ocfs2/dlm: Clean up messages in o2dlm
authorSunil Mushran <sunil.mushran@oracle.com>
Sun, 24 Jul 2011 17:23:54 +0000 (10:23 -0700)
committerSunil Mushran <sunil.mushran@oracle.com>
Sun, 24 Jul 2011 17:23:54 +0000 (10:23 -0700)
o2dlm messages needed a facelift.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
fs/ocfs2/dlm/dlmdomain.c
fs/ocfs2/dlm/dlmlock.c
fs/ocfs2/dlm/dlmmaster.c
fs/ocfs2/dlm/dlmrecovery.c
fs/ocfs2/dlm/dlmthread.c

index 6ed6b95dcf935a6516e935b85a3ca9ffc0b8a9d8..ce225fb0cea71069d44c4dcc2dc87be207fd1140 100644 (file)
@@ -539,17 +539,17 @@ again:
 
 static void __dlm_print_nodes(struct dlm_ctxt *dlm)
 {
-       int node = -1;
+       int node = -1, num = 0;
 
        assert_spin_locked(&dlm->spinlock);
 
-       printk(KERN_NOTICE "o2dlm: Nodes in domain %s: ", dlm->name);
-
+       printk("( ");
        while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES,
                                     node + 1)) < O2NM_MAX_NODES) {
                printk("%d ", node);
+               ++num;
        }
-       printk("\n");
+       printk(") %u nodes\n", num);
 }
 
 static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
@@ -566,11 +566,10 @@ static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
 
        node = exit_msg->node_idx;
 
-       printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s\n", node, dlm->name);
-
        spin_lock(&dlm->spinlock);
        clear_bit(node, dlm->domain_map);
        clear_bit(node, dlm->exit_domain_map);
+       printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s ", node, dlm->name);
        __dlm_print_nodes(dlm);
 
        /* notify anything attached to the heartbeat events */
@@ -755,6 +754,7 @@ void dlm_unregister_domain(struct dlm_ctxt *dlm)
 
                dlm_mark_domain_leaving(dlm);
                dlm_leave_domain(dlm);
+               printk(KERN_NOTICE "o2dlm: Leaving domain %s\n", dlm->name);
                dlm_force_free_mles(dlm);
                dlm_complete_dlm_shutdown(dlm);
        }
@@ -970,7 +970,7 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
                clear_bit(assert->node_idx, dlm->exit_domain_map);
                __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
 
-               printk(KERN_NOTICE "o2dlm: Node %u joins domain %s\n",
+               printk(KERN_NOTICE "o2dlm: Node %u joins domain %s ",
                       assert->node_idx, dlm->name);
                __dlm_print_nodes(dlm);
 
@@ -1701,8 +1701,10 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm)
 bail:
        spin_lock(&dlm->spinlock);
        __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
-       if (!status)
+       if (!status) {
+               printk(KERN_NOTICE "o2dlm: Joining domain %s ", dlm->name);
                __dlm_print_nodes(dlm);
+       }
        spin_unlock(&dlm->spinlock);
 
        if (ctxt) {
index 8d39e0fd66f7379b8fb08ac40502efa0d4a51cc6..c7f3e22bda1ea35c75ac780e725c35aef9015fe4 100644 (file)
@@ -319,27 +319,23 @@ static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
        tmpret = o2net_send_message(DLM_CREATE_LOCK_MSG, dlm->key, &create,
                                    sizeof(create), res->owner, &status);
        if (tmpret >= 0) {
-               // successfully sent and received
-               ret = status;  // this is already a dlm_status
+               ret = status;
                if (ret == DLM_REJECTED) {
-                       mlog(ML_ERROR, "%s:%.*s: BUG.  this is a stale lockres "
-                            "no longer owned by %u.  that node is coming back "
-                            "up currently.\n", dlm->name, create.namelen,
+                       mlog(ML_ERROR, "%s: res %.*s, Stale lockres no longer "
+                            "owned by node %u. That node is coming back up "
+                            "currently.\n", dlm->name, create.namelen,
                             create.name, res->owner);
                        dlm_print_one_lock_resource(res);
                        BUG();
                }
        } else {
-               mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
-                    "node %u\n", tmpret, DLM_CREATE_LOCK_MSG, dlm->key,
-                    res->owner);
-               if (dlm_is_host_down(tmpret)) {
+               mlog(ML_ERROR, "%s: res %.*s, Error %d send CREATE LOCK to "
+                    "node %u\n", dlm->name, create.namelen, create.name,
+                    tmpret, res->owner);
+               if (dlm_is_host_down(tmpret))
                        ret = DLM_RECOVERING;
-                       mlog(0, "node %u died so returning DLM_RECOVERING "
-                            "from lock message!\n", res->owner);
-               } else {
+               else
                        ret = dlm_err_to_dlm_status(tmpret);
-               }
        }
 
        return ret;
index 11eefb8c12e98fb418f41c31a3b0a32201be3ca1..9f3b093b629ac7047a7da96f506e9035b1090aa4 100644 (file)
@@ -829,8 +829,8 @@ lookup:
                 * but they might own this lockres.  wait on them. */
                bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
                if (bit < O2NM_MAX_NODES) {
-                       mlog(ML_NOTICE, "%s:%.*s: at least one node (%d) to "
-                            "recover before lock mastery can begin\n",
+                       mlog(0, "%s: res %.*s, At least one node (%d) "
+                            "to recover before lock mastery can begin\n",
                             dlm->name, namelen, (char *)lockid, bit);
                        wait_on_recovery = 1;
                }
@@ -864,8 +864,8 @@ redo_request:
                 * dlm spinlock would be detectable be a change on the mle,
                 * so we only need to clear out the recovery map once. */
                if (dlm_is_recovery_lock(lockid, namelen)) {
-                       mlog(ML_NOTICE, "%s: recovery map is not empty, but "
-                            "must master $RECOVERY lock now\n", dlm->name);
+                       mlog(0, "%s: Recovery map is not empty, but must "
+                            "master $RECOVERY lock now\n", dlm->name);
                        if (!dlm_pre_master_reco_lockres(dlm, res))
                                wait_on_recovery = 0;
                        else {
@@ -883,8 +883,8 @@ redo_request:
                spin_lock(&dlm->spinlock);
                bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
                if (bit < O2NM_MAX_NODES) {
-                       mlog(ML_NOTICE, "%s:%.*s: at least one node (%d) to "
-                            "recover before lock mastery can begin\n",
+                       mlog(0, "%s: res %.*s, At least one node (%d) "
+                            "to recover before lock mastery can begin\n",
                             dlm->name, namelen, (char *)lockid, bit);
                        wait_on_recovery = 1;
                } else
@@ -913,8 +913,8 @@ redo_request:
                         * yet, keep going until it does.  this is how the
                         * master will know that asserts are needed back to
                         * the lower nodes. */
-                       mlog(0, "%s:%.*s: requests only up to %u but master "
-                            "is %u, keep going\n", dlm->name, namelen,
+                       mlog(0, "%s: res %.*s, Requests only up to %u but "
+                            "master is %u, keep going\n", dlm->name, namelen,
                             lockid, nodenum, mle->master);
                }
        }
@@ -924,13 +924,12 @@ wait:
        ret = dlm_wait_for_lock_mastery(dlm, res, mle, &blocked);
        if (ret < 0) {
                wait_on_recovery = 1;
-               mlog(0, "%s:%.*s: node map changed, redo the "
-                    "master request now, blocked=%d\n",
-                    dlm->name, res->lockname.len,
+               mlog(0, "%s: res %.*s, Node map changed, redo the master "
+                    "request now, blocked=%d\n", dlm->name, res->lockname.len,
                     res->lockname.name, blocked);
                if (++tries > 20) {
-                       mlog(ML_ERROR, "%s:%.*s: spinning on "
-                            "dlm_wait_for_lock_mastery, blocked=%d\n",
+                       mlog(ML_ERROR, "%s: res %.*s, Spinning on "
+                            "dlm_wait_for_lock_mastery, blocked = %d\n",
                             dlm->name, res->lockname.len,
                             res->lockname.name, blocked);
                        dlm_print_one_lock_resource(res);
@@ -940,7 +939,8 @@ wait:
                goto redo_request;
        }
 
-       mlog(0, "lockres mastered by %u\n", res->owner);
+       mlog(0, "%s: res %.*s, Mastered by %u\n", dlm->name, res->lockname.len,
+            res->lockname.name, res->owner);
        /* make sure we never continue without this */
        BUG_ON(res->owner == O2NM_MAX_NODES);
 
@@ -2187,8 +2187,6 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
        namelen = res->lockname.len;
        BUG_ON(namelen > O2NM_MAX_NAME_LEN);
 
-       mlog(0, "%s:%.*s: sending deref to %d\n",
-            dlm->name, namelen, lockname, res->owner);
        memset(&deref, 0, sizeof(deref));
        deref.node_idx = dlm->node_num;
        deref.namelen = namelen;
@@ -2197,14 +2195,12 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
        ret = o2net_send_message(DLM_DEREF_LOCKRES_MSG, dlm->key,
                                 &deref, sizeof(deref), res->owner, &r);
        if (ret < 0)
-               mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
-                    "node %u\n", ret, DLM_DEREF_LOCKRES_MSG, dlm->key,
-                    res->owner);
+               mlog(ML_ERROR, "%s: res %.*s, error %d send DEREF to node %u\n",
+                    dlm->name, namelen, lockname, ret, res->owner);
        else if (r < 0) {
                /* BAD.  other node says I did not have a ref. */
-               mlog(ML_ERROR,"while dropping ref on %s:%.*s "
-                   "(master=%u) got %d.\n", dlm->name, namelen,
-                   lockname, res->owner, r);
+               mlog(ML_ERROR, "%s: res %.*s, DEREF to node %u got %d\n",
+                    dlm->name, namelen, lockname, res->owner, r);
                dlm_print_one_lock_resource(res);
                BUG();
        }
@@ -2916,9 +2912,9 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
                                         &migrate, sizeof(migrate), nodenum,
                                         &status);
                if (ret < 0) {
-                       mlog(ML_ERROR, "Error %d when sending message %u (key "
-                            "0x%x) to node %u\n", ret, DLM_MIGRATE_REQUEST_MSG,
-                            dlm->key, nodenum);
+                       mlog(ML_ERROR, "%s: res %.*s, Error %d send "
+                            "MIGRATE_REQUEST to node %u\n", dlm->name,
+                            migrate.namelen, migrate.name, ret, nodenum);
                        if (!dlm_is_host_down(ret)) {
                                mlog(ML_ERROR, "unhandled error=%d!\n", ret);
                                BUG();
index 7efab6d28a21b4ee6a8376559d70f739a4e1da90..a3c312c43b90e12ec1394859f8d803dae018152f 100644 (file)
@@ -430,6 +430,8 @@ 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 +442,18 @@ 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 +516,8 @@ 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 +528,13 @@ 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 +575,7 @@ 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 +648,7 @@ 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 +810,9 @@ 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 +963,9 @@ 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 +1134,11 @@ 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;
@@ -2324,9 +2333,9 @@ 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,
index 1d6d1d22c4715e3c89bef69570916cc5bf44c259..46faac20f16c98323ac611a44626526ac8bbce03 100644 (file)
@@ -185,8 +185,6 @@ static void dlm_purge_lockres(struct dlm_ctxt *dlm,
                /* clear our bit from the master's refmap, ignore errors */
                ret = dlm_drop_lockres_ref(dlm, res);
                if (ret < 0) {
-                       mlog(ML_ERROR, "%s: deref %.*s failed %d\n", dlm->name,
-                            res->lockname.len, res->lockname.name, ret);
                        if (!dlm_is_host_down(ret))
                                BUG();
                }