aboutsummaryrefslogtreecommitdiff
path: root/fs/ocfs2/dlm/dlmrecovery.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ocfs2/dlm/dlmrecovery.c')
-rw-r--r--fs/ocfs2/dlm/dlmrecovery.c164
1 files changed, 82 insertions, 82 deletions
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index 7efab6d28a2..01ebfd0bdad 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -362,40 +362,38 @@ static int dlm_is_node_recovered(struct dlm_ctxt *dlm, u8 node)
}
-int dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout)
+void dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout)
{
- if (timeout) {
- mlog(ML_NOTICE, "%s: waiting %dms for notification of "
- "death of node %u\n", dlm->name, timeout, node);
+ if (dlm_is_node_dead(dlm, node))
+ return;
+
+ printk(KERN_NOTICE "o2dlm: Waiting on the death of node %u in "
+ "domain %s\n", node, dlm->name);
+
+ if (timeout)
wait_event_timeout(dlm->dlm_reco_thread_wq,
- dlm_is_node_dead(dlm, node),
- msecs_to_jiffies(timeout));
- } else {
- mlog(ML_NOTICE, "%s: waiting indefinitely for notification "
- "of death of node %u\n", dlm->name, node);
+ dlm_is_node_dead(dlm, node),
+ msecs_to_jiffies(timeout));
+ else
wait_event(dlm->dlm_reco_thread_wq,
dlm_is_node_dead(dlm, node));
- }
- /* for now, return 0 */
- return 0;
}
-int dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout)
+void dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout)
{
- if (timeout) {
- mlog(0, "%s: waiting %dms for notification of "
- "recovery of node %u\n", dlm->name, timeout, node);
+ if (dlm_is_node_recovered(dlm, node))
+ return;
+
+ printk(KERN_NOTICE "o2dlm: Waiting on the recovery of node %u in "
+ "domain %s\n", node, dlm->name);
+
+ if (timeout)
wait_event_timeout(dlm->dlm_reco_thread_wq,
- dlm_is_node_recovered(dlm, node),
- msecs_to_jiffies(timeout));
- } else {
- mlog(0, "%s: waiting indefinitely for notification "
- "of recovery of node %u\n", dlm->name, node);
+ dlm_is_node_recovered(dlm, node),
+ msecs_to_jiffies(timeout));
+ else
wait_event(dlm->dlm_reco_thread_wq,
dlm_is_node_recovered(dlm, node));
- }
- /* for now, return 0 */
- return 0;
}
/* callers of the top-level api calls (dlmlock/dlmunlock) should
@@ -430,6 +428,8 @@ static void dlm_begin_recovery(struct dlm_ctxt *dlm)
{
spin_lock(&dlm->spinlock);
BUG_ON(dlm->reco.state & DLM_RECO_STATE_ACTIVE);
+ printk(KERN_NOTICE "o2dlm: Begin recovery on domain %s for node %u\n",
+ dlm->name, dlm->reco.dead_node);
dlm->reco.state |= DLM_RECO_STATE_ACTIVE;
spin_unlock(&dlm->spinlock);
}
@@ -440,9 +440,18 @@ static void dlm_end_recovery(struct dlm_ctxt *dlm)
BUG_ON(!(dlm->reco.state & DLM_RECO_STATE_ACTIVE));
dlm->reco.state &= ~DLM_RECO_STATE_ACTIVE;
spin_unlock(&dlm->spinlock);
+ printk(KERN_NOTICE "o2dlm: End recovery on domain %s\n", dlm->name);
wake_up(&dlm->reco.event);
}
+static void dlm_print_recovery_master(struct dlm_ctxt *dlm)
+{
+ printk(KERN_NOTICE "o2dlm: Node %u (%s) is the Recovery Master for the "
+ "dead node %u in domain %s\n", dlm->reco.new_master,
+ (dlm->node_num == dlm->reco.new_master ? "me" : "he"),
+ dlm->reco.dead_node, dlm->name);
+}
+
static int dlm_do_recovery(struct dlm_ctxt *dlm)
{
int status = 0;
@@ -505,9 +514,8 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm)
}
mlog(0, "another node will master this recovery session.\n");
}
- mlog(0, "dlm=%s (%d), new_master=%u, this node=%u, dead_node=%u\n",
- dlm->name, task_pid_nr(dlm->dlm_reco_thread_task), dlm->reco.new_master,
- dlm->node_num, dlm->reco.dead_node);
+
+ dlm_print_recovery_master(dlm);
/* it is safe to start everything back up here
* because all of the dead node's lock resources
@@ -518,15 +526,13 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm)
return 0;
master_here:
- mlog(ML_NOTICE, "(%d) Node %u is the Recovery Master for the Dead Node "
- "%u for Domain %s\n", task_pid_nr(dlm->dlm_reco_thread_task),
- dlm->node_num, dlm->reco.dead_node, dlm->name);
+ dlm_print_recovery_master(dlm);
status = dlm_remaster_locks(dlm, dlm->reco.dead_node);
if (status < 0) {
/* we should never hit this anymore */
- mlog(ML_ERROR, "error %d remastering locks for node %u, "
- "retrying.\n", status, dlm->reco.dead_node);
+ mlog(ML_ERROR, "%s: Error %d remastering locks for node %u, "
+ "retrying.\n", dlm->name, status, dlm->reco.dead_node);
/* yield a bit to allow any final network messages
* to get handled on remaining nodes */
msleep(100);
@@ -567,7 +573,7 @@ static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node)
BUG_ON(ndata->state != DLM_RECO_NODE_DATA_INIT);
ndata->state = DLM_RECO_NODE_DATA_REQUESTING;
- mlog(0, "requesting lock info from node %u\n",
+ mlog(0, "%s: Requesting lock info from node %u\n", dlm->name,
ndata->node_num);
if (ndata->node_num == dlm->node_num) {
@@ -640,7 +646,7 @@ static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node)
spin_unlock(&dlm_reco_state_lock);
}
- mlog(0, "done requesting all lock info\n");
+ mlog(0, "%s: Done requesting all lock info\n", dlm->name);
/* nodes should be sending reco data now
* just need to wait */
@@ -802,10 +808,9 @@ static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from,
/* negative status is handled by caller */
if (ret < 0)
- mlog(ML_ERROR, "Error %d when sending message %u (key "
- "0x%x) to node %u\n", ret, DLM_LOCK_REQUEST_MSG,
- dlm->key, request_from);
-
+ mlog(ML_ERROR, "%s: Error %d send LOCK_REQUEST to node %u "
+ "to recover dead node %u\n", dlm->name, ret,
+ request_from, dead_node);
// return from here, then
// sleep until all received or error
return ret;
@@ -956,9 +961,9 @@ static int dlm_send_all_done_msg(struct dlm_ctxt *dlm, u8 dead_node, u8 send_to)
ret = o2net_send_message(DLM_RECO_DATA_DONE_MSG, dlm->key, &done_msg,
sizeof(done_msg), send_to, &tmpret);
if (ret < 0) {
- mlog(ML_ERROR, "Error %d when sending message %u (key "
- "0x%x) to node %u\n", ret, DLM_RECO_DATA_DONE_MSG,
- dlm->key, send_to);
+ mlog(ML_ERROR, "%s: Error %d send RECO_DATA_DONE to node %u "
+ "to recover dead node %u\n", dlm->name, ret, send_to,
+ dead_node);
if (!dlm_is_host_down(ret)) {
BUG();
}
@@ -1127,9 +1132,11 @@ static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm,
if (ret < 0) {
/* XXX: negative status is not handled.
* this will end up killing this node. */
- mlog(ML_ERROR, "Error %d when sending message %u (key "
- "0x%x) to node %u\n", ret, DLM_MIG_LOCKRES_MSG,
- dlm->key, send_to);
+ mlog(ML_ERROR, "%s: res %.*s, Error %d send MIG_LOCKRES to "
+ "node %u (%s)\n", dlm->name, mres->lockname_len,
+ mres->lockname, ret, send_to,
+ (orig_flags & DLM_MRES_MIGRATION ?
+ "migration" : "recovery"));
} else {
/* might get an -ENOMEM back here */
ret = status;
@@ -1767,7 +1774,7 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm,
dlm->name, mres->lockname_len, mres->lockname,
from);
spin_lock(&res->spinlock);
- dlm_lockres_set_refmap_bit(from, res);
+ dlm_lockres_set_refmap_bit(dlm, res, from);
spin_unlock(&res->spinlock);
added++;
break;
@@ -1965,7 +1972,7 @@ skip_lvb:
mlog(0, "%s:%.*s: added lock for node %u, "
"setting refmap bit\n", dlm->name,
res->lockname.len, res->lockname.name, ml->node);
- dlm_lockres_set_refmap_bit(ml->node, res);
+ dlm_lockres_set_refmap_bit(dlm, res, ml->node);
added++;
}
spin_unlock(&res->spinlock);
@@ -2084,6 +2091,9 @@ static void dlm_finish_local_lockres_recovery(struct dlm_ctxt *dlm,
list_for_each_entry_safe(res, next, &dlm->reco.resources, recovering) {
if (res->owner == dead_node) {
+ mlog(0, "%s: res %.*s, Changing owner from %u to %u\n",
+ dlm->name, res->lockname.len, res->lockname.name,
+ res->owner, new_master);
list_del_init(&res->recovering);
spin_lock(&res->spinlock);
/* new_master has our reference from
@@ -2105,40 +2115,30 @@ static void dlm_finish_local_lockres_recovery(struct dlm_ctxt *dlm,
for (i = 0; i < DLM_HASH_BUCKETS; i++) {
bucket = dlm_lockres_hash(dlm, i);
hlist_for_each_entry(res, hash_iter, bucket, hash_node) {
- if (res->state & DLM_LOCK_RES_RECOVERING) {
- if (res->owner == dead_node) {
- mlog(0, "(this=%u) res %.*s owner=%u "
- "was not on recovering list, but "
- "clearing state anyway\n",
- dlm->node_num, res->lockname.len,
- res->lockname.name, new_master);
- } else if (res->owner == dlm->node_num) {
- mlog(0, "(this=%u) res %.*s owner=%u "
- "was not on recovering list, "
- "owner is THIS node, clearing\n",
- dlm->node_num, res->lockname.len,
- res->lockname.name, new_master);
- } else
- continue;
+ if (!(res->state & DLM_LOCK_RES_RECOVERING))
+ 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);
- dlm_lockres_put(res);
- }
- spin_lock(&res->spinlock);
- /* new_master has our reference from
- * the lock state sent during recovery */
- dlm_change_lockres_owner(dlm, res, new_master);
- res->state &= ~DLM_LOCK_RES_RECOVERING;
- if (__dlm_lockres_has_locks(res))
- __dlm_dirty_lockres(dlm, res);
- spin_unlock(&res->spinlock);
- wake_up(&res->wq);
+ if (res->owner != dead_node &&
+ res->owner != dlm->node_num)
+ continue;
+
+ if (!list_empty(&res->recovering)) {
+ list_del_init(&res->recovering);
+ dlm_lockres_put(res);
}
+
+ /* new_master has our reference from
+ * the lock state sent during recovery */
+ mlog(0, "%s: res %.*s, Changing owner from %u to %u\n",
+ dlm->name, res->lockname.len, res->lockname.name,
+ res->owner, new_master);
+ spin_lock(&res->spinlock);
+ dlm_change_lockres_owner(dlm, res, new_master);
+ res->state &= ~DLM_LOCK_RES_RECOVERING;
+ if (__dlm_lockres_has_locks(res))
+ __dlm_dirty_lockres(dlm, res);
+ spin_unlock(&res->spinlock);
+ wake_up(&res->wq);
}
}
}
@@ -2252,12 +2252,12 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm,
res->lockname.len, res->lockname.name, freed, dead_node);
__dlm_print_one_lock_resource(res);
}
- dlm_lockres_clear_refmap_bit(dead_node, res);
+ dlm_lockres_clear_refmap_bit(dlm, res, dead_node);
} else if (test_bit(dead_node, res->refmap)) {
mlog(0, "%s:%.*s: dead node %u had a ref, but had "
"no locks and had not purged before dying\n", dlm->name,
res->lockname.len, res->lockname.name, dead_node);
- dlm_lockres_clear_refmap_bit(dead_node, res);
+ dlm_lockres_clear_refmap_bit(dlm, res, dead_node);
}
/* do not kick thread yet */
@@ -2324,9 +2324,9 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node)
dlm_revalidate_lvb(dlm, res, dead_node);
if (res->owner == dead_node) {
if (res->state & DLM_LOCK_RES_DROPPING_REF) {
- mlog(ML_NOTICE, "Ignore %.*s for "
+ mlog(ML_NOTICE, "%s: res %.*s, Skip "
"recovery as it is being freed\n",
- res->lockname.len,
+ dlm->name, res->lockname.len,
res->lockname.name);
} else
dlm_move_lockres_to_recovery_list(dlm,