static int dlm_recovery_thread(void *data);
void dlm_complete_recovery_thread(struct dlm_ctxt *dlm);
int dlm_launch_recovery_thread(struct dlm_ctxt *dlm);
-static void dlm_kick_recovery_thread(struct dlm_ctxt *dlm);
+void dlm_kick_recovery_thread(struct dlm_ctxt *dlm);
static int dlm_do_recovery(struct dlm_ctxt *dlm);
static int dlm_pick_recovery_master(struct dlm_ctxt *dlm);
u8 send_to,
struct dlm_lock_resource *res,
int total_locks);
-static int dlm_lockres_master_requery(struct dlm_ctxt *dlm,
- struct dlm_lock_resource *res,
- u8 *real_master);
static int dlm_process_recovery_data(struct dlm_ctxt *dlm,
struct dlm_lock_resource *res,
struct dlm_migratable_lockres *mres);
-static int dlm_do_master_requery(struct dlm_ctxt *dlm,
- struct dlm_lock_resource *res,
- u8 nodenum, u8 *real_master);
static int dlm_send_finalize_reco_message(struct dlm_ctxt *dlm);
static int dlm_send_all_done_msg(struct dlm_ctxt *dlm,
u8 dead_node, u8 send_to);
* RECOVERY THREAD
*/
-static void dlm_kick_recovery_thread(struct dlm_ctxt *dlm)
+void dlm_kick_recovery_thread(struct dlm_ctxt *dlm)
{
/* wake the recovery thread
* this will wake the reco thread in one of three places
dlm->name, dlm->reco.dead_node, dlm->reco.new_master,
dead_node, reco_master);
mlog(ML_ERROR, "%s: name=%.*s master=%u locks=%u/%u flags=%u "
- "entry[0]={c=%"MLFu64",l=%u,f=%u,t=%d,ct=%d,hb=%d,n=%u}\n",
+ "entry[0]={c=%u:%llu,l=%u,f=%u,t=%d,ct=%d,hb=%d,n=%u}\n",
dlm->name, mres->lockname_len, mres->lockname, mres->master,
mres->num_locks, mres->total_locks, mres->flags,
- mres->ml[0].cookie, mres->ml[0].list, mres->ml[0].flags,
+ dlm_get_lock_cookie_node(mres->ml[0].cookie),
+ dlm_get_lock_cookie_seq(mres->ml[0].cookie),
+ mres->ml[0].list, mres->ml[0].flags,
mres->ml[0].type, mres->ml[0].convert_type,
mres->ml[0].highest_blocked, mres->ml[0].node);
BUG();
mlog(0, "found lockres owned by dead node while "
"doing recovery for node %u. sending it.\n",
dead_node);
- list_del_init(&res->recovering);
- list_add_tail(&res->recovering, list);
+ list_move_tail(&res->recovering, list);
} else if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) {
mlog(0, "found UNKNOWN owner while doing recovery "
"for node %u. sending it.\n", dead_node);
- list_del_init(&res->recovering);
- list_add_tail(&res->recovering, list);
+ list_move_tail(&res->recovering, list);
}
}
spin_unlock(&dlm->spinlock);
-static int dlm_lockres_master_requery(struct dlm_ctxt *dlm,
- struct dlm_lock_resource *res,
- u8 *real_master)
+int dlm_lockres_master_requery(struct dlm_ctxt *dlm,
+ struct dlm_lock_resource *res, u8 *real_master)
{
struct dlm_node_iter iter;
int nodenum;
ret = dlm_do_master_requery(dlm, res, nodenum, real_master);
if (ret < 0) {
mlog_errno(ret);
- BUG();
- /* TODO: need to figure a way to restart this */
+ if (!dlm_is_host_down(ret))
+ BUG();
+ /* host is down, so answer for that node would be
+ * DLM_LOCK_RES_OWNER_UNKNOWN. continue. */
}
if (*real_master != DLM_LOCK_RES_OWNER_UNKNOWN) {
mlog(0, "lock master is %u\n", *real_master);
}
-static int dlm_do_master_requery(struct dlm_ctxt *dlm,
- struct dlm_lock_resource *res,
- u8 nodenum, u8 *real_master)
+int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+ u8 nodenum, u8 *real_master)
{
int ret = -EINVAL;
struct dlm_master_requery req;
struct dlm_ctxt *dlm = data;
struct dlm_master_requery *req = (struct dlm_master_requery *)msg->buf;
struct dlm_lock_resource *res = NULL;
+ unsigned int hash;
int master = DLM_LOCK_RES_OWNER_UNKNOWN;
u32 flags = DLM_ASSERT_MASTER_REQUERY;
return master;
}
+ hash = dlm_lockid_hash(req->name, req->namelen);
+
spin_lock(&dlm->spinlock);
- res = __dlm_lookup_lockres(dlm, req->name, req->namelen);
+ res = __dlm_lookup_lockres(dlm, req->name, req->namelen, hash);
if (res) {
spin_lock(&res->spinlock);
master = res->owner;
/* lock is always created locally first, and
* destroyed locally last. it must be on the list */
if (!lock) {
+ u64 c = ml->cookie;
mlog(ML_ERROR, "could not find local lock "
- "with cookie %"MLFu64"!\n",
- ml->cookie);
+ "with cookie %u:%llu!\n",
+ dlm_get_lock_cookie_node(c),
+ dlm_get_lock_cookie_seq(c));
BUG();
}
BUG_ON(lock->ml.node != ml->node);
/* move the lock to its proper place */
/* do not alter lock refcount. switching lists. */
- list_del_init(&lock->list);
- list_add_tail(&lock->list, queue);
+ list_move_tail(&lock->list, queue);
spin_unlock(&res->spinlock);
mlog(0, "just reordered a local lock!\n");
} else
continue;
+ if (!list_empty(&res->recovering)) {
+ mlog(0, "%s:%.*s: lockres was "
+ "marked RECOVERING, owner=%u\n",
+ dlm->name, res->lockname.len,
+ res->lockname.name, res->owner);
+ list_del_init(&res->recovering);
+ }
spin_lock(&res->spinlock);
dlm_change_lockres_owner(dlm, res, new_master);
res->state &= ~DLM_LOCK_RES_RECOVERING;
mlog(0, "%u not in domain/live_nodes map "
"so setting it in reco map manually\n",
br->dead_node);
- set_bit(br->dead_node, dlm->recovery_map);
+ /* force the recovery cleanup in __dlm_hb_node_down
+ * both of these will be cleared in a moment */
+ set_bit(br->dead_node, dlm->domain_map);
+ set_bit(br->dead_node, dlm->live_nodes_map);
__dlm_hb_node_down(dlm, br->dead_node);
}
spin_unlock(&dlm->spinlock);