Re: "deadlock" between smc91x driver and link_watch

[Nicolas Pitre <nico@xxxxxxx>]

On Wed, 24 Nov 2004, Ian Campbell wrote:

> Quite right. Fixed patch included.

Small question:

> + * smc_phy_configure_wq
> + *
> + * The net_device is referenced when the work was scheduled to avoid
> + * the need for a flush_scheduled_work() in smc_close(). Drop the
> + * reference and then do the configuration.

You probably want to invert the comment here too.

> +static void smc_phy_configure_wq(void *data)
> +{
> +     struct net_device *dev = data;
> +     smc_phy_configure(data);
> +     dev_put(dev);
> +}

[...]

> @@ -1536,10 +1553,8 @@
>       /* clear everything */
>       smc_shutdown(dev);
>  
> -     if (lp->phy_type != 0) {
> -             flush_scheduled_work();
> +     if (lp->phy_type != 0)
>               smc_phy_powerdown(dev, lp->mii.phy_id);


How do you ensure that smc_phy_configure() can't end up being called 
after smc_phy_powerdown() here?


Nicolas