Dangling reference from dmu_objset_upgrade

After porting the fix for openzfs#5295
over to illumos, we started hitting an assertion failure when running
the testsuite:

	assertion failed: rc->rc_count == number, file: .../refcount.c

and the unexpected hold has this stack:

	dsl_dataset_long_hold+0x59 dmu_objset_upgrade+0x73
dmu_objset_id_quota_upgrade+0x15 dmu_objset_own+0x14f

The simplest reproducer for this in illumos is

    zpool create -f -O version=1 testpool c3t0d0; zpool destroy testpool

which is run as part of the zpool_create_tempname test, but I can't get
this to trigger on FreeBSD. This appears to be because of the call to
txg_wait_synced() in dmu_objset_upgrade_stop() (which was missing in
illumos), slows down dmu_objset_disown() enough to avoid the condition.

Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Andy Fiddaman <[email protected]>
Closes openzfs#11368

module/zfs/dmu_objset.c

@@ -1428,10 +1428,15 @@ dmu_objset_upgrade_task_cb(void *data)
 	mutex_enter(&os->os_upgrade_lock);
 	os->os_upgrade_status = EINTR;
 	if (!os->os_upgrade_exit) {
+		int status;
+
 		mutex_exit(&os->os_upgrade_lock);
 
-		os->os_upgrade_status = os->os_upgrade_cb(os);
+		status = os->os_upgrade_cb(os);
+
 		mutex_enter(&os->os_upgrade_lock);
+
+		os->os_upgrade_status = status;
 	}
 	os->os_upgrade_exit = B_TRUE;
 	os->os_upgrade_id = 0;
@@ -1459,6 +1464,8 @@ dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb)
 			dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
 			os->os_upgrade_status = ENOMEM;
 		}
+	} else {
+		dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
 	}
 	mutex_exit(&os->os_upgrade_lock);
 }