Merge branch 'tipc-failover-fixes'

Jon Maloy says:

====================
tipc: fix link failover/synch problems

We fix three problems with the new link failover/synch implementation,
which was introduced earlier in this release cycle. They are all related
to situations where there is a very short interval between the disabling
and enabling of interfaces.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
David S. Miller 2015-08-23 16:14:46 -07:00
commit c5f98b5669
2 changed files with 24 additions and 8 deletions

View File

@ -351,11 +351,11 @@ int tipc_link_fsm_evt(struct tipc_link *l, int evt)
l->state = LINK_RESET;
break;
case LINK_ESTABLISH_EVT:
case LINK_SYNCH_END_EVT:
break;
case LINK_SYNCH_BEGIN_EVT:
l->state = LINK_SYNCHING;
break;
case LINK_SYNCH_END_EVT:
case LINK_FAILOVER_BEGIN_EVT:
case LINK_FAILOVER_END_EVT:
default:
@ -1330,6 +1330,7 @@ static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
u16 peers_snd_nxt = msg_next_sent(hdr);
u16 peers_tol = msg_link_tolerance(hdr);
u16 peers_prio = msg_linkprio(hdr);
u16 rcv_nxt = l->rcv_nxt;
char *if_name;
int rc = 0;
@ -1393,7 +1394,7 @@ static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
break;
/* Send NACK if peer has sent pkts we haven't received yet */
if (more(peers_snd_nxt, l->rcv_nxt))
if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
rcvgap = peers_snd_nxt - l->rcv_nxt;
if (rcvgap || (msg_probe(hdr)))
tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,

View File

@ -423,6 +423,8 @@ static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
/* There is still a working link => initiate failover */
tnl = node_active_link(n, 0);
tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
tipc_link_reset(l);
@ -565,6 +567,8 @@ void tipc_node_check_dest(struct net *net, u32 onode,
goto exit;
}
tipc_link_reset(l);
if (n->state == NODE_FAILINGOVER)
tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
le->link = l;
n->link_cnt++;
tipc_node_calculate_timer(n, l);
@ -1075,7 +1079,7 @@ static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
u16 exp_pkts = msg_msgcnt(hdr);
u16 rcv_nxt, syncpt, dlv_nxt;
int state = n->state;
struct tipc_link *l, *pl = NULL;
struct tipc_link *l, *tnl, *pl = NULL;
struct tipc_media_addr *maddr;
int i, pb_id;
@ -1129,7 +1133,7 @@ static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
}
/* Open parallel link when tunnel link reaches synch point */
if ((n->state == NODE_FAILINGOVER) && !tipc_link_is_failingover(l)) {
if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
if (!more(rcv_nxt, n->sync_point))
return true;
tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
@ -1138,6 +1142,10 @@ static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
return true;
}
/* No synching needed if only one link */
if (!pl || !tipc_link_is_up(pl))
return true;
/* Initiate or update synch mode if applicable */
if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG)) {
syncpt = iseqno + exp_pkts - 1;
@ -1156,13 +1164,20 @@ static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
/* Open tunnel link when parallel link reaches synch point */
if ((n->state == NODE_SYNCHING) && tipc_link_is_synching(l)) {
if (pl)
dlv_nxt = mod(pl->rcv_nxt - skb_queue_len(pl->inputq));
if (!pl || more(dlv_nxt, n->sync_point)) {
tipc_link_fsm_evt(l, LINK_SYNCH_END_EVT);
if (tipc_link_is_synching(l)) {
tnl = l;
} else {
tnl = pl;
pl = l;
}
dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
if (more(dlv_nxt, n->sync_point)) {
tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
return true;
}
if (l == pl)
return true;
if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
return true;
if (usr == LINK_PROTOCOL)