12  * that a node is broken and should be recovered.  They can't both recover each
17  * So we declare that a node which has given up on connecting to a majority
20  * There are huge opportunities for races here.  After we give up on a node's
22  * to declare the node as truly dead.  We also need to be careful with the
23  * race between when we see a node start heartbeating and when we connect
83  * go away as our node would be fenced externally before other nodes
115 			mlog(ML_ERROR, "fencing this node because it is "  in o2quo_make_decision()
126 		 * the lowest numbered node */  in o2quo_make_decision()
129 			mlog(ML_ERROR, "fencing this node because it is "  in o2quo_make_decision()
138 			mlog(ML_ERROR, "fencing this node because it is "  in o2quo_make_decision()
141 			     "node %u\n", quorum, qs->qs_heartbeating,  in o2quo_make_decision()
152 		mlog(ML_NOTICE, "not fencing this node, heartbeating: %d, "  in o2quo_make_decision()
162 static void o2quo_set_hold(struct o2quo_state *qs, u8 node)  in o2quo_set_hold()  argument
166 	if (!test_and_set_bit(node, qs->qs_hold_bm)) {  in o2quo_set_hold()
169 			        "node %u\n", node);  in o2quo_set_hold()
170 		mlog(0, "node %u, %d total\n", node, qs->qs_holds);  in o2quo_set_hold()
174 static void o2quo_clear_hold(struct o2quo_state *qs, u8 node)  in o2quo_clear_hold()  argument
178 	if (test_and_clear_bit(node, qs->qs_hold_bm)) {  in o2quo_clear_hold()
179 		mlog(0, "node %u, %d total\n", node, qs->qs_holds - 1);  in o2quo_clear_hold()
186 		mlog_bug_on_msg(qs->qs_holds < 0, "node %u, holds %d\n",  in o2quo_clear_hold()
187 				node, qs->qs_holds);  in o2quo_clear_hold()
191 /* as a node comes up we delay the quorum decision until we know the fate of
195 void o2quo_hb_up(u8 node)  in o2quo_hb_up()  argument
203 		        "node %u\n", node);  in o2quo_hb_up()
204 	mlog_bug_on_msg(test_bit(node, qs->qs_hb_bm), "node %u\n", node);  in o2quo_hb_up()
205 	set_bit(node, qs->qs_hb_bm);  in o2quo_hb_up()
207 	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);  in o2quo_hb_up()
209 	if (!test_bit(node, qs->qs_conn_bm))  in o2quo_hb_up()
210 		o2quo_set_hold(qs, node);  in o2quo_hb_up()
212 		o2quo_clear_hold(qs, node);  in o2quo_hb_up()
217 /* hb going down releases any holds we might have had due to this node from
219 void o2quo_hb_down(u8 node)  in o2quo_hb_down()  argument
227 			"node %u, %d heartbeating\n",  in o2quo_hb_down()
228 			node, qs->qs_heartbeating);  in o2quo_hb_down()
229 	mlog_bug_on_msg(!test_bit(node, qs->qs_hb_bm), "node %u\n", node);  in o2quo_hb_down()
230 	clear_bit(node, qs->qs_hb_bm);  in o2quo_hb_down()
232 	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);  in o2quo_hb_down()
234 	o2quo_clear_hold(qs, node);  in o2quo_hb_down()
239 /* this tells us that we've decided that the node is still heartbeating
244 void o2quo_hb_still_up(u8 node)  in o2quo_hb_still_up()  argument
250 	mlog(0, "node %u\n", node);  in o2quo_hb_still_up()
253 	o2quo_clear_hold(qs, node);  in o2quo_hb_still_up()
258 /* This is analogous to hb_up.  as a node's connection comes up we delay the
263 void o2quo_conn_up(u8 node)  in o2quo_conn_up()  argument
271 		        "node %u\n", node);  in o2quo_conn_up()
272 	mlog_bug_on_msg(test_bit(node, qs->qs_conn_bm), "node %u\n", node);  in o2quo_conn_up()
273 	set_bit(node, qs->qs_conn_bm);  in o2quo_conn_up()
275 	mlog(0, "node %u, %d total\n", node, qs->qs_connected);  in o2quo_conn_up()
277 	if (!test_bit(node, qs->qs_hb_bm))  in o2quo_conn_up()
278 		o2quo_set_hold(qs, node);  in o2quo_conn_up()
280 		o2quo_clear_hold(qs, node);  in o2quo_conn_up()
285 /* we've decided that we won't ever be connecting to the node again.  if it's
287  * node stops heartbeating from hb_down or the caller decides that the node is
289 void o2quo_conn_err(u8 node)  in o2quo_conn_err()  argument
295 	if (test_bit(node, qs->qs_conn_bm)) {  in o2quo_conn_err()
298 				"node %u, connected %d\n",  in o2quo_conn_err()
299 				node, qs->qs_connected);  in o2quo_conn_err()
301 		clear_bit(node, qs->qs_conn_bm);  in o2quo_conn_err()
303 		if (test_bit(node, qs->qs_hb_bm))  in o2quo_conn_err()
304 			o2quo_set_hold(qs, node);  in o2quo_conn_err()
307 	mlog(0, "node %u, %d total\n", node, qs->qs_connected);  in o2quo_conn_err()