summaryrefslogtreecommitdiffstats
path: root/fs/ocfs2/dlm/dlmdomain.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ocfs2/dlm/dlmdomain.c')
-rw-r--r--fs/ocfs2/dlm/dlmdomain.c95
1 files changed, 50 insertions, 45 deletions
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 8b3382abf840..02d315fef432 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -43,8 +43,6 @@
#include "dlmdomain.h"
#include "dlmdebug.h"
-#include "dlmver.h"
-
#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_DOMAIN)
#include "cluster/masklog.h"
@@ -841,7 +839,7 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
* to back off and try again. This gives heartbeat a chance
* to catch up.
*/
- if (!o2hb_check_node_heartbeating(query->node_idx)) {
+ if (!o2hb_check_node_heartbeating_no_sem(query->node_idx)) {
mlog(0, "node %u is not in our live map yet\n",
query->node_idx);
@@ -961,6 +959,14 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
* domain. Set him in the map and clean up our
* leftover join state. */
BUG_ON(dlm->joining_node != assert->node_idx);
+
+ if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) {
+ mlog(0, "dlm recovery is ongoing, disallow join\n");
+ spin_unlock(&dlm->spinlock);
+ spin_unlock(&dlm_domain_lock);
+ return -EAGAIN;
+ }
+
set_bit(assert->node_idx, dlm->domain_map);
clear_bit(assert->node_idx, dlm->exit_domain_map);
__dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
@@ -1125,7 +1131,6 @@ static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
struct dlm_ctxt *dlm = NULL;
char *local = NULL;
int status = 0;
- int locked = 0;
qr = (struct dlm_query_region *) msg->buf;
@@ -1134,10 +1139,8 @@ static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
/* buffer used in dlm_mast_regions() */
local = kmalloc(sizeof(qr->qr_regions), GFP_KERNEL);
- if (!local) {
- status = -ENOMEM;
- goto bail;
- }
+ if (!local)
+ return -ENOMEM;
status = -EINVAL;
@@ -1146,16 +1149,15 @@ static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
if (!dlm) {
mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
"before join domain\n", qr->qr_node, qr->qr_domain);
- goto bail;
+ goto out_domain_lock;
}
spin_lock(&dlm->spinlock);
- locked = 1;
if (dlm->joining_node != qr->qr_node) {
mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
"but joining node is %d\n", qr->qr_node, qr->qr_domain,
dlm->joining_node);
- goto bail;
+ goto out_dlm_lock;
}
/* Support for global heartbeat was added in 1.1 */
@@ -1165,14 +1167,15 @@ static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
"but active dlm protocol is %d.%d\n", qr->qr_node,
qr->qr_domain, dlm->dlm_locking_proto.pv_major,
dlm->dlm_locking_proto.pv_minor);
- goto bail;
+ goto out_dlm_lock;
}
status = dlm_match_regions(dlm, qr, local, sizeof(qr->qr_regions));
-bail:
- if (locked)
- spin_unlock(&dlm->spinlock);
+out_dlm_lock:
+ spin_unlock(&dlm->spinlock);
+
+out_domain_lock:
spin_unlock(&dlm_domain_lock);
kfree(local);
@@ -1522,6 +1525,7 @@ static int dlm_send_one_join_assert(struct dlm_ctxt *dlm,
unsigned int node)
{
int status;
+ int ret;
struct dlm_assert_joined assert_msg;
mlog(0, "Sending join assert to node %u\n", node);
@@ -1533,11 +1537,13 @@ static int dlm_send_one_join_assert(struct dlm_ctxt *dlm,
status = o2net_send_message(DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
&assert_msg, sizeof(assert_msg), node,
- NULL);
+ &ret);
if (status < 0)
mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
"node %u\n", status, DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
node);
+ else
+ status = ret;
return status;
}
@@ -1879,19 +1885,19 @@ static int dlm_join_domain(struct dlm_ctxt *dlm)
goto bail;
}
- status = dlm_debug_init(dlm);
+ status = dlm_launch_thread(dlm);
if (status < 0) {
mlog_errno(status);
goto bail;
}
- status = dlm_launch_thread(dlm);
+ status = dlm_launch_recovery_thread(dlm);
if (status < 0) {
mlog_errno(status);
goto bail;
}
- status = dlm_launch_recovery_thread(dlm);
+ status = dlm_debug_init(dlm);
if (status < 0) {
mlog_errno(status);
goto bail;
@@ -1917,12 +1923,11 @@ static int dlm_join_domain(struct dlm_ctxt *dlm)
goto bail;
}
- if (total_backoff >
- msecs_to_jiffies(DLM_JOIN_TIMEOUT_MSECS)) {
+ if (total_backoff > DLM_JOIN_TIMEOUT_MSECS) {
status = -ERESTARTSYS;
mlog(ML_NOTICE, "Timed out joining dlm domain "
"%s after %u msecs\n", dlm->name,
- jiffies_to_msecs(total_backoff));
+ total_backoff);
goto bail;
}
@@ -1970,24 +1975,22 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
dlm = kzalloc(sizeof(*dlm), GFP_KERNEL);
if (!dlm) {
- mlog_errno(-ENOMEM);
+ ret = -ENOMEM;
+ mlog_errno(ret);
goto leave;
}
dlm->name = kstrdup(domain, GFP_KERNEL);
if (dlm->name == NULL) {
- mlog_errno(-ENOMEM);
- kfree(dlm);
- dlm = NULL;
+ ret = -ENOMEM;
+ mlog_errno(ret);
goto leave;
}
dlm->lockres_hash = (struct hlist_head **)dlm_alloc_pagevec(DLM_HASH_PAGES);
if (!dlm->lockres_hash) {
- mlog_errno(-ENOMEM);
- kfree(dlm->name);
- kfree(dlm);
- dlm = NULL;
+ ret = -ENOMEM;
+ mlog_errno(ret);
goto leave;
}
@@ -1997,11 +2000,8 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
dlm->master_hash = (struct hlist_head **)
dlm_alloc_pagevec(DLM_HASH_PAGES);
if (!dlm->master_hash) {
- mlog_errno(-ENOMEM);
- dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES);
- kfree(dlm->name);
- kfree(dlm);
- dlm = NULL;
+ ret = -ENOMEM;
+ mlog_errno(ret);
goto leave;
}
@@ -2012,14 +2012,8 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
dlm->node_num = o2nm_this_node();
ret = dlm_create_debugfs_subroot(dlm);
- if (ret < 0) {
- dlm_free_pagevec((void **)dlm->master_hash, DLM_HASH_PAGES);
- dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES);
- kfree(dlm->name);
- kfree(dlm);
- dlm = NULL;
+ if (ret < 0)
goto leave;
- }
spin_lock_init(&dlm->spinlock);
spin_lock_init(&dlm->master_lock);
@@ -2028,7 +2022,6 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
INIT_LIST_HEAD(&dlm->list);
INIT_LIST_HEAD(&dlm->dirty_list);
INIT_LIST_HEAD(&dlm->reco.resources);
- INIT_LIST_HEAD(&dlm->reco.received);
INIT_LIST_HEAD(&dlm->reco.node_data);
INIT_LIST_HEAD(&dlm->purge_list);
INIT_LIST_HEAD(&dlm->dlm_domain_handlers);
@@ -2081,6 +2074,19 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
atomic_read(&dlm->dlm_refs.refcount));
leave:
+ if (ret < 0 && dlm) {
+ if (dlm->master_hash)
+ dlm_free_pagevec((void **)dlm->master_hash,
+ DLM_HASH_PAGES);
+
+ if (dlm->lockres_hash)
+ dlm_free_pagevec((void **)dlm->lockres_hash,
+ DLM_HASH_PAGES);
+
+ kfree(dlm->name);
+ kfree(dlm);
+ dlm = NULL;
+ }
return dlm;
}
@@ -2328,8 +2334,6 @@ static int __init dlm_init(void)
{
int status;
- dlm_print_version();
-
status = dlm_init_mle_cache();
if (status) {
mlog(ML_ERROR, "Could not create o2dlm_mle slabcache\n");
@@ -2379,6 +2383,7 @@ static void __exit dlm_exit (void)
MODULE_AUTHOR("Oracle");
MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("OCFS2 Distributed Lock Management");
module_init(dlm_init);
module_exit(dlm_exit);
OpenPOWER on IntegriCloud