[DLM] fix user unlocking

When a user process exits, we clear all the locks it holds.  There is a
problem, though, with locks that the process had begun unlocking before it
exited.  We couldn't find the lkb's that were in the process of being
unlocked remotely, to flag that they are DEAD.  To solve this, we move
lkb's being unlocked onto a new list in the per-process structure that
tracks what locks the process is holding.  We can then go through this
list to flag the necessary lkb's when clearing locks for a process when it
exits.
Signed-off-by: NDavid Teigland <teigland@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

fs/dlm/dlm_internal.h

@@ -526,6 +526,7 @@ struct dlm_user_proc {
 	spinlock_t		asts_spin;
 	struct list_head	locks;
 	spinlock_t		locks_spin;
+	struct list_head	unlocking;
 	wait_queue_head_t	wait;
 };
 

fs/dlm/lock.c

@@ -3772,12 +3772,10 @@ int dlm_user_unlock(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
 		goto out_put;
 
 	spin_lock(&ua->proc->locks_spin);
-	list_del_init(&lkb->lkb_ownqueue);
+	/* dlm_user_add_ast() may have already taken lkb off the proc list */
+	if (!list_empty(&lkb->lkb_ownqueue))
+		list_move(&lkb->lkb_ownqueue, &ua->proc->unlocking);
 	spin_unlock(&ua->proc->locks_spin);
-
-	/* this removes the reference for the proc->locks list added by
-	   dlm_user_request */
-	unhold_lkb(lkb);
  out_put:
 	dlm_put_lkb(lkb);
  out:
@@ -3817,9 +3815,8 @@ int dlm_user_cancel(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
 	/* this lkb was removed from the WAITING queue */
 	if (lkb->lkb_grmode == DLM_LOCK_IV) {
 		spin_lock(&ua->proc->locks_spin);
-		list_del_init(&lkb->lkb_ownqueue);
+		list_move(&lkb->lkb_ownqueue, &ua->proc->unlocking);
 		spin_unlock(&ua->proc->locks_spin);
-		unhold_lkb(lkb);
 	}
  out_put:
 	dlm_put_lkb(lkb);
@@ -3880,11 +3877,6 @@ void dlm_clear_proc_locks(struct dlm_ls *ls, struct dlm_user_proc *proc)
 	mutex_lock(&ls->ls_clear_proc_locks);
 
 	list_for_each_entry_safe(lkb, safe, &proc->locks, lkb_ownqueue) {
-		if (lkb->lkb_ast_type) {
-			list_del(&lkb->lkb_astqueue);
-			unhold_lkb(lkb);
-		}
-
 		list_del_init(&lkb->lkb_ownqueue);
 
 		if (lkb->lkb_exflags & DLM_LKF_PERSISTENT) {
@@ -3901,6 +3893,20 @@ void dlm_clear_proc_locks(struct dlm_ls *ls, struct dlm_user_proc *proc)
 
 		dlm_put_lkb(lkb);
 	}
+
+	/* in-progress unlocks */
+	list_for_each_entry_safe(lkb, safe, &proc->unlocking, lkb_ownqueue) {
+		list_del_init(&lkb->lkb_ownqueue);
+		lkb->lkb_flags |= DLM_IFL_DEAD;
+		dlm_put_lkb(lkb);
+	}
+
+	list_for_each_entry_safe(lkb, safe, &proc->asts, lkb_astqueue) {
+		list_del(&lkb->lkb_astqueue);
+		dlm_put_lkb(lkb);
+	}
+
 	mutex_unlock(&ls->ls_clear_proc_locks);
 	unlock_recovery(ls);
 }
+

fs/dlm/user.c

@@ -180,6 +180,14 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, int type)
 	    ua->lksb.sb_status == -EAGAIN && !list_empty(&lkb->lkb_ownqueue))
 		remove_ownqueue = 1;
 
+	/* unlocks or cancels of waiting requests need to be removed from the
+	   proc's unlocking list, again there must be a better way...  */
+
+	if (ua->lksb.sb_status == -DLM_EUNLOCK ||
+	    (ua->lksb.sb_status == -DLM_ECANCEL &&
+	     lkb->lkb_grmode == DLM_LOCK_IV))
+		remove_ownqueue = 1;
+
 	/* We want to copy the lvb to userspace when the completion
 	   ast is read if the status is 0, the lock has an lvb and
 	   lvb_ops says we should.  We could probably have set_lvb_lock()
@@ -523,6 +531,7 @@ static int device_open(struct inode *inode, struct file *file)
 	proc->lockspace = ls->ls_local_handle;
 	INIT_LIST_HEAD(&proc->asts);
 	INIT_LIST_HEAD(&proc->locks);
+	INIT_LIST_HEAD(&proc->unlocking);
 	spin_lock_init(&proc->asts_spin);
 	spin_lock_init(&proc->locks_spin);
 	init_waitqueue_head(&proc->wait);