On Fri, Sep 06, 2024 at 06:06:14AM +0000, Jacky Chou wrote:
> Hello,
>
> >
> > We might not have a phy so we need to check for NULL before calling
> > phy_stop(netdev->phydev) or it could lead to an Oops.
> >
> > Fixes: e24a6c874601 ("net: ftgmac100: Get link speed and duplex for NC-SI")
> > Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > ---
> > drivers/net/ethernet/faraday/ftgmac100.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/faraday/ftgmac100.c
> > b/drivers/net/ethernet/faraday/ftgmac100.c
> > index f3cc14cc757d..0e873e6f60d6 100644
> > --- a/drivers/net/ethernet/faraday/ftgmac100.c
> > +++ b/drivers/net/ethernet/faraday/ftgmac100.c
> > @@ -1565,7 +1565,8 @@ static int ftgmac100_open(struct net_device
> > *netdev)
> > return 0;
> >
> > err_ncsi:
> > - phy_stop(netdev->phydev);
> > + if (netdev->phydev)
> > + phy_stop(netdev->phydev);
> When using " use-ncsi" property, the driver will register a fixed-link phy device and
> bind to netdev at probe stage.
>
> if (np && of_get_property(np, "use-ncsi", NULL)) {
>
> ......
>
> phydev = fixed_phy_register(PHY_POLL, &ncsi_phy_status, NULL);
> err = phy_connect_direct(netdev, phydev, ftgmac100_adjust_link,
> PHY_INTERFACE_MODE_MII);
This is another bug. There needs to be error checking in case fixed_phy_register()
fails, other wise it crashes when we call phy_connect_direct(). For example,
if the probe() ordering is unlucky fixed_phy_register() can return -EPROBE_DEFER
so it's not even unusual error cases, which can lead to a crash but just normal
stuff.
> if (err) {
> dev_err(&pdev->dev, "Connecting PHY failed\n");
> goto err_phy_connect;
> }
> } else if (np && of_phy_is_fixed_link(np)) {
>
> Therefore, it does not need to check if the point is NULL in this error handling.
> Thanks.
It's really unsafe to assume that we will never add more gotos to the
ftgmac100_open() function. If you insist, I could remove the Fixes tag... Let
me know.
regards,
dan carpenter
