We cannot do the cancel_work_sync from after the unregister_netdev, as
the dev pointer is no longer valid, causing a uaf on ldisc unregister
(or device close).
Instead, do the cancel_work_sync from the ndo_uninit op, where the dev
still exists, but the queue has stopped.
Fixes: 584bcfe236c0 ("mctp: serial: cancel tx work on ldisc close")
Reported-by: Luo Likang <luolikang@nsfocus.com>
Tested-by: Luo Likang <luolikang@nsfocus.com>
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au>
Link: https://lore.kernel.org/r/20220211011552.1861886-1-jk@codeconstruct.com.au
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
mctp_serial_push(dev, c[i]);
}
+static void mctp_serial_uninit(struct net_device *ndev)
+{
+ struct mctp_serial *dev = netdev_priv(ndev);
+
+ cancel_work_sync(&dev->tx_work);
+}
+
static const struct net_device_ops mctp_serial_netdev_ops = {
.ndo_start_xmit = mctp_serial_tx,
+ .ndo_uninit = mctp_serial_uninit,
};
static void mctp_serial_setup(struct net_device *ndev)
int idx = dev->idx;
unregister_netdev(dev->netdev);
- cancel_work_sync(&dev->tx_work);
ida_free(&mctp_serial_ida, idx);
}