Re: [PATCH] Netpoll support for Sibyte MAC

[Sergei Shtylyov <sshtylyov@ru.mvista.com>]

Hello.

Deepak Saxena wrote:

> NETPOLL support for Sibyte MAC

> Index: linux-2.6.18/drivers/net/sb1250-mac.c
> ===================================================================
> --- linux-2.6.18.orig/drivers/net/sb1250-mac.c
> +++ linux-2.6.18/drivers/net/sb1250-mac.c
> @@ -1128,6 +1128,26 @@ static void sbdma_fillring(sbmacdma_t *d
>  	}
>  }
>  
> +#ifdef CONFIG_NET_POLL_CONTROLLER
> +static void sbmac_netpoll(struct net_device *netdev)
> +{
> +	struct sbmac_softc *sc = netdev_priv(netdev);
> +	int irq = sc->sbm_dev->irq;
> +
> +	__raw_writeq(0, sc->sbm_imr);
> +

    Thinking about it again, I'm not sure that blindly writing to the chip 
reg. to mask interrupts is SMP-safe enough (versus disable_irq()).  I know why 
it was done this way -- to quell the BUG emitted in the realtime mode on SMP 
-- because of scheduling with disabled interrupts in synchoronize_irq(), but 
after having spent much time on netpoll, I'm no longer sure that anything but 
disable_irq() is safe enough for SMP since -- otherwise there's no warranty 
that sbmac_intr() is not running on another CPU...

> +	sbmac_intr(irq, netdev, NULL);
> +
> +#ifdef CONFIG_SBMAC_COALESCE
> +	__raw_writeq(((M_MAC_INT_EOP_COUNT | M_MAC_INT_EOP_TIMER) << S_MAC_TX_CH0) |
> +	((M_MAC_INT_EOP_COUNT | M_MAC_INT_EOP_TIMER) << S_MAC_RX_CH0),
> +	sc->sbm_imr);
> +#else
> +	__raw_writeq((M_MAC_INT_CHANNEL << S_MAC_TX_CH0) | 
> +	(M_MAC_INT_CHANNEL << S_MAC_RX_CH0), sc->sbm_imr);
> +#endif
> +}
> +#endif

WBR, Sergei