On Thu, Nov 30, 2017 at 03:33:32PM +0200, Petr Machata wrote:
> Dan Carpenter <dan.carpenter@xxxxxxxxxx> writes:
>
> > Hello Petr Machata,
> >
> > The patch 0063587d3587: "mlxsw: spectrum: Support decap-only IP-in-IP
> > tunnels" from Oct 16, 2017, leads to the following static checker
> > warning:
> >
> > drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c:1306 mlxsw_sp_netdevice_ipip_ol_reg_event()
> > error: uninitialized symbol 'ipipt'.
> >
> > drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
> > 1296 static int mlxsw_sp_netdevice_ipip_ol_reg_event(struct mlxsw_sp *mlxsw_sp,
> > 1297 struct net_device *ol_dev)
> > 1298 {
> > 1299 struct mlxsw_sp_ipip_entry *ipip_entry;
> > 1300 enum mlxsw_sp_l3proto ul_proto;
> > 1301 enum mlxsw_sp_ipip_type ipipt;
> > 1302 union mlxsw_sp_l3addr saddr;
> > 1303 u32 ul_tb_id;
> > 1304
> > 1305 mlxsw_sp_netdev_ipip_type(mlxsw_sp, ol_dev, &ipipt);
> > ^^^^^
> > Not initialized on error path.
>
> That's a "can't happen" scenario. That's the reason I'm not checking
> return code from mlxsw_sp_netdev_ipip_type--we won't get there unless
> it's safe to call it.
>
> How should this be handled? I can make it a WARN_ON type of thing, but
> my impression was that function contracts are not checked like this in
> the kernel. (E.g. mlxsw_sp is not checked for not-NULLness.)
Ah. I see how it's checked now in mlxsw_sp_netdevice_event(). Just
ignore the warning. I'm pretty sure that eventually I will be able to
teach Smatch to follow that code.
regards,
dan carpenter
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
