tipc: fix unexpected link reset due to discovery messages

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2160540
Tested: compile only

commit c244c09
Author: Tung Nguyen <tung.q.nguyen@dektech.com.au>
Date:   Thu Jan 5 06:02:51 2023 +0000

    tipc: fix unexpected link reset due to discovery messages

    This unexpected behavior is observed:

    node 1                    | node 2
    ------                    | ------
    link is established       | link is established
    reboot                    | link is reset
    up                        | send discovery message
    receive discovery message |
    link is established       | link is established
    send discovery message    |
                              | receive discovery message
                              | link is reset (unexpected)
                              | send reset message
    link is reset             |

    It is due to delayed re-discovery as described in function
    tipc_node_check_dest(): "this link endpoint has already reset
    and re-established contact with the peer, before receiving a
    discovery message from that node."

    However, commit 598411d has changed the condition for calling
    tipc_node_link_down() which was the acceptance of new media address.

    This commit fixes this by restoring the old and correct behavior.

    Fixes: 598411d ("tipc: make resetting of links non-atomic")
    Acked-by: Jon Maloy <jmaloy@redhat.com>
    Signed-off-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
    Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Xin Long <lxin@redhat.com>

Author: Xin Long

net/tipc/node.c

@@ -1179,8 +1179,9 @@ void tipc_node_check_dest(struct net *net, u32 addr,
 	bool addr_match = false;
 	bool sign_match = false;
 	bool link_up = false;
+	bool link_is_reset = false;
 	bool accept_addr = false;
-	bool reset = true;
+	bool reset = false;
 	char *if_name;
 	unsigned long intv;
 	u16 session;
@@ -1200,14 +1201,14 @@ void tipc_node_check_dest(struct net *net, u32 addr,
 	/* Prepare to validate requesting node's signature and media address */
 	l = le->link;
 	link_up = l && tipc_link_is_up(l);
+	link_is_reset = l && tipc_link_is_reset(l);
 	addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
 	sign_match = (signature == n->signature);
 
 	/* These three flags give us eight permutations: */
 
 	if (sign_match && addr_match && link_up) {
-		/* All is fine. Do nothing. */
-		reset = false;
+		/* All is fine. Ignore requests. */
 		/* Peer node is not a container/local namespace */
 		if (!n->peer_hash_mix)
 			n->peer_hash_mix = hash_mixes;
@@ -1232,6 +1233,7 @@ void tipc_node_check_dest(struct net *net, u32 addr,
 		 */
 		accept_addr = true;
 		*respond = true;
+		reset = true;
 	} else if (!sign_match && addr_match && link_up) {
 		/* Peer node rebooted. Two possibilities:
 		 *  - Delayed re-discovery; this link endpoint has already
@@ -1263,6 +1265,7 @@ void tipc_node_check_dest(struct net *net, u32 addr,
 		n->signature = signature;
 		accept_addr = true;
 		*respond = true;
+		reset = true;
 	}
 
 	if (!accept_addr)
@@ -1291,6 +1294,7 @@ void tipc_node_check_dest(struct net *net, u32 addr,
 		tipc_link_fsm_evt(l, LINK_RESET_EVT);
 		if (n->state == NODE_FAILINGOVER)
 			tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
+		link_is_reset = tipc_link_is_reset(l);
 		le->link = l;
 		n->link_cnt++;
 		tipc_node_calculate_timer(n, l);
@@ -1303,7 +1307,7 @@ void tipc_node_check_dest(struct net *net, u32 addr,
 	memcpy(&le->maddr, maddr, sizeof(*maddr));
 exit:
 	tipc_node_write_unlock(n);
-	if (reset && l && !tipc_link_is_reset(l))
+	if (reset && !link_is_reset)
 		tipc_node_link_down(n, b->identity, false);
 	tipc_node_put(n);
 }