oct 2018 23:24:26 -0700 Subject: tcp: do not release socket ownership in tcp_close() From: Eric Dumazet <edumazet@xxxxxxxxxx> commit 8873c064d1de579ea23412a6d3eee972593f142b upstream. syzkaller was able to hit the WARN_ON(sock_owned_by_user(sk)); in tcp_close() While a socket is being closed, it is very possible other threads find it in rtnetlink dump. tcp_get_info() will acquire the socket lock for a short amount of time (slow = lock_sock_fast(sk)/unlock_sock_fast(sk, slow);), enough to trigger the warning. Fixes: 67db3e4bfbc9 ("tcp: no longer hold ehash lock while calling tcp_get_info()") Signed-off-by: Eric Dumazet <edumazet@xxxxxxxxxx> Reported-by: syzbot <syzkaller@xxxxxxxxxxxxxxxx> Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- include/net/sock.h | 1 + net/core/sock.c | 2 +- net/ipv4/tcp.c | 11 +++-------- 3 files changed, 5 insertions(+), 9 deletions(-) --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1452,6 +1452,7 @@ static inline void lock_sock(struct sock lock_sock_nested(sk, 0); } +void __release_sock(struct sock *sk); void release_sock(struct sock *sk); /* BH context may only use the following locking interface. */ --- a/net/core/sock.c +++ b/net/core/sock.c @@ -2242,7 +2242,7 @@ static void __lock_sock(struct sock *sk) finish_wait(&sk->sk_lock.wq, &wait); } -static void __release_sock(struct sock *sk) +void __release_sock(struct sock *sk) __releases(&sk->sk_lock.slock) __acquires(&sk->sk_lock.slock) { --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2217,16 +2217,10 @@ adjudge_to_death: sock_hold(sk); sock_orphan(sk); - /* It is the last release_sock in its life. It will remove backlog. */ - release_sock(sk); - - - /* Now socket is owned by kernel and we acquire BH lock - * to finish close. No need to check for user refs. - */ local_bh_disable(); bh_lock_sock(sk); - WARN_ON(sock_owned_by_user(sk)); + /* remove backlog if any, without releasing ownership. */ + __release_sock(sk); percpu_counter_inc(sk->sk_prot->orphan_count); @@ -2295,6 +2289,7 @@ adjudge_to_death: out: bh_unlock_sock(sk); local_bh_enable(); + release_sock(sk); sock_put(sk); } EXPORT_SYMBOL(tcp_close); Patches currently in stable-queue which might be from edumazet@xxxxxxxxxx are queue-4.14/llc-do-not-use-sk_eat_skb.patch queue-4.14/tcp-do-not-release-socket-ownership-in-tcp_close.patch