Re: [PATCH net-next] net: Drop unlikely before IS_ERR(_OR_NULL)

[Kefeng Wang]

On 2019/6/6 0:13, Jesse Brandeburg wrote:
> On Wed, 5 Jun 2019 22:24:26 +0800 Kefeng wrote:
>> IS_ERR(_OR_NULL) already contain an 'unlikely' compiler flag,
>> so no need to do that again from its callers. Drop it.
>>
> <snip>
>
>>  	segs = __skb_gso_segment(skb, features, false);
>> -	if (unlikely(IS_ERR_OR_NULL(segs))) {
>> +	if (IS_ERR_OR_NULL(segs)) {
>>  		int segs_nr = skb_shinfo(skb)->gso_segs;
>>  
> The change itself seems reasonable, but did you check to see if the
> paths changed are faster/slower with your fix?  Did you look at any
> assembly output to see if the compiler actually generated different
> code?  Is there a set of similar changes somewhere else in the kernel
> we can refer to?

+Enrico Weigelt

There is no different in assembly output (only check the x86/arm64), and

the Enrico Weigelt have finished a cocci script to do this cleanup.


>
> I'm not sure in the end that the change is worth it, so would like you
> to prove it is, unless davem overrides me. :-)
>
>
> .
>