On Mon, Nov 2, 2015 at 1:20 PM, Eric Dumazet <eric.duma...@gmail.com> wrote: > > Well, before your changes, the qdisc_tree_decrease_qlen(sch, 0) would > have been useless, since qdisc_tree_decrease_qlen() does nothing in this > case [1] > > But after your changes, we need to change the backlog, even if the qlen > does not change. > > Maybe I missed something, but your patch is not really obvious ;) >
Oh, very good point... You actually mean in some cases we enqueue one packet and then drop one packet which ends up with dropping 0 packet but the backlog size goes down by (drop_len - enqueue_len). :) Hmm, I need to double check for this case for sure, but for sfq case, I just did: - qdisc_tree_decrease_qlen(sch, 1); + qdisc_tree_reduce_backlog(sch, 1, dropped); which isn't the case where n==0, like you mentioned. I am auditing that case for all the qdisc's anyway. Thanks. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
