Signed-off-by: Sunil Mushran <[email protected]>
---
fs/ocfs2/cluster/ocfs2_nodemanager.h | 2 +
fs/ocfs2/dlm/dlmcommon.h | 12 ++-
fs/ocfs2/dlm/dlmdomain.c | 222 +++++++++++++++++++++++++++++++++-
3 files changed, 234 insertions(+), 2 deletions(-)
diff --git a/fs/ocfs2/cluster/ocfs2_nodemanager.h
b/fs/ocfs2/cluster/ocfs2_nodemanager.h
index 5b9854b..1829c01 100644
--- a/fs/ocfs2/cluster/ocfs2_nodemanager.h
+++ b/fs/ocfs2/cluster/ocfs2_nodemanager.h
@@ -36,4 +36,6 @@
/* host name, group name, cluster name all 64 bytes */
#define O2NM_MAX_NAME_LEN 64 // __NEW_UTS_LEN
+#define O2NM_MAX_HBREGIONS 16
+
#endif /* _OCFS2_NODEMANAGER_H */
diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index a13292a..2c05138 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -445,7 +445,8 @@ enum {
DLM_LOCK_REQUEST_MSG, /* 515 */
DLM_RECO_DATA_DONE_MSG, /* 516 */
DLM_BEGIN_RECO_MSG, /* 517 */
- DLM_FINALIZE_RECO_MSG /* 518 */
+ DLM_FINALIZE_RECO_MSG, /* 518 */
+ DLM_QUERY_HBREGION, /* 519 */
};
struct dlm_reco_node_data
@@ -727,6 +728,15 @@ struct dlm_cancel_join
u8 domain[O2NM_MAX_NAME_LEN];
};
+struct dlm_query_hbregion {
+ u8 qhb_node;
+ u8 qhb_numregions;
+ u8 qhb_namelen;
+ u8 pad1;
+ u8 qhb_domain[O2NM_MAX_NAME_LEN];
+ u8 qhb_hbregions[O2HB_MAX_REGION_NAME_LEN * O2NM_MAX_HBREGIONS];
+};
+
struct dlm_exit_domain
{
u8 node_idx;
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 2408b9f..3521a00 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -128,10 +128,13 @@ static DECLARE_WAIT_QUEUE_HEAD(dlm_domain_events);
* will have a negotiated version with the same major number and a minor
* number equal or smaller. The dlm_ctxt->dlm_locking_proto field should
* be used to determine what a running domain is actually using.
+ *
+ * New in version 1.1:
+ * - Message DLM_QUERY_HBREGION added to support global heartbeat
*/
static const struct dlm_protocol_version dlm_protocol = {
.pv_major = 1,
- .pv_minor = 0,
+ .pv_minor = 1,
};
#define DLM_DOMAIN_BACKOFF_MS 200
@@ -142,6 +145,8 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg,
u32 len, void *data,
void **ret_data);
static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data,
void **ret_data);
+static int dlm_query_hbregion_handler(struct o2net_msg *msg, u32 len,
+ void *data, void **ret_data);
static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
void **ret_data);
static int dlm_protocol_compare(struct dlm_protocol_version *existing,
@@ -918,6 +923,205 @@ static int dlm_assert_joined_handler(struct o2net_msg
*msg, u32 len, void *data,
return 0;
}
+static int dlm_match_hbregions(struct dlm_ctxt *dlm,
+ struct dlm_query_hbregion *qhb)
+{
+ char *local = NULL, *remote = qhb->qhb_hbregions;
+ char *l, *r;
+ int localnr, i, j, foundit;
+ int status = 0;
+
+ if (!o2hb_global_heartbeat_active()) {
+ if (qhb->qhb_numregions) {
+ mlog(ML_ERROR, "Domain %s: Joining node %d has global "
+ "heartbeat enabled but local node %d does not\n",
+ qhb->qhb_domain, qhb->qhb_node, dlm->node_num);
+ status = -EINVAL;
+ }
+ goto bail;
+ }
+
+ if (o2hb_global_heartbeat_active() && !qhb->qhb_numregions) {
+ mlog(ML_ERROR, "Domain %s: Local node %d has global "
+ "heartbeat enabled but joining node %d does not\n",
+ qhb->qhb_domain, dlm->node_num, qhb->qhb_node);
+ status = -EINVAL;
+ goto bail;
+ }
+
+ r = remote;
+ for (i = 0; i < qhb->qhb_numregions; ++i) {
+ mlog(ML_NOTICE, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, r);
+ r += O2HB_MAX_REGION_NAME_LEN;
+ }
+
+ local = kmalloc(sizeof(qhb->qhb_hbregions), GFP_KERNEL);
+ if (!local) {
+ status = -ENOMEM;
+ goto bail;
+ }
+
+ localnr = o2hb_get_all_regions(local, O2NM_MAX_HBREGIONS);
+
+ /* compare local regions with remote */
+ l = local;
+ for (i = 0; i < localnr; ++i) {
+ foundit = 0;
+ r = remote;
+ for (j = 0; j <= qhb->qhb_numregions; ++j) {
+ if (!memcmp(l, r, O2HB_MAX_REGION_NAME_LEN)) {
+ foundit = 1;
+ break;
+ }
+ r += O2HB_MAX_REGION_NAME_LEN;
+ }
+ if (!foundit) {
+ status = -EINVAL;
+ mlog(ML_ERROR, "Domain %s: Region '%.*s' registered "
+ "in local node %d but not in joining node %d\n",
+ qhb->qhb_domain, O2HB_MAX_REGION_NAME_LEN, l,
+ dlm->node_num, qhb->qhb_node);
+ goto bail;
+ }
+ l += O2HB_MAX_REGION_NAME_LEN;
+ }
+
+ /* compare remote with local regions */
+ r = remote;
+ for (i = 0; i < qhb->qhb_numregions; ++i) {
+ foundit = 0;
+ l = local;
+ for (j = 0; j < localnr; ++j) {
+ if (!memcmp(r, l, O2HB_MAX_REGION_NAME_LEN)) {
+ foundit = 1;
+ break;
+ }
+ l += O2HB_MAX_REGION_NAME_LEN;
+ }
+ if (!foundit) {
+ status = -EINVAL;
+ mlog(ML_ERROR, "Domain %s: Region '%.*s' registered "
+ "in joining node %d but not in local node %d\n",
+ qhb->qhb_domain, O2HB_MAX_REGION_NAME_LEN, r,
+ qhb->qhb_node, dlm->node_num);
+ goto bail;
+ }
+ r += O2HB_MAX_REGION_NAME_LEN;
+ }
+
+bail:
+ kfree(local);
+
+ return status;
+}
+
+static int dlm_send_hbregions(struct dlm_ctxt *dlm, unsigned long *node_map)
+{
+ struct dlm_query_hbregion *qhb = NULL;
+ int status, ret = 0, i;
+ char *p;
+
+ if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES)
+ goto bail;
+
+ qhb = kmalloc(sizeof(struct dlm_query_hbregion), GFP_KERNEL);
+ if (!qhb) {
+ ret = -ENOMEM;
+ mlog_errno(ret);
+ goto bail;
+ }
+
+ memset(qhb, 0, sizeof(struct dlm_query_hbregion));
+
+ qhb->qhb_node = dlm->node_num;
+ qhb->qhb_namelen = strlen(dlm->name);
+ memcpy(qhb->qhb_domain, dlm->name, qhb->qhb_namelen);
+ /* if local hb, the numregions will be zero */
+ if (o2hb_global_heartbeat_active())
+ qhb->qhb_numregions = o2hb_get_all_regions(qhb->qhb_hbregions,
+ O2NM_MAX_HBREGIONS);
+
+ p = qhb->qhb_hbregions;
+ for (i = 0; i < qhb->qhb_numregions; ++i, p += O2HB_MAX_REGION_NAME_LEN)
+ mlog(ML_NOTICE, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, p);
+
+ i = -1;
+ while ((i = find_next_bit(node_map, O2NM_MAX_NODES,
+ i + 1)) < O2NM_MAX_NODES) {
+ if (i == dlm->node_num)
+ continue;
+
+ mlog(ML_NOTICE, "Sending hbregion to node %d\n", i);
+
+ ret = o2net_send_message(DLM_QUERY_HBREGION, DLM_MOD_KEY, qhb,
+ sizeof(struct dlm_query_hbregion),
+ i, &status);
+ if (ret >= 0)
+ ret = status;
+ if (ret) {
+ mlog(ML_ERROR, "hbregion mismatch %d, node %d\n",
+ ret, i);
+ break;
+ }
+ }
+
+bail:
+ kfree(qhb);
+ return ret;
+}
+
+static int dlm_query_hbregion_handler(struct o2net_msg *msg, u32 len,
+ void *data, void **ret_data)
+{
+ struct dlm_query_hbregion *qhb;
+ struct dlm_ctxt *dlm = NULL;
+ int status = 0;
+ int locked = 0;
+
+ qhb = (struct dlm_query_hbregion *) msg->buf;
+
+ mlog(ML_NOTICE, "Node %u queries hb regions on domain %s\n",
+ qhb->qhb_node, qhb->qhb_domain);
+
+ status = -EINVAL;
+
+ spin_lock(&dlm_domain_lock);
+ dlm = __dlm_lookup_domain_full(qhb->qhb_domain, qhb->qhb_namelen);
+ if (!dlm) {
+ mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
+ "before join domain\n", qhb->qhb_node, qhb->qhb_domain);
+ goto bail;
+ }
+
+ spin_lock(&dlm->spinlock);
+ locked = 1;
+ if (dlm->joining_node != qhb->qhb_node) {
+ mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
+ "but joining node is %d\n", qhb->qhb_node, qhb->qhb_domain,
+ dlm->joining_node);
+ goto bail;
+ }
+
+ /* Support for global heartbeat was added in 1.1 */
+ if (dlm->dlm_locking_proto.pv_major == 1 &&
+ dlm->dlm_locking_proto.pv_minor == 0) {
+ mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
+ "but active dlm protocol is %d.%d\n", qhb->qhb_node,
+ qhb->qhb_domain, dlm->dlm_locking_proto.pv_major,
+ dlm->dlm_locking_proto.pv_minor);
+ goto bail;
+ }
+
+ status = dlm_match_hbregions(dlm, qhb);
+
+bail:
+ if (locked)
+ spin_unlock(&dlm->spinlock);
+ spin_unlock(&dlm_domain_lock);
+
+ return status;
+}
+
static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data,
void **ret_data)
{
@@ -1233,6 +1437,15 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm)
set_bit(dlm->node_num, dlm->domain_map);
spin_unlock(&dlm->spinlock);
+ /* Support for global heartbeat was added in 1.1 */
+ if (dlm_protocol.pv_major >= 1 && dlm_protocol.pv_minor > 0) {
+ status = dlm_send_hbregions(dlm, ctxt->yes_resp_map);
+ if (status) {
+ mlog_errno(status);
+ goto bail;
+ }
+ }
+
dlm_send_join_asserts(dlm, ctxt->yes_resp_map);
/* Joined state *must* be set before the joining node
@@ -1799,6 +2012,13 @@ static int dlm_register_net_handlers(void)
sizeof(struct dlm_cancel_join),
dlm_cancel_join_handler,
NULL, NULL, &dlm_join_handlers);
+ if (status)
+ goto bail;
+
+ status = o2net_register_handler(DLM_QUERY_HBREGION, DLM_MOD_KEY,
+ sizeof(struct dlm_query_hbregion),
+ dlm_query_hbregion_handler,
+ NULL, NULL, &dlm_join_handlers);
bail:
if (status < 0)
--
1.7.0.4
_______________________________________________
Ocfs2-devel mailing list
[email protected]
http://oss.oracle.com/mailman/listinfo/ocfs2-devel
