On Sat, May 4, 2019 at 8:41 AM Eric Dumazet <eric.duma...@gmail.com> wrote:
>
>
>
> On 5/2/19 2:06 PM, Cong Wang wrote:
> > In commit 3c75f6ee139d ("net_sched: sch_htb: add per class overlimits
> > counter")
> > we added an overlimits counter for each HTB class which could
> > properly reflect how many times we use up all the bandwidth
> > on each class. However, the overlimits counter in HTB qdisc
> > does not, it is way bigger than the sum of each HTB class.
> > In fact, this qdisc overlimits counter increases when we have
> > no skb to dequeue, which happens more often than we run out of
> > bandwidth.
> >
> > It makes more sense to make this qdisc overlimits counter just
> > be a sum of each HTB class, in case people still get confused.
> >
> > I have verified this patch with one single HTB class, where HTB
> > qdisc counters now always match HTB class counters as expected.
> >
> > Cc: Eric Dumazet <eduma...@google.com>
> > Signed-off-by: Cong Wang <xiyou.wangc...@gmail.com>
> > ---
> > net/sched/sch_htb.c | 7 +++++--
> > 1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
> > index d27d9bc9d010..cece0d455985 100644
> > --- a/net/sched/sch_htb.c
> > +++ b/net/sched/sch_htb.c
> > @@ -177,6 +177,7 @@ struct htb_sched {
> > int row_mask[TC_HTB_MAXDEPTH];
> >
> > struct htb_level hlevel[TC_HTB_MAXDEPTH];
> > + u32 overlimits;
> > };
>
>
> Hi Cong, your patch makes sense, but I am not sure about the location of this
> field,
> as this might need another cache line miss.
>
> Maybe instead reduce 'long direct_pkts;' to 'u32 direct_pkts;'
> since we only export
> direct_pkts as 32bit quantity in the struct tc_htb_glob.
>
> Then move the 'u32 overlimits;' right after direct_pkts to fill the new hole.
Sure, v2 is coming. :)
Thanks for catching it.
