ocfs2: o2hb: add NEGOTIATE_APPROVE message
authorJunxiao Bi <junxiao.bi@oracle.com>
Fri, 27 May 2016 21:27:04 +0000 (14:27 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 27 May 2016 21:49:37 +0000 (14:49 -0700)
This message is used to re-queue write timeout timer and negotiate timer
when all nodes suffer a write hung to storage, this makes node not fence
self if storage down.

Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com>
Reviewed-by: Ryan Ding <ryan.ding@oracle.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Gang He <ghe@suse.com>
Cc: rwxybh <rwxybh@126.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/cluster/heartbeat.c

index 454c8907683328ce9b48e3dc381ea199654d8963..e929b15d6162041c9e45f7794d2ed5ef9390efa7 100644 (file)
@@ -294,6 +294,7 @@ struct o2hb_bio_wait_ctxt {
 
 enum {
        O2HB_NEGO_TIMEOUT_MSG = 1,
+       O2HB_NEGO_APPROVE_MSG = 2,
 };
 
 struct o2hb_nego_msg {
@@ -388,7 +389,7 @@ again:
 static void o2hb_nego_timeout(struct work_struct *work)
 {
        unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
-       int master_node;
+       int master_node, i;
        struct o2hb_region *reg;
 
        reg = container_of(work, struct o2hb_region, hr_nego_timeout_work.work);
@@ -410,6 +411,17 @@ static void o2hb_nego_timeout(struct work_struct *work)
                }
 
                /* approve negotiate timeout request. */
+               o2hb_arm_timeout(reg);
+
+               i = -1;
+               while ((i = find_next_bit(live_node_bitmap,
+                               O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
+                       if (i == master_node)
+                               continue;
+
+                       o2hb_send_nego_msg(reg->hr_key,
+                                       O2HB_NEGO_APPROVE_MSG, i);
+               }
        } else {
                /* negotiate timeout with master node. */
                o2hb_send_nego_msg(reg->hr_key, O2HB_NEGO_TIMEOUT_MSG,
@@ -432,6 +444,13 @@ static int o2hb_nego_timeout_handler(struct o2net_msg *msg, u32 len, void *data,
        return 0;
 }
 
+static int o2hb_nego_approve_handler(struct o2net_msg *msg, u32 len, void *data,
+                               void **ret_data)
+{
+       o2hb_arm_timeout(data);
+       return 0;
+}
+
 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc)
 {
        atomic_set(&wc->wc_num_reqs, 1);
@@ -2099,6 +2118,13 @@ static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *g
        if (ret)
                goto free;
 
+       ret = o2net_register_handler(O2HB_NEGO_APPROVE_MSG, reg->hr_key,
+                       sizeof(struct o2hb_nego_msg),
+                       o2hb_nego_approve_handler,
+                       reg, NULL, &reg->hr_handler_list);
+       if (ret)
+               goto unregister_handler;
+
        ret = o2hb_debug_region_init(reg, o2hb_debug_dir);
        if (ret) {
                config_item_put(&reg->hr_item);