On 14/01/2020 18:49, David Ahern wrote:
> On 1/14/20 9:45 AM, Nikolay Aleksandrov wrote:
>> On 14/01/2020 18:36, Nikolay Aleksandrov wrote:
>>> On 14/01/2020 17:34, David Ahern wrote:
>>>> On 1/14/20 6:55 AM, Jakub Kicinski wrote:
>>>>> On Mon, 13 Jan 2020 17:52:28 +0200, Nikolay Aleksandrov wrote:
>>>>>> +static int br_vlan_rtm_dump(struct sk_buff *skb, struct netlink_callback *cb)
>>>>>> +{
>>>>>> + int idx = 0, err = 0, s_idx = cb->args[0];
>>>>>> + struct net *net = sock_net(skb->sk);
>>>>>> + struct br_vlan_msg *bvm;
>>>>>> + struct net_device *dev;
>>>>>> +
>>>>>> + if (cb->nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bvm))) {
>>>>>
>>>>> I wonder if it'd be useful to make this a strict != check? At least
>>>>> when strict validation is on? Perhaps we'll one day want to extend
>>>>> the request?
>>>>>
>>>>
>>>> +1. All new code should be using the strict checks.
>>>>
>>>
>>> IIRC, I did it to be able to add filter attributes later, but it should just use nlmsg_parse()
>>> instead and all will be taken care of.
>>> I'll respin v2 with that change.
>>>
>>> Thanks,
>>> Nik
>>>
>>
>> Actually nlmsg_parse() uses the same "<" check for the size before parsing. :)
>> If I change to it and with no attributes to parse would be essentially equal to the
>> current situation, but if I make it strict "!=" then we won't be able to add
>> filter attributes later as we won't be backwards compatible. I'll continue looking
>> into it, but IMO we should leave it as it is in order to be able to add the filtering later.
>>
>> Thoughts ?
>>
>>
>>
>>
>
> If the header is > sizeof(*bvm) I expect this part of
> __nla_validate_parse() to kick in:
>
> if (unlikely(rem > 0)) {
> pr_warn_ratelimited("netlink: %d bytes leftover after
> parsing attributes in process `%s'.\n",
> rem, current->comm);
> NL_SET_ERR_MSG(extack, "bytes leftover after parsing
> attributes");
> if (validate & NL_VALIDATE_TRAILING)
> return -EINVAL;
> }
>
Ah fair enough, so nlmsg_parse() would be better even without attrs.
