/*
* Check if this node is heartbeating and is connected to all other
* heartbeating nodes.
*/
static int o2cb_cluster_check(void)
{
u8 node_num;
int i;
unsigned long hbmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
unsigned long netmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
node_num = o2nm_this_node();
if (node_num == O2NM_MAX_NODES) {
printk(KERN_ERR "o2cb: This node has not been configured.\n");
return -EINVAL;
}
/*
* o2dlm expects o2net sockets to be created. If not, then
* dlm_join_domain() fails with a stack of errors which are both cryptic
* and incomplete. The idea here is to detect upfront whether we have
* managed to connect to all nodes or not. If not, then list the nodes
* to allow the user to check the configuration (incorrect IP, firewall,
* etc.) Yes, this is racy. But its not the end of the world.
*/
#define O2CB_MAP_STABILIZE_COUNT 60
for (i = 0; i < O2CB_MAP_STABILIZE_COUNT; ++i) {
o2hb_fill_node_map(hbmap, sizeof(hbmap));
if (!test_bit(node_num, hbmap)) {
printk(KERN_ERR "o2cb: %s heartbeat has not been "
"started.\n", (o2hb_global_heartbeat_active() ?
"Global" : "Local"));
return -EINVAL;
}
o2net_fill_node_map(netmap, sizeof(netmap));
/* Force set the current node to allow easy compare */
set_bit(node_num, netmap);
if (!memcmp(hbmap, netmap, sizeof(hbmap)))
return 0;
if (i < O2CB_MAP_STABILIZE_COUNT)
msleep(1000);
}
printk(KERN_ERR "o2cb: This node could not connect to nodes:");
i = -1;
while ((i = find_next_bit(hbmap, O2NM_MAX_NODES,
i + 1)) < O2NM_MAX_NODES) {
if (!test_bit(i, netmap))
printk(" %u", i);
}
printk(".\n");
return -ENOTCONN;
}
static int o2cb_cluster_this_node(unsigned int *node)
{
int node_num;
node_num = o2nm_this_node();
if (node_num == O2NM_INVALID_NODE_NUM)
return -ENOENT;
if (node_num >= O2NM_MAX_NODES)
return -EOVERFLOW;
*node = node_num;
return 0;
}
static void o2quo_make_decision(struct work_struct *work)
{
int quorum;
int lowest_hb, lowest_reachable = 0, fence = 0;
struct o2quo_state *qs = &o2quo_state;
spin_lock(&qs->qs_lock);
lowest_hb = find_first_bit(qs->qs_hb_bm, O2NM_MAX_NODES);
if (lowest_hb != O2NM_MAX_NODES)
lowest_reachable = test_bit(lowest_hb, qs->qs_conn_bm);
mlog(0, "heartbeating: %d, connected: %d, "
"lowest: %d (%sreachable)\n", qs->qs_heartbeating,
qs->qs_connected, lowest_hb, lowest_reachable ? "" : "un");
if (!test_bit(o2nm_this_node(), qs->qs_hb_bm) ||
qs->qs_heartbeating == 1)
goto out;
if (qs->qs_heartbeating & 1) {
/* the odd numbered cluster case is straight forward --
* if we can't talk to the majority we're hosed */
quorum = (qs->qs_heartbeating + 1)/2;
if (qs->qs_connected < quorum) {
mlog(ML_ERROR, "fencing this node because it is "
"only connected to %u nodes and %u is needed "
"to make a quorum out of %u heartbeating nodes\n",
qs->qs_connected, quorum,
qs->qs_heartbeating);
fence = 1;
}
} else {
/* the even numbered cluster adds the possibility of each half
* of the cluster being able to talk amongst themselves.. in
* that case we're hosed if we can't talk to the group that has
* the lowest numbered node */
quorum = qs->qs_heartbeating / 2;
if (qs->qs_connected < quorum) {
mlog(ML_ERROR, "fencing this node because it is "
"only connected to %u nodes and %u is needed "
"to make a quorum out of %u heartbeating nodes\n",
qs->qs_connected, quorum,
qs->qs_heartbeating);
fence = 1;
}
else if ((qs->qs_connected == quorum) &&
!lowest_reachable) {
mlog(ML_ERROR, "fencing this node because it is "
"connected to a half-quorum of %u out of %u "
"nodes which doesn't include the lowest active "
"node %u\n", quorum, qs->qs_heartbeating,
lowest_hb);
fence = 1;
}
}
out:
spin_unlock(&qs->qs_lock);
if (fence)
o2quo_fence_self();
}
