In its attempt to run only the minimal amount of tear down steps,
qeth_l2_stop_card() fails to reset the "is dev_addr registered?" flag
in some rare scenarios. But a future change to the tear down sequence
would cause us to _always_ hit this issue, so patch it up before that
code lands.
Fix it by unconditionally clearing the flag bit. This also allows us to
remove the additional cleanup step in qeth_dev_layer2_store().
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
goto out;
}
- card->info.mac_bits = 0;
if (card->discipline) {
/* start with a new, pristine netdevice: */
ndev = qeth_clone_netdev(card->dev);
dev_close(card->dev);
rtnl_unlock();
}
- card->info.mac_bits &= ~QETH_LAYER2_MAC_REGISTERED;
card->state = CARD_STATE_SOFTSETUP;
}
if (card->state == CARD_STATE_SOFTSETUP) {
}
flush_workqueue(card->event_wq);
+ card->info.mac_bits &= ~QETH_LAYER2_MAC_REGISTERED;
}
static int qeth_l2_process_inbound_buffer(struct qeth_card *card,