Re: [PATCH] sb1000: prevent a potential NULL pointer dereference in sb1000_dev_ioctl()

[Satyam Sharma <satyam@infradead.org>]

Hi Michael,


On Sun, 29 Jul 2007, Michael Buesch wrote:

> On Sunday 29 July 2007 20:34:46 Satyam Sharma wrote:
> > (2) !(dev->flags & IFF_UP) is bogus because the functions of this ioctl
> > can (and should) be allowed even when the interface is not up and running.
> 
> Are you _sure_? This function does poke with the device hardware.
> It might return crap or even machinecheck when not initialized.
> Hardware is probably powered down, if not IFF_UP. (I don't know if that's
> the case here, though).

IFF_UP checks if the _interface_ is up -- the hardware / card could still
be powered up, but the interface down (ifconfing eth0 down or ip link set
eth0 down).

Probably what we want here is netif_device_present()? -- I think that
should return true only when the *device* itself is up (as in powered)
but the interface itself could be down ...

Let's wait for comments from the netdev people Cc:'ed here, in that case.


> >  drivers/net/sb1000.c |    3 ---
> >  1 files changed, 0 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/net/sb1000.c b/drivers/net/sb1000.c
> > index 1de3eec..f60fe98 100644
> > --- a/drivers/net/sb1000.c
> > +++ b/drivers/net/sb1000.c
> > @@ -993,9 +993,6 @@ static int sb1000_dev_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
> >  	unsigned int stats[5];
> >  	struct sb1000_private *lp = netdev_priv(dev);
> >  
> > -	if (!(dev && dev->flags & IFF_UP))
> > -		return -ENODEV;
> > -


Satyam