re: af_mpls: fix undefined reference to ip6_route_output

re: af_mpls: fix undefined reference to ip6_route_output

[Dan Carpenter]

Hello Roopa,

I have a concern about patch bf21563acc1d: "af_mpls: fix undefined
reference to ip6_route_output" from Jul 30, 2015.

net/mpls/af_mpls.c
   450  
   451          dev = find_outdev(net, cfg);
   452          if (IS_ERR(dev)) {

find_outdev() used to return NULL pointers but now it only returns NULL
if cfg->rc_via_table == NEIGH_LINK_TABLE or dev_get_by_index() fails.
I think it could lead to a NULL dereference and we don't check for that
here.

Returning a mix of error pointers and NULL is bad style, it needs a big
comment at the top of the function if it's deliberate.

   453                  err = PTR_ERR(dev);
   454                  dev = NULL;
   455                  goto errout;
   456          }

regards,
dan carpenter

Re: af_mpls: fix undefined reference to ip6_route_output

[roopa]

On 8/4/15, 12:44 AM, Dan Carpenter wrote:
> Hello Roopa,
>
> I have a concern about patch bf21563acc1d: "af_mpls: fix undefined
> reference to ip6_route_output" from Jul 30, 2015.
>
> net/mpls/af_mpls.c
>     450
>     451          dev = find_outdev(net, cfg);
>     452          if (IS_ERR(dev)) {
>
> find_outdev() used to return NULL pointers but now it only returns NULL
> if cfg->rc_via_table == NEIGH_LINK_TABLE or dev_get_by_index() fails.
> I think it could lead to a NULL dereference and we don't check for that
> here.
>
> Returning a mix of error pointers and NULL is bad style, it needs a big
> comment at the top of the function if it's deliberate.
>
>     453                  err = PTR_ERR(dev);
>     454                  dev = NULL;
>     455                  goto errout;
>     456          }
>
>
You are right. The cfg->rc_via_table == NEIGH_LINK_TABLE or 
dev_get_by_index() case, needs a ERR_PTR(-ENODEV).
I will get a patch out shortly.

thanks. I just submitted a patch.