Florian Fainelli <f.fainelli@xxxxxxxxx> wrote :
> 2014-03-18 15:16 GMT-07:00 Byungho An <bh74.an@xxxxxxxxxxx>:
> > Florian Fainelli <f.fainelli@xxxxxxxxx> wrote:
> >> 2014-03-18 11:19 GMT-07:00 Byungho An <bh74.an@xxxxxxxxxxx>:
> >> > From: Siva Reddy <siva.kallam@xxxxxxxxxxx>
> >> >
> >> > This patch adds support for Samsung 10Gb ethernet driver(sxgbe).
> >> >
> >> > - sxgbe core initialization
> >> > - Tx and Rx support
> >> > - MDIO support
> >> > - ISRs for Tx and Rx
> >> > - ifconfig support to driver
> >>
> >> Too many files to review at once, the diffstat was around 5000+ lines!
> >> You should split this into logical parts, or just submit the basic
> >> bits for now and add the features later.
> > Yes. but I think it is one logically, after this patch there won't be much code at once.
>
> Since you are adding multiple objects to the Makefile at once, you could very well split the patches to reflect that and submit one
> object as a separate patch. Also, sometimes, even if the driver is feature complete, it helps submitting the basic support first, and
> then start adding new features to it progressively.
But it is basic framework itself.
Since this #2 patch contains changes that are required for its basic functionality, it is not easy to split.
>
> >
> >>
> >> [snip]
> >>
> >> > + ret = register_netdev(ndev);
> >> > + if (ret) {
> >> > + pr_err("%s: ERROR %i registering the device\n", __func__, ret);
> >> > + goto error_netdev_register;
> >> > + }
> >> > +
> >> > + priv->sxgbe_clk = clk_get(priv->device, SXGBE_RESOURCE_NAME);
> >> > + if (IS_ERR(priv->sxgbe_clk)) {
> >> > + netdev_warn(ndev, "%s: warning: cannot get CSR clock\n",
> >> > + __func__);
> >> > + goto error_clk_get;
> >> > + }
> >>
> >> This is racy, after register_netdev() is called, the network stack is
> >> free to use the interface, which means that as much as possible needs
> >> to be initialized before you call it, including clocks, and MDIO.
> >>
> >> > +
> >> > + /* If a specific clk_csr value is passed from the platform
> >> > + * this means that the CSR Clock Range selection cannot be
> >> > + * changed at run-time and it is fixed. Viceversa the driver'll try to
> >> > + * set the MDC clock dynamically according to the csr actual
> >> > + * clock input.
> >> > + */
> >> > + if (!priv->plat->clk_csr)
> >> > + sxgbe_clk_csr_set(priv);
> >> > + else
> >> > + priv->clk_csr = priv->plat->clk_csr;
> >> > +
> >> > + /* MDIO bus Registration */
> >> > + ret = sxgbe_mdio_register(ndev);
> >> > + if (ret < 0) {
> >> > + netdev_dbg(ndev, "%s: MDIO bus (id: %d) registration failed\n",
> >> > + __func__, priv->plat->bus_id);
> >> > + goto error_mdio_register;
> >> > + }
> >>
> >> Don't do this, register your MDIO bus before calling register_netdev().
> > OK.
> >
> >>
> >> > +
> >> > + sxgbe_check_ether_addr(priv);
> >> > +
> >> > + return priv;
> >> > +
> >> > +error_mdio_register:
> >> > + clk_put(priv->sxgbe_clk);
> >> > +error_clk_get:
> >> > + unregister_netdev(ndev);
> >> > +error_netdev_register:
> >> > + netif_napi_del(&priv->napi);
> >> > +error_free_netdev:
> >> > + free_netdev(ndev);
> >> > +
> >> > + return NULL;
> >> > +}
> >> > +
> >> > +/**
> >> > + * sxgbe_dvr_remove
> >> > + * @ndev: net device pointer
> >> > + * Description: this function resets the TX/RX processes, disables
> >> > +the MAC RX/TX
> >> > + * changes the link status, releases the DMA descriptor rings.
> >> > + */
> >> > +int sxgbe_dvr_remove(struct net_device *ndev) {
> >> > + struct sxgbe_priv_data *priv = netdev_priv(ndev);
> >> > +
> >> > + netdev_info(ndev, "%s: removing driver\n", __func__);
> >> > +
> >> > + priv->hw->dma->stop_rx(priv->ioaddr, SXGBE_RX_QUEUES);
> >> > + priv->hw->dma->stop_tx(priv->ioaddr, SXGBE_TX_QUEUES);
> >> > +
> >> > + priv->hw->mac->enable_tx(priv->ioaddr, false);
> >> > + priv->hw->mac->enable_rx(priv->ioaddr, false);
> >> > +
> >> > + netif_napi_del(&priv->napi);
> >> > +
> >> > + sxgbe_mdio_unregister(ndev);
> >> > +
> >> > + netif_carrier_off(ndev);
> >>
> >> This is not required, both the PHY library and the network stack will
> >> do that for you.
> > You mean netif_napi_del and netif_carrier_off? or all of sxgbe_dvr_remove function?
> > I think dma/mac control are needed.
>
> I mean netif_carrier_off() is not required, everything else is.
>
> >
> >>
> >> [snip]
> >>
> >> > + /* register with kernel subsystem */
> >> > + err = mdiobus_register(mdio_bus);
> >> > + if (err != 0) {
> >> > + netdev_err(ndev, "mdiobus register failed\n");
> >> > + goto mdiobus_err;
> >> > + }
> >> > +
> >> > + for (phy_addr = 0; phy_addr < PHY_MAX_ADDR; phy_addr++) {
> >> > + struct phy_device *phy = mdio_bus->phy_map[phy_addr];
> >> > + if (phy) {
> >> > + char irq_num[4];
> >> > + char *irq_str;
> >> > + /* If an IRQ was provided to be assigned after
> >> > + * the bus probe, do it here.
> >> > + */
> >> > + if ((mdio_data->irqs == NULL) &&
> >> > + (mdio_data->probed_phy_irq > 0)) {
> >> > + irqlist[phy_addr] = mdio_data->probed_phy_irq;
> >> > + phy->irq = mdio_data->probed_phy_irq;
> >> > + }
> >> > +
> >> > + /* If we're going to bind the MAC to this PHY bus,
> >> > + * and no PHY number was provided to the MAC,
> >> > + * use the one probed here.
> >> > + */
> >> > + if (priv->plat->phy_addr == -1)
> >> > + priv->plat->phy_addr = phy_addr;
> >> > +
> >> > + act = (priv->plat->phy_addr == phy_addr);
> >> > + switch (phy->irq) {
> >> > + case PHY_POLL:
> >> > + irq_str = "POLL";
> >> > + break;
> >> > + case PHY_IGNORE_INTERRUPT:
> >> > + irq_str = "IGNORE";
> >> > + break;
> >> > + default:
> >> > + sprintf(irq_num, "%d", phy->irq);
> >> > + irq_str = irq_num;
> >> > + break;
> >> > + }
> >> > + netdev_info(ndev, "PHY ID %08x at %d IRQ %s (%s)%s\n",
> >> > + phy->phy_id, phy_addr, irq_str,
> >> > + dev_name(&phy->dev), act ? " active" : "");
> >> > + phy_found = 1;
> >> > + }
> >>
> >> This is clunky, and you are duplicating what of_mdiobus_register()
> >> will do for you, but since you are not using the standard Ethernet
> >> PHY device tree node, you could not use that code. So please use the
> >> standard Ethernet PHY device tree node such that you can remove that code and leverage it instead.
> > I'll consider it. but I think after this patch set.
>
> Humm, I would rather you do this know this will eliminate all this code in your driver which duplicates an existing function in the kernel.
> --
> Florian
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@xxxxxxxxxxxxxxx More
> majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
