14  * that a node is broken and should be recovered.  They can't both recover each
19  * So we declare that a node which has given up on connecting to a majority
22  * There are huge opportunities for races here.  After we give up on a node's
24  * to declare the node as truly dead.  We also need to be careful with the
25  * race between when we see a node start heartbeating and when we connect
85  * go away as our node would be fenced externally before other nodes
117 			mlog(ML_ERROR, "fencing this node because it is "  in o2quo_make_decision()
128 		 * the lowest numbered node */  in o2quo_make_decision()
131 			mlog(ML_ERROR, "fencing this node because it is "  in o2quo_make_decision()
140 			mlog(ML_ERROR, "fencing this node because it is "  in o2quo_make_decision()
143 			     "node %u\n", quorum, qs->qs_heartbeating,  in o2quo_make_decision()
154 		mlog(ML_NOTICE, "not fencing this node, heartbeating: %d, "  in o2quo_make_decision()
164 static void o2quo_set_hold(struct o2quo_state *qs, u8 node)  in o2quo_set_hold()  argument
168 	if (!test_and_set_bit(node, qs->qs_hold_bm)) {  in o2quo_set_hold()
171 			        "node %u\n", node);  in o2quo_set_hold()
172 		mlog(0, "node %u, %d total\n", node, qs->qs_holds);  in o2quo_set_hold()
176 static void o2quo_clear_hold(struct o2quo_state *qs, u8 node)  in o2quo_clear_hold()  argument
180 	if (test_and_clear_bit(node, qs->qs_hold_bm)) {  in o2quo_clear_hold()
181 		mlog(0, "node %u, %d total\n", node, qs->qs_holds - 1);  in o2quo_clear_hold()
188 		mlog_bug_on_msg(qs->qs_holds < 0, "node %u, holds %d\n",  in o2quo_clear_hold()
189 				node, qs->qs_holds);  in o2quo_clear_hold()
193 /* as a node comes up we delay the quorum decision until we know the fate of
197 void o2quo_hb_up(u8 node)  in o2quo_hb_up()  argument
205 		        "node %u\n", node);  in o2quo_hb_up()
206 	mlog_bug_on_msg(test_bit(node, qs->qs_hb_bm), "node %u\n", node);  in o2quo_hb_up()
207 	set_bit(node, qs->qs_hb_bm);  in o2quo_hb_up()
209 	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);  in o2quo_hb_up()
211 	if (!test_bit(node, qs->qs_conn_bm))  in o2quo_hb_up()
212 		o2quo_set_hold(qs, node);  in o2quo_hb_up()
214 		o2quo_clear_hold(qs, node);  in o2quo_hb_up()
219 /* hb going down releases any holds we might have had due to this node from
221 void o2quo_hb_down(u8 node)  in o2quo_hb_down()  argument
229 			"node %u, %d heartbeating\n",  in o2quo_hb_down()
230 			node, qs->qs_heartbeating);  in o2quo_hb_down()
231 	mlog_bug_on_msg(!test_bit(node, qs->qs_hb_bm), "node %u\n", node);  in o2quo_hb_down()
232 	clear_bit(node, qs->qs_hb_bm);  in o2quo_hb_down()
234 	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);  in o2quo_hb_down()
236 	o2quo_clear_hold(qs, node);  in o2quo_hb_down()
241 /* this tells us that we've decided that the node is still heartbeating
246 void o2quo_hb_still_up(u8 node)  in o2quo_hb_still_up()  argument
252 	mlog(0, "node %u\n", node);  in o2quo_hb_still_up()
255 	o2quo_clear_hold(qs, node);  in o2quo_hb_still_up()
260 /* This is analogous to hb_up.  as a node's connection comes up we delay the
265 void o2quo_conn_up(u8 node)  in o2quo_conn_up()  argument
273 		        "node %u\n", node);  in o2quo_conn_up()
274 	mlog_bug_on_msg(test_bit(node, qs->qs_conn_bm), "node %u\n", node);  in o2quo_conn_up()
275 	set_bit(node, qs->qs_conn_bm);  in o2quo_conn_up()
277 	mlog(0, "node %u, %d total\n", node, qs->qs_connected);  in o2quo_conn_up()
279 	if (!test_bit(node, qs->qs_hb_bm))  in o2quo_conn_up()
280 		o2quo_set_hold(qs, node);  in o2quo_conn_up()
282 		o2quo_clear_hold(qs, node);  in o2quo_conn_up()
287 /* we've decided that we won't ever be connecting to the node again.  if it's
289  * node stops heartbeating from hb_down or the caller decides that the node is
291 void o2quo_conn_err(u8 node)  in o2quo_conn_err()  argument
297 	if (test_bit(node, qs->qs_conn_bm)) {  in o2quo_conn_err()
300 				"node %u, connected %d\n",  in o2quo_conn_err()
301 				node, qs->qs_connected);  in o2quo_conn_err()
303 		clear_bit(node, qs->qs_conn_bm);  in o2quo_conn_err()
305 		if (test_bit(node, qs->qs_hb_bm))  in o2quo_conn_err()
306 			o2quo_set_hold(qs, node);  in o2quo_conn_err()
309 	mlog(0, "node %u, %d total\n", node, qs->qs_connected);  in o2quo_conn_err()