We should destroy the switch workqueue only after deregistering the
switchdev notifiers. Without this fix, we could end up with switchdev
notifications on a draining workqueue and also with a lock up since the
netdevice reference count is increased (in port_switchdev_event) and not
decreased ever (since the workqueue did not run).
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Link: https://lore.kernel.org/r/20200721091919.20394-5-ioana.ciornei@nxp.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ethsw_teardown_irqs(sw_dev);
- destroy_workqueue(ethsw->workqueue);
-
dpsw_disable(ethsw->mc_io, 0, ethsw->dpsw_handle);
for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
kfree(ethsw->ports);
ethsw_takedown(sw_dev);
+
+ destroy_workqueue(ethsw->workqueue);
+
fsl_mc_portal_free(ethsw->mc_io);
kfree(ethsw);