On 9/25/19 5:25 AM, Daniel Borkmann wrote:
> On Wed, Sep 25, 2019 at 10:40:43AM +0800, xiaolinkui wrote:
>> This is an unlikely case, use unlikely() on it seems logical.
>>
>> Signed-off-by: xiaolinkui <xiaolin...@kylinos.cn>
>
> It's already here [0], but should probably rather get reverted instead
> due to lack of a more elaborate reasoning on why it needs to be done
> this way instead of letting compiler do it's job in this case. "Seems
> logical" is never a good technical explanation. Do you have any better
> analysis you performed prior to submitting the patch (twice by now)?
>
Yes, we need more details here.
We could probably save more cpu cycles checking if we can move
the smb_mb() after the dql_avail() check :)
> Thanks,
> Daniel
>
> [0]
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=f3acd33d840d3ea3e1233d234605c85cbbf26054
>
>> ---
>> include/linux/netdevice.h | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
>> index 88292953aa6f..005f3da1b13d 100644
>> --- a/include/linux/netdevice.h
>> +++ b/include/linux/netdevice.h
>> @@ -3270,7 +3270,7 @@ static inline void netdev_tx_completed_queue(struct
>> netdev_queue *dev_queue,
>> */
>> smp_mb();
>>
>> - if (dql_avail(&dev_queue->dql) < 0)
>> + if (unlikely(dql_avail(&dev_queue->dql) < 0))
>> return;
>>
>> if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
>> --
>> 2.17.1
>>
>>
>>
