1 files changed, 76 insertions, 50 deletions

diff --git a/fs/ocfs2/dlm/dlmunlock.c b/fs/ocfs2/dlm/dlmunlock.c
index cec2ce1cd31..2e3c9dbab68 100644
--- a/fs/ocfs2/dlm/dlmunlock.c
+++ b/fs/ocfs2/dlm/dlmunlock.c
@@ -28,9 +28,7 @@
 #include <linux/module.h>
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
-#include <linux/utsname.h>
 #include <linux/init.h>
 #include <linux/sysctl.h>
 #include <linux/random.h>
@@ -117,12 +115,12 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
 	else
 		BUG_ON(res->owner == dlm->node_num);
 
-	spin_lock(&dlm->spinlock);
+	spin_lock(&dlm->ast_lock);
 	/* We want to be sure that we're not freeing a lock
 	 * that still has AST's pending... */
 	in_use = !list_empty(&lock->ast_list);
-	spin_unlock(&dlm->spinlock);
-	if (in_use) {
+	spin_unlock(&dlm->ast_lock);
+	if (in_use && !(flags & LKM_CANCEL)) {
 	       mlog(ML_ERROR, "lockres %.*s: Someone is calling dlmunlock "
 		    "while waiting for an ast!", res->lockname.len,
 		    res->lockname.name);
@@ -131,7 +129,7 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
 
 	spin_lock(&res->spinlock);
 	if (res->state & DLM_LOCK_RES_IN_PROGRESS) {
-		if (master_node) {
+		if (master_node && !(flags & LKM_CANCEL)) {
 			mlog(ML_ERROR, "lockres in progress!\n");
 			spin_unlock(&res->spinlock);
 			return DLM_FORWARD;
@@ -147,6 +145,10 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
 		goto leave;
 	}
 
+	if (res->state & DLM_LOCK_RES_MIGRATING) {
+		status = DLM_MIGRATING;
+		goto leave;
+	}
 
 	/* see above for what the spec says about
 	 * LKM_CANCEL and the lock queue state */
@@ -155,7 +157,7 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
 	else
 		status = dlm_get_unlock_actions(dlm, res, lock, lksb, &actions);
 
-	if (status != DLM_NORMAL)
+	if (status != DLM_NORMAL && (status != DLM_CANCELGRANT || !master_node))
 		goto leave;
 
 	/* By now this has been masked out of cancel requests. */
@@ -183,11 +185,26 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
 		spin_lock(&lock->spinlock);
 		/* if the master told us the lock was already granted,
 		 * let the ast handle all of these actions */
-		if (status == DLM_NORMAL &&
-		    lksb->status == DLM_CANCELGRANT) {
+		if (status == DLM_CANCELGRANT) {
 			actions &= ~(DLM_UNLOCK_REMOVE_LOCK|
 				     DLM_UNLOCK_REGRANT_LOCK|
 				     DLM_UNLOCK_CLEAR_CONVERT_TYPE);
+		} else if (status == DLM_RECOVERING ||
+			   status == DLM_MIGRATING ||
+			   status == DLM_FORWARD ||
+			   status == DLM_NOLOCKMGR
+			   ) {
+			/* must clear the actions because this unlock
+			 * is about to be retried.  cannot free or do
+			 * any list manipulation. */
+			mlog(0, "%s:%.*s: clearing actions, %s\n",
+			     dlm->name, res->lockname.len,
+			     res->lockname.name,
+			     status==DLM_RECOVERING?"recovering":
+			     (status==DLM_MIGRATING?"migrating":
+				(status == DLM_FORWARD ? "forward" :
+						"nolockmanager")));
+			actions = 0;
 		}
 		if (flags & LKM_CANCEL)
 			lock->cancel_pending = 0;
@@ -231,8 +248,10 @@ leave:
 	if (actions & DLM_UNLOCK_FREE_LOCK) {
 		/* this should always be coupled with list removal */
 		BUG_ON(!(actions & DLM_UNLOCK_REMOVE_LOCK));
-		mlog(0, "lock %"MLFu64" should be gone now! refs=%d\n",
-		     lock->ml.cookie, atomic_read(&lock->lock_refs.refcount)-1);
+		mlog(0, "lock %u:%llu should be gone now! refs=%d\n",
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+		     atomic_read(&lock->lock_refs.refcount)-1);
 		dlm_lock_put(lock);
 	}
 	if (actions & DLM_UNLOCK_CALL_AST)
@@ -256,8 +275,7 @@ void dlm_commit_pending_unlock(struct dlm_lock_resource *res,
 void dlm_commit_pending_cancel(struct dlm_lock_resource *res,
 			       struct dlm_lock *lock)
 {
-	list_del_init(&lock->list);
-	list_add_tail(&lock->list, &res->granted);
+	list_move_tail(&lock->list, &res->granted);
 	lock->ml.convert_type = LKM_IVMODE;
 }
 
@@ -302,7 +320,17 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
 	struct kvec vec[2];
 	size_t veclen = 1;
 
-	mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);
+	mlog(0, "%.*s\n", res->lockname.len, res->lockname.name);
+
+	if (owner == dlm->node_num) {
+		/* ended up trying to contact ourself.  this means
+		 * that the lockres had been remote but became local
+		 * via a migration.  just retry it, now as local */
+		mlog(0, "%s:%.*s: this node became the master due to a "
+		     "migration, re-evaluate now\n", dlm->name,
+		     res->lockname.len, res->lockname.name);
+		return DLM_FORWARD;
+	}
 
 	memset(&unlock, 0, sizeof(unlock));
 	unlock.node_idx = dlm->node_num;
@@ -325,16 +353,12 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
 					vec, veclen, owner, &status);
 	if (tmpret >= 0) {
 		// successfully sent and received
-		if (status == DLM_CANCELGRANT)
-			ret = DLM_NORMAL;
-		else if (status == DLM_FORWARD) {
+		if (status == DLM_FORWARD)
 			mlog(0, "master was in-progress.  retry\n");
-			ret = DLM_FORWARD;
-		} else
-			ret = status;
-		lksb->status = status;
+		ret = status;
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_UNLOCK_LOCK_MSG, dlm->key, owner);
 		if (dlm_is_host_down(tmpret)) {
 			/* NOTE: this seems strange, but it is what we want.
 			 * when the master goes down during a cancel or
@@ -343,12 +367,14 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
 			 * updated state to the recovery master.  this thread
 			 * just needs to finish out the operation and call
 			 * the unlockast. */
-			ret = DLM_NORMAL;
+			if (dlm_is_node_dead(dlm, owner))
+				ret = DLM_NORMAL;
+			else
+				ret = DLM_NOLOCKMGR;
 		} else {
 			/* something bad.  this will BUG in ocfs2 */
 			ret = dlm_err_to_dlm_status(tmpret);
 		}
-		lksb->status = ret;
 	}
 
 	return ret;
@@ -362,12 +388,12 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
  * returns: DLM_NORMAL, DLM_BADARGS, DLM_IVLOCKID,
  *          return value from dlmunlock_master
  */
-int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data)
+int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data,
+			    void **ret_data)
 {
 	struct dlm_ctxt *dlm = data;
 	struct dlm_unlock_lock *unlock = (struct dlm_unlock_lock *)msg->buf;
 	struct dlm_lock_resource *res = NULL;
-	struct list_head *iter;
 	struct dlm_lock *lock = NULL;
 	enum dlm_status status = DLM_NORMAL;
 	int found = 0, i;
@@ -437,8 +463,7 @@ int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data)
 	}
 
 	for (i=0; i<3; i++) {
-		list_for_each(iter, queue) {
-			lock = list_entry(iter, struct dlm_lock, list);
+		list_for_each_entry(lock, queue, list) {
 			if (lock->ml.cookie == unlock->cookie &&
 		    	    lock->ml.node == unlock->node_idx) {
 				dlm_lock_get(lock);
@@ -459,6 +484,10 @@ int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data)
 
 	/* lock was found on queue */
 	lksb = lock->lksb;
+	if (flags & (LKM_VALBLK|LKM_PUT_LVB) &&
+	    lock->ml.type != LKM_EXMODE)
+		flags &= ~(LKM_VALBLK|LKM_PUT_LVB);
+
 	/* unlockast only called on originating node */
 	if (flags & LKM_PUT_LVB) {
 		lksb->flags |= DLM_LKSB_PUT_LVB;
@@ -480,13 +509,11 @@ int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data)
 not_found:
 	if (!found)
 		mlog(ML_ERROR, "failed to find lock to unlock! "
-			       "cookie=%"MLFu64"\n",
-		     unlock->cookie);
-	else {
-		/* send the lksb->status back to the other node */
-		status = lksb->status;
+			       "cookie=%u:%llu\n",
+		     dlm_get_lock_cookie_node(be64_to_cpu(unlock->cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(unlock->cookie)));
+	else
 		dlm_lock_put(lock);
-	}
 
 leave:
 	if (res)
@@ -508,26 +535,22 @@ static enum dlm_status dlm_get_cancel_actions(struct dlm_ctxt *dlm,
 
 	if (dlm_lock_on_list(&res->blocked, lock)) {
 		/* cancel this outright */
-		lksb->status = DLM_NORMAL;
 		status = DLM_NORMAL;
 		*actions = (DLM_UNLOCK_CALL_AST |
 			    DLM_UNLOCK_REMOVE_LOCK);
 	} else if (dlm_lock_on_list(&res->converting, lock)) {
 		/* cancel the request, put back on granted */
-		lksb->status = DLM_NORMAL;
 		status = DLM_NORMAL;
 		*actions = (DLM_UNLOCK_CALL_AST |
 			    DLM_UNLOCK_REMOVE_LOCK |
 			    DLM_UNLOCK_REGRANT_LOCK |
 			    DLM_UNLOCK_CLEAR_CONVERT_TYPE);
 	} else if (dlm_lock_on_list(&res->granted, lock)) {
-		/* too late, already granted.  DLM_CANCELGRANT */
-		lksb->status = DLM_CANCELGRANT;
-		status = DLM_NORMAL;
+		/* too late, already granted. */
+		status = DLM_CANCELGRANT;
 		*actions = DLM_UNLOCK_CALL_AST;
 	} else {
 		mlog(ML_ERROR, "lock to cancel is not on any list!\n");
-		lksb->status = DLM_IVLOCKID;
 		status = DLM_IVLOCKID;
 		*actions = 0;
 	}
@@ -544,13 +567,11 @@ static enum dlm_status dlm_get_unlock_actions(struct dlm_ctxt *dlm,
 
 	/* unlock request */
 	if (!dlm_lock_on_list(&res->granted, lock)) {
-		lksb->status = DLM_DENIED;
 		status = DLM_DENIED;
 		dlm_error(status);
 		*actions = 0;
 	} else {
 		/* unlock granted lock */
-		lksb->status = DLM_NORMAL;
 		status = DLM_NORMAL;
 		*actions = (DLM_UNLOCK_FREE_LOCK |
 			    DLM_UNLOCK_CALL_AST |
@@ -571,8 +592,6 @@ enum dlm_status dlmunlock(struct dlm_ctxt *dlm, struct dlm_lockstatus *lksb,
 	struct dlm_lock *lock = NULL;
 	int call_ast, is_master;
 
-	mlog_entry_void();
-
 	if (!lksb) {
 		dlm_error(DLM_BADARGS);
 		return DLM_BADARGS;
@@ -607,6 +626,8 @@ retry:
 
 	spin_lock(&res->spinlock);
 	is_master = (res->owner == dlm->node_num);
+	if (flags & LKM_VALBLK && lock->ml.type != LKM_EXMODE)
+		flags &= ~LKM_VALBLK;
 	spin_unlock(&res->spinlock);
 
 	if (is_master) {
@@ -623,7 +644,9 @@ retry:
 
 	if (status == DLM_RECOVERING ||
 	    status == DLM_MIGRATING ||
-	    status == DLM_FORWARD) {
+	    status == DLM_FORWARD ||
+	    status == DLM_NOLOCKMGR) {
+
 		/* We want to go away for a tiny bit to allow recovery
 		 * / migration to complete on this resource. I don't
 		 * know of any wait queue we could sleep on as this
@@ -635,26 +658,29 @@ retry:
 		msleep(50);
 
 		mlog(0, "retrying unlock due to pending recovery/"
-		     "migration/in-progress\n");
+		     "migration/in-progress/reconnect\n");
 		goto retry;
 	}
 
 	if (call_ast) {
-		mlog(0, "calling unlockast(%p, %d)\n", data, lksb->status);
+		mlog(0, "calling unlockast(%p, %d)\n", data, status);
 		if (is_master) {
-			/* it is possible that there is one last bast 
+			/* it is possible that there is one last bast
 			 * pending.  make sure it is flushed, then
 			 * call the unlockast.
 			 * not an issue if this is a mastered remotely,
 			 * since this lock has been removed from the
 			 * lockres queues and cannot be found. */
 			dlm_kick_thread(dlm, NULL);
-			wait_event(dlm->ast_wq, 
+			wait_event(dlm->ast_wq,
 				   dlm_lock_basts_flushed(dlm, lock));
 		}
-		(*unlockast)(data, lksb->status);
+		(*unlockast)(data, status);
 	}
 
+	if (status == DLM_CANCELGRANT)
+		status = DLM_NORMAL;
+
 	if (status == DLM_NORMAL) {
 		mlog(0, "kicking the thread\n");
 		dlm_kick_thread(dlm, res);