Re: net: Initialize return value in gro_cells_receive

[Gregory Rose <gvrose8192@gmail.com>]

On 10/7/2020 9:37 AM, Eric Dumazet wrote:
> 
> 
> On 10/7/20 5:50 PM, Gregory Rose wrote:
>>
>>
>> On 10/7/2020 1:21 AM, Eric Dumazet wrote:
>>>
>>>
>>> On 10/6/20 8:53 PM, Gregory Rose wrote:
>>>> The 'res' return value is uninitalized and may be returned with
>>>> some random value.  Initialize to NET_RX_DROP as the default
>>>> return value.
>>>>
>>>> Signed-off-by: Greg Rose <gvrose8192@gmail.com>
>>>>
>>>> diff --git a/net/core/gro_cells.c b/net/core/gro_cells.c
>>>> index e095fb871d91..4e835960db07 100644
>>>> --- a/net/core/gro_cells.c
>>>> +++ b/net/core/gro_cells.c
>>>> @@ -13,7 +13,7 @@ int gro_cells_receive(struct gro_cells *gcells, struct sk_buff *skb)
>>>>    {
>>>>           struct net_device *dev = skb->dev;
>>>>           struct gro_cell *cell;
>>>> -       int res;
>>>> +       int res = NET_RX_DROP;
>>>>
>>>>           rcu_read_lock();
>>>>           if (unlikely(!(dev->flags & IFF_UP)))
>>>
>>> I do not think this is needed.
>>>
>>> Also, when/if sending a patch fixing a bug, we require a Fixes: tag.
>>>
>>> Thanks.
>>>
>> If it's not needed then feel free to ignore it.  It just looked like
>> the unlikely case returns without setting the return value.
> 
> Can you elaborate ? I do not see this problem in current upstream code.
> 
> If a compiler gave you a warning, please give its version, thanks.
> 

No, it's my misreading of the code - it jumps to the drop that is in the
middle of an if statement, sets res to NET_RX_DROP there and then jumps 
to the unlock label.

My apologies.

- Greg