Re: [PATCH bpf-next 2/2] netkit: use netlink policy for mode and policy attributes validation

[Nikolay Aleksandrov <razor@blackwall.org>]

On 10/26/23 17:23, Nikolay Aleksandrov wrote:
> On 10/26/23 17:11, Ido Schimmel wrote:
>> On Thu, Oct 26, 2023 at 12:41:06PM +0300, Nikolay Aleksandrov wrote:
>>>   static const struct nla_policy netkit_policy[IFLA_NETKIT_MAX + 1] = {
>>>       [IFLA_NETKIT_PEER_INFO]        = { .len = sizeof(struct 
>>> ifinfomsg) },
>>> -    [IFLA_NETKIT_POLICY]        = { .type = NLA_U32 },
>>> -    [IFLA_NETKIT_MODE]        = { .type = NLA_U32 },
>>> -    [IFLA_NETKIT_PEER_POLICY]    = { .type = NLA_U32 },
>>> +    [IFLA_NETKIT_POLICY]        = NLA_POLICY_VALIDATE_FN(NLA_U32,
>>> +                                 netkit_check_policy),
>>
>> Nik, it's problematic to use NLA_POLICY_VALIDATE_FN() with anything
>> other than NLA_BINARY. See commit 9e17f99220d1 ("net/sched: act_mpls:
>> Fix warning during failed attribute validation").
>>
> 
> But how is that code called at all? The validation type is 
> NLA_VALIDATE_FUNCTION(), not NLA_VALIDATE_MIN/MAX/RANGE/RANGE_WARN...
> nla_validate_int_range() is called only on:
>          case NLA_VALIDATE_RANGE_PTR:
>          case NLA_VALIDATE_RANGE:
>          case NLA_VALIDATE_RANGE_WARN_TOO_LONG:
>          case NLA_VALIDATE_MIN:
>          case NLA_VALIDATE_MAX:
> 

Ah, I'm looking at the wrong thing.. I saw the problem. :)

> Anyway, I'll switch to NLA_BINARY in a bit to make sure it's ok. Thanks 
> for the pointer.
> 
>>> +    [IFLA_NETKIT_MODE]        = NLA_POLICY_VALIDATE_FN(NLA_U32,
>>> +                                 netkit_check_mode),
>>> +    [IFLA_NETKIT_PEER_POLICY]    = NLA_POLICY_VALIDATE_FN(NLA_U32,
>>> +                                 netkit_check_policy),
>>>       [IFLA_NETKIT_PRIMARY]        = { .type = NLA_REJECT,
>>>                           .reject_message = "Primary attribute is 
>>> read-only" },
>>>   };
>>> -- 
>>> 2.38.1
>>>
>>>
>