Re: bonding: fix device leak on error in bond_create()

[Andy Gospodarek <andy@greyhouse.net>]

On Fri, Feb 26, 2010 at 05:17:40PM +0100, Patrick McHardy wrote:
> Andy Gospodarek wrote:
> > On Fri, Feb 26, 2010 at 04:35:52PM +0100, Patrick McHardy wrote:
> > 
> >> commit f357a5caa0ca29b74811a0fc08efb5ae4aade959
> >> Author: Patrick McHardy <kaber@trash.net>
> >> Date:   Thu Feb 25 20:21:10 2010 +0100
> >>
> >>     bonding: fix device leak on error in bond_create()
> >>     
> >>     When the register_netdevice() call fails, the newly allocated device is
> >>     not freed.
> >>     
> >>     Signed-off-by: Patrick McHardy <kaber@trash.net>
> >>
> >> diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> >> index 1787e3c..430c022 100644
> >> --- a/drivers/net/bonding/bond_main.c
> >> +++ b/drivers/net/bonding/bond_main.c
> >> @@ -4946,6 +4946,8 @@ int bond_create(struct net *net, const char *name)
> >>  	}
> >>  
> >>  	res = register_netdevice(bond_dev);
> >> +	if (res < 0)
> >> +		goto out_netdev;
> >>  
> >>  out:
> >>  	rtnl_unlock();
> > 
> > Acked-by: Andy Gospodarek <andy@greyhouse.net>
> > 
> > Seems like a good idea.  I know register_netdevice is always supposed to
> > return a value < 0 on error, but this would work if it didn't for some
> > reason.
> 
> I don't care much, but that would be a bug since we'd return
> invalid errno codes to userspace.

Then I guess the bonding driver would probably not be the last to know
if someone snuck a bug into register_netdevice....