aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ocfs2/dlm
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ocfs2/dlm')
-rw-r--r--fs/ocfs2/dlm/dlmcommon.h26
-rw-r--r--fs/ocfs2/dlm/dlmdomain.c13
-rw-r--r--fs/ocfs2/dlm/dlmmaster.c127
-rw-r--r--fs/ocfs2/dlm/dlmrecovery.c40
-rw-r--r--fs/ocfs2/dlm/dlmthread.c13
5 files changed, 199 insertions, 20 deletions
diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index 68c607e63ff6..004f2cbe8f71 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -282,6 +282,7 @@ static inline void __dlm_set_joining_node(struct dlm_ctxt *dlm,
#define DLM_LOCK_RES_DROPPING_REF 0x00000040
#define DLM_LOCK_RES_BLOCK_DIRTY 0x00001000
#define DLM_LOCK_RES_SETREF_INPROG 0x00002000
+#define DLM_LOCK_RES_RECOVERY_WAITING 0x00004000
/* max milliseconds to wait to sync up a network failure with a node death */
#define DLM_NODE_DEATH_WAIT_MAX (5 * 1000)
@@ -451,6 +452,7 @@ enum {
DLM_QUERY_REGION = 519,
DLM_QUERY_NODEINFO = 520,
DLM_BEGIN_EXIT_DOMAIN_MSG = 521,
+ DLM_DEREF_LOCKRES_DONE = 522,
};
struct dlm_reco_node_data
@@ -545,7 +547,7 @@ struct dlm_master_requery
* };
*
* from ../cluster/tcp.h
- * NET_MAX_PAYLOAD_BYTES (4096 - sizeof(net_msg))
+ * O2NET_MAX_PAYLOAD_BYTES (4096 - sizeof(net_msg))
* (roughly 4080 bytes)
* and sizeof(dlm_migratable_lockres) = 112 bytes
* and sizeof(dlm_migratable_lock) = 16 bytes
@@ -586,7 +588,7 @@ struct dlm_migratable_lockres
/* from above, 128 bytes
* for some undetermined future use */
-#define DLM_MIG_LOCKRES_RESERVED (NET_MAX_PAYLOAD_BYTES - \
+#define DLM_MIG_LOCKRES_RESERVED (O2NET_MAX_PAYLOAD_BYTES - \
DLM_MIG_LOCKRES_MAX_LEN)
struct dlm_create_lock
@@ -782,6 +784,20 @@ struct dlm_deref_lockres
u8 name[O2NM_MAX_NAME_LEN];
};
+enum {
+ DLM_DEREF_RESPONSE_DONE = 0,
+ DLM_DEREF_RESPONSE_INPROG = 1,
+};
+
+struct dlm_deref_lockres_done {
+ u32 pad1;
+ u16 pad2;
+ u8 node_idx;
+ u8 namelen;
+
+ u8 name[O2NM_MAX_NAME_LEN];
+};
+
static inline enum dlm_status
__dlm_lockres_state_to_status(struct dlm_lock_resource *res)
{
@@ -789,7 +805,8 @@ __dlm_lockres_state_to_status(struct dlm_lock_resource *res)
assert_spin_locked(&res->spinlock);
- if (res->state & DLM_LOCK_RES_RECOVERING)
+ if (res->state & (DLM_LOCK_RES_RECOVERING|
+ DLM_LOCK_RES_RECOVERY_WAITING))
status = DLM_RECOVERING;
else if (res->state & DLM_LOCK_RES_MIGRATING)
status = DLM_MIGRATING;
@@ -968,6 +985,8 @@ int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data,
void dlm_assert_master_post_handler(int status, void *data, void *ret_data);
int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
void **ret_data);
+int dlm_deref_lockres_done_handler(struct o2net_msg *msg, u32 len, void *data,
+ void **ret_data);
int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data,
void **ret_data);
int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
@@ -1009,6 +1028,7 @@ static inline void __dlm_wait_on_lockres(struct dlm_lock_resource *res)
{
__dlm_wait_on_lockres_flags(res, (DLM_LOCK_RES_IN_PROGRESS|
DLM_LOCK_RES_RECOVERING|
+ DLM_LOCK_RES_RECOVERY_WAITING|
DLM_LOCK_RES_MIGRATING));
}
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 2ee7fe747cea..12e064b8be9a 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -132,10 +132,13 @@ static DECLARE_WAIT_QUEUE_HEAD(dlm_domain_events);
* - Message DLM_QUERY_NODEINFO added to allow online node removes
* New in version 1.2:
* - Message DLM_BEGIN_EXIT_DOMAIN_MSG added to mark start of exit domain
+ * New in version 1.3:
+ * - Message DLM_DEREF_LOCKRES_DONE added to inform non-master that the
+ * refmap is cleared
*/
static const struct dlm_protocol_version dlm_protocol = {
.pv_major = 1,
- .pv_minor = 2,
+ .pv_minor = 3,
};
#define DLM_DOMAIN_BACKOFF_MS 200
@@ -1396,7 +1399,7 @@ static int dlm_send_join_cancels(struct dlm_ctxt *dlm,
unsigned int map_size)
{
int status, tmpstat;
- unsigned int node;
+ int node;
if (map_size != (BITS_TO_LONGS(O2NM_MAX_NODES) *
sizeof(unsigned long))) {
@@ -1853,7 +1856,13 @@ static int dlm_register_domain_handlers(struct dlm_ctxt *dlm)
sizeof(struct dlm_exit_domain),
dlm_begin_exit_domain_handler,
dlm, NULL, &dlm->dlm_domain_handlers);
+ if (status)
+ goto bail;
+ status = o2net_register_handler(DLM_DEREF_LOCKRES_DONE, dlm->key,
+ sizeof(struct dlm_deref_lockres_done),
+ dlm_deref_lockres_done_handler,
+ dlm, NULL, &dlm->dlm_domain_handlers);
bail:
if (status)
dlm_unregister_domain_handlers(dlm);
diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
index 9477d6e1de37..9aed6e202201 100644
--- a/fs/ocfs2/dlm/dlmmaster.c
+++ b/fs/ocfs2/dlm/dlmmaster.c
@@ -2278,7 +2278,7 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
dlm_print_one_lock_resource(res);
BUG();
}
- return ret;
+ return ret ? ret : r;
}
int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
@@ -2345,7 +2345,7 @@ int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
res->lockname.len, res->lockname.name, node);
dlm_print_one_lock_resource(res);
}
- ret = 0;
+ ret = DLM_DEREF_RESPONSE_DONE;
goto done;
}
@@ -2365,7 +2365,7 @@ int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
spin_unlock(&dlm->work_lock);
queue_work(dlm->dlm_worker, &dlm->dispatched_work);
- return 0;
+ return DLM_DEREF_RESPONSE_INPROG;
done:
if (res)
@@ -2375,6 +2375,122 @@ done:
return ret;
}
+int dlm_deref_lockres_done_handler(struct o2net_msg *msg, u32 len, void *data,
+ void **ret_data)
+{
+ struct dlm_ctxt *dlm = data;
+ struct dlm_deref_lockres_done *deref
+ = (struct dlm_deref_lockres_done *)msg->buf;
+ struct dlm_lock_resource *res = NULL;
+ char *name;
+ unsigned int namelen;
+ int ret = -EINVAL;
+ u8 node;
+ unsigned int hash;
+
+ if (!dlm_grab(dlm))
+ return 0;
+
+ name = deref->name;
+ namelen = deref->namelen;
+ node = deref->node_idx;
+
+ if (namelen > DLM_LOCKID_NAME_MAX) {
+ mlog(ML_ERROR, "Invalid name length!");
+ goto done;
+ }
+ if (deref->node_idx >= O2NM_MAX_NODES) {
+ mlog(ML_ERROR, "Invalid node number: %u\n", node);
+ goto done;
+ }
+
+ hash = dlm_lockid_hash(name, namelen);
+
+ spin_lock(&dlm->spinlock);
+ res = __dlm_lookup_lockres_full(dlm, name, namelen, hash);
+ if (!res) {
+ spin_unlock(&dlm->spinlock);
+ mlog(ML_ERROR, "%s:%.*s: bad lockres name\n",
+ dlm->name, namelen, name);
+ goto done;
+ }
+
+ spin_lock(&res->spinlock);
+ BUG_ON(!(res->state & DLM_LOCK_RES_DROPPING_REF));
+ if (!list_empty(&res->purge)) {
+ mlog(0, "%s: Removing res %.*s from purgelist\n",
+ dlm->name, res->lockname.len, res->lockname.name);
+ list_del_init(&res->purge);
+ dlm_lockres_put(res);
+ dlm->purge_count--;
+ }
+
+ if (!__dlm_lockres_unused(res)) {
+ mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
+ dlm->name, res->lockname.len, res->lockname.name);
+ __dlm_print_one_lock_resource(res);
+ BUG();
+ }
+
+ __dlm_unhash_lockres(dlm, res);
+
+ spin_lock(&dlm->track_lock);
+ if (!list_empty(&res->tracking))
+ list_del_init(&res->tracking);
+ else {
+ mlog(ML_ERROR, "%s: Resource %.*s not on the Tracking list\n",
+ dlm->name, res->lockname.len, res->lockname.name);
+ __dlm_print_one_lock_resource(res);
+ }
+ spin_unlock(&dlm->track_lock);
+
+ /* lockres is not in the hash now. drop the flag and wake up
+ * any processes waiting in dlm_get_lock_resource.
+ */
+ res->state &= ~DLM_LOCK_RES_DROPPING_REF;
+ spin_unlock(&res->spinlock);
+ wake_up(&res->wq);
+
+ dlm_lockres_put(res);
+
+ spin_unlock(&dlm->spinlock);
+
+done:
+ dlm_put(dlm);
+ return ret;
+}
+
+static void dlm_drop_lockres_ref_done(struct dlm_ctxt *dlm,
+ struct dlm_lock_resource *res, u8 node)
+{
+ struct dlm_deref_lockres_done deref;
+ int ret = 0, r;
+ const char *lockname;
+ unsigned int namelen;
+
+ lockname = res->lockname.name;
+ namelen = res->lockname.len;
+ BUG_ON(namelen > O2NM_MAX_NAME_LEN);
+
+ memset(&deref, 0, sizeof(deref));
+ deref.node_idx = dlm->node_num;
+ deref.namelen = namelen;
+ memcpy(deref.name, lockname, namelen);
+
+ ret = o2net_send_message(DLM_DEREF_LOCKRES_DONE, dlm->key,
+ &deref, sizeof(deref), node, &r);
+ if (ret < 0) {
+ mlog(ML_ERROR, "%s: res %.*s, error %d send DEREF DONE "
+ " to node %u\n", dlm->name, namelen,
+ lockname, ret, node);
+ } else if (r < 0) {
+ /* ignore the error */
+ mlog(ML_ERROR, "%s: res %.*s, DEREF to node %u got %d\n",
+ dlm->name, namelen, lockname, node, r);
+ dlm_print_one_lock_resource(res);
+ }
+}
+
static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data)
{
struct dlm_ctxt *dlm;
@@ -2395,6 +2511,8 @@ static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data)
}
spin_unlock(&res->spinlock);
+ dlm_drop_lockres_ref_done(dlm, res, node);
+
if (cleared) {
mlog(0, "%s:%.*s node %u ref dropped in dispatch\n",
dlm->name, res->lockname.len, res->lockname.name, node);
@@ -2432,7 +2550,8 @@ static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm,
return 0;
/* delay migration when the lockres is in RECOCERING state */
- if (res->state & DLM_LOCK_RES_RECOVERING)
+ if (res->state & (DLM_LOCK_RES_RECOVERING|
+ DLM_LOCK_RES_RECOVERY_WAITING))
return 0;
if (res->owner != dlm->node_num)
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index b94a425f0175..cd38488a10fc 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -1403,12 +1403,24 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
* and RECOVERY flag changed when it completes. */
hash = dlm_lockid_hash(mres->lockname, mres->lockname_len);
spin_lock(&dlm->spinlock);
- res = __dlm_lookup_lockres(dlm, mres->lockname, mres->lockname_len,
+ res = __dlm_lookup_lockres_full(dlm, mres->lockname, mres->lockname_len,
hash);
if (res) {
/* this will get a ref on res */
/* mark it as recovering/migrating and hash it */
spin_lock(&res->spinlock);
+ if (res->state & DLM_LOCK_RES_DROPPING_REF) {
+ mlog(0, "%s: node is attempting to migrate "
+ "lockres %.*s, but marked as dropping "
+ " ref!\n", dlm->name,
+ mres->lockname_len, mres->lockname);
+ ret = -EINVAL;
+ spin_unlock(&res->spinlock);
+ spin_unlock(&dlm->spinlock);
+ dlm_lockres_put(res);
+ goto leave;
+ }
+
if (mres->flags & DLM_MRES_RECOVERY) {
res->state |= DLM_LOCK_RES_RECOVERING;
} else {
@@ -2163,6 +2175,13 @@ 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, bucket, hash_node) {
+ if (res->state & DLM_LOCK_RES_RECOVERY_WAITING) {
+ spin_lock(&res->spinlock);
+ res->state &= ~DLM_LOCK_RES_RECOVERY_WAITING;
+ spin_unlock(&res->spinlock);
+ wake_up(&res->wq);
+ }
+
if (!(res->state & DLM_LOCK_RES_RECOVERING))
continue;
@@ -2300,6 +2319,7 @@ 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);
}
+ res->state |= DLM_LOCK_RES_RECOVERY_WAITING;
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 "
@@ -2377,14 +2397,16 @@ 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, "%s: res %.*s, Skip "
- "recovery as it is being freed\n",
- dlm->name, res->lockname.len,
- res->lockname.name);
- } else
- dlm_move_lockres_to_recovery_list(dlm,
- res);
-
+ mlog(0, "%s:%.*s: owned by "
+ "dead node %u, this node was "
+ "dropping its ref when it died. "
+ "continue, dropping the flag.\n",
+ dlm->name, res->lockname.len,
+ res->lockname.name, dead_node);
+ }
+ res->state &= ~DLM_LOCK_RES_DROPPING_REF;
+ dlm_move_lockres_to_recovery_list(dlm,
+ res);
} else if (res->owner == dlm->node_num) {
dlm_free_dead_locks(dlm, res, dead_node);
__dlm_lockres_calc_usage(dlm, res);
diff --git a/fs/ocfs2/dlm/dlmthread.c b/fs/ocfs2/dlm/dlmthread.c
index c5f6c241ecd7..68d239ba0c63 100644
--- a/fs/ocfs2/dlm/dlmthread.c
+++ b/fs/ocfs2/dlm/dlmthread.c
@@ -106,7 +106,8 @@ int __dlm_lockres_unused(struct dlm_lock_resource *res)
if (!list_empty(&res->dirty) || res->state & DLM_LOCK_RES_DIRTY)
return 0;
- if (res->state & DLM_LOCK_RES_RECOVERING)
+ if (res->state & (DLM_LOCK_RES_RECOVERING|
+ DLM_LOCK_RES_RECOVERY_WAITING))
return 0;
/* Another node has this resource with this node as the master */
@@ -202,6 +203,13 @@ static void dlm_purge_lockres(struct dlm_ctxt *dlm,
dlm->purge_count--;
}
+ if (!master && ret != 0) {
+ mlog(0, "%s: deref %.*s in progress or master goes down\n",
+ dlm->name, res->lockname.len, res->lockname.name);
+ spin_unlock(&res->spinlock);
+ return;
+ }
+
if (!__dlm_lockres_unused(res)) {
mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
dlm->name, res->lockname.len, res->lockname.name);
@@ -700,7 +708,8 @@ static int dlm_thread(void *data)
* dirty for a short while. */
BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
if (res->state & (DLM_LOCK_RES_IN_PROGRESS |
- DLM_LOCK_RES_RECOVERING)) {
+ DLM_LOCK_RES_RECOVERING |
+ DLM_LOCK_RES_RECOVERY_WAITING)) {
/* move it to the tail and keep going */
res->state &= ~DLM_LOCK_RES_DIRTY;
spin_unlock(&res->spinlock);
Copyright © 1996 – 2023 Jason A. Donenfeld. All Rights Reverse Engineered.