Re: [PATCH V2 1/4] ofnet: add error check for grub_netbuff_reserve

[Andrei Borzenkov]

On Mon, Dec 12, 2016 at 1:34 PM, Stanislav Kholmanskikh
<address@hidden> wrote:
>
>
> On 12/10/2016 09:08 PM, Andrei Borzenkov wrote:
>> 02.12.2016 18:10, Stanislav Kholmanskikh пишет:
>>> Signed-off-by: Stanislav Kholmanskikh <address@hidden>
>>> ---
>>>  grub-core/net/drivers/ieee1275/ofnet.c |    6 +++++-
>>>  1 files changed, 5 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/grub-core/net/drivers/ieee1275/ofnet.c 
>>> b/grub-core/net/drivers/ieee1275/ofnet.c
>>> index 6bd3b92..8332d34 100644
>>> --- a/grub-core/net/drivers/ieee1275/ofnet.c
>>> +++ b/grub-core/net/drivers/ieee1275/ofnet.c
>>> @@ -90,7 +90,11 @@ get_card_packet (struct grub_net_card *dev)
>>>      return NULL;
>>>    /* Reserve 2 bytes so that 2 + 14/18 bytes of ethernet header is 
>>> divisible
>>>       by 4. So that IP header is aligned on 4 bytes. */
>>> -  grub_netbuff_reserve (nb, 2);
>>> +  if (grub_netbuff_reserve (nb, 2))
>>> +    {
>>> +      grub_netbuff_free (nb);
>>> +      return NULL;
>>> +    }
>>>
>>>    start_time = grub_get_time_ms ();
>>>    do
>>>
>>
>> We already account for this reserve when we allocate netbuf. So this is
>> redundant. May be short comment before grub_netbuf_alloc to explain how
>> we get at final size.
>
> So your message is that since nb = grub_netbuff_alloc(some_len + 2)
> succeeds (i.e. returns a valid buffer), then following
> grub_netbuff_reserve(nb, 2) will never fail and needs no check for
> error. Right?
>

In this place - yes.

> Personally I don't like skipping such checks, but if you insist on
> removing the check, then, I think, all other drivers should be updated
> as well, because they all have the check.
>

The code is rather inconsistent, I agree, but this is not bug fix that
is required as part of your patch series, so any cleanup belongs
elsewhere.