From: Tariq Toukan <tariqt@mellanox.com>
Date: Mon, 22 Jun 2020 23:26:04 +0300
Subject: net: Do not clear the sock TX queue in sk_set_socket()
Git-commit: 41b14fb8724d5a4b382a63cb4a1a61880347ccb8
Patch-mainline: 5.8-rc3
References: networking-stable-20_06_28

Clearing the sock TX queue in sk_set_socket() might cause unexpected
out-of-order transmit when called from sock_orphan(), as outstanding
packets can pick a different TX queue and bypass the ones already queued.

This is undesired in general. More specifically, it breaks the in-order
scheduling property guarantee for device-offloaded TLS sockets.

Remove the call to sk_tx_queue_clear() in sk_set_socket(), and add it
explicitly only where needed.

Fixes: e022f0b4a03f ("net: Introduce sk_tx_queue_mapping")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
---
 include/net/sock.h |    1 -
 net/core/sock.c    |    2 ++
 2 files changed, 2 insertions(+), 1 deletion(-)

--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1751,7 +1751,6 @@ static inline int sk_rx_queue_get(const
 
 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
 {
-	sk_tx_queue_clear(sk);
 	sk->sk_socket = sock;
 }
 
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1457,6 +1457,7 @@ struct sock *sk_alloc(struct net *net, i
 		cgroup_sk_alloc(&sk->sk_cgrp_data);
 		sock_update_classid(&sk->sk_cgrp_data);
 		sock_update_netprioidx(&sk->sk_cgrp_data);
+		sk_tx_queue_clear(sk);
 	}
 
 	return sk;
@@ -1658,6 +1659,7 @@ struct sock *sk_clone_lock(const struct
 		 */
 		sk_refcnt_debug_inc(newsk);
 		sk_set_socket(newsk, NULL);
+		sk_tx_queue_clear(newsk);
 		newsk->sk_wq = NULL;
 
 		if (newsk->sk_prot->sockets_allocated)