Re: [Patch net-next] net_sched: kill u32_node pointer in Qdisc

[Jiri Pirko <jiri@resnulli.us>]

Wed, Aug 23, 2017 at 11:31:23PM CEST, xiyou.wangcong@gmail.com wrote:
>On Wed, Aug 23, 2017 at 2:20 PM, Jiri Pirko <jiri@resnulli.us> wrote:
>> Wed, Aug 23, 2017 at 11:14:15PM CEST, xiyou.wangcong@gmail.com wrote:
>>>On Wed, Aug 23, 2017 at 1:25 PM, Jiri Pirko <jiri@resnulli.us> wrote:
>>>>>+static struct hlist_head *tc_u_common_hash;
>>>>
>>>> Why not use rhashtable?
>>>>
>>>
>>>It doesn't have to be so complicated, it is not fast path and
>>>we don't have so many qdisc's and u32 filters in system
>>>relatively.
>>
>> Well, it is easier to work with. So why not use it?
>>
>
>OMG... You must have a different definition for "easier".

Apparently.


>
>
>>
>>>
>>>
>>>>>+      tc_u_common_hash = kvmalloc_array(U32_HASH_SIZE, sizeof(struct hlist_head), GFP_KERNEL);
>>>>
>>>> This is over 80cols.
>>>>
>>>
>>>Yeah, it doesn't harm anything, but I can fix it.
>>
>> At least checkpatch warns you about it.
>
>Not every checkpatch.pl warning worth your time.
>
>Jiri, spend your time on something more value than
>80-cols. ;)

I would not have to spend any time on it, if you would just follow the
usual workflow. Clearly, you have some problem with that. I cannot
say I understand it :/