RE: [PATCH net-next 1/4] drivers: net: xgene-v2: Add MDIO support

[Keyur Chudgar <kchudgar@apm.com>]

Hi Iyappan,

I agree with Andrew.

Let's re-post the patch after addressing Andrew's comment.

Thanks,
Keyur

> -----Original Message-----
> From: Iyappan Subramanian [mailto:isubramanian@apm.com]
> Sent: Tuesday, March 21, 2017 4:19 PM
> To: Andrew Lunn
> Cc: David Miller; netdev; Florian Fainelli; David Laight; linux-arm-
> kernel@lists.infradead.org; patches; Keyur Chudgar
> Subject: Re: [PATCH net-next 1/4] drivers: net: xgene-v2: Add MDIO
> support
>
> On Tue, Mar 21, 2017 at 1:35 PM, Andrew Lunn <andrew@lunn.ch> wrote:
> >> @@ -511,9 +512,9 @@ static int xge_close(struct net_device *ndev)  {
> >>       struct xge_pdata *pdata = netdev_priv(ndev);
> >>
> >> -     netif_carrier_off(ndev);
> >>       netif_stop_queue(ndev);
> >>       xge_mac_disable(pdata);
> >> +     phy_stop(ndev->phydev);
> >>
> >>       xge_intr_disable(pdata);
> >>       xge_free_irq(ndev);
> >> @@ -683,9 +684,14 @@ static int xge_probe(struct platform_device
> *pdev)
> >>       if (ret)
> >>               goto err;
> >>
> >> +     spin_lock_init(&pdata->mdio_lock);
> >> +
> >
> > ...
> >
> >> +static int xge_mdio_write(struct mii_bus *bus, int phy_id, int reg,
> >> +u16 data) {
> >> +     struct xge_pdata *pdata = bus->priv;
> >> +     u32 done, val = 0;
> >> +     u8 wait = 10;
> >> +     int ret = 0;
> >> +
> >> +     spin_lock(&pdata->mdio_lock);
> >> +
> >> +     SET_REG_BITS(&val, PHY_ADDR, phy_id);
> >> +     SET_REG_BITS(&val, REG_ADDR, reg);
> >> +     xge_wr_csr(pdata, MII_MGMT_ADDRESS, val);
> >> +
> >> +     xge_wr_csr(pdata, MII_MGMT_CONTROL, data);
> >> +     do {
> >> +             usleep_range(5, 10);
> >> +             done = xge_rd_csr(pdata, MII_MGMT_INDICATORS);
> >> +     } while ((done & MII_MGMT_BUSY) && wait--);
> >> +
> >> +     if (done & MII_MGMT_BUSY) {
> >> +             dev_err(&bus->dev, "MII_MGMT write failed\n");
> >> +             ret = -ETIMEDOUT;
> >> +     }
> >> +
> >> +     spin_unlock(&pdata->mdio_lock);
> >> +
> >> +     return ret;
> >> +}
> >> +
> >> +static int xge_mdio_read(struct mii_bus *bus, int phy_id, int reg)
> {
> >> +     struct xge_pdata *pdata = bus->priv;
> >> +     u32 data, done, val = 0;
> >> +     u8 wait = 10;
> >> +
> >> +     spin_lock(&pdata->mdio_lock);
> >> +
> >
> > Hi Iyappan
> >
> > Please could you explain what this lock is protecting which the
> > mii_bus mdio_lock in mdio_bus.c is not protecting?
>
> Hi Keyur,
>
> Please could you explain what this lock is protecting which the mii_bus
> mdio_lock in mdio_bus.c is not protecting?
>
> I agree with him.  Actually there is a mutex on mdio_bus.  So the mdio
> bus read and write are locked.  we don't need the lock.
>
> Do you agree ?
>
>
> >
> > Thanks
> >         Andrew