[RFC PATCH 12/12] net: sched: pfifo_fast new option to deque multiple pkts

[John Fastabend]

Now that pfifo_fast is using the alf_queue data structures we can
dequeue multiple skbs and save some overhead.

This works because the bulk dequeue logic accepts skb lists already.

Signed-off-by: John Fastabend <john.r.fastab...@intel.com>
---
 include/net/sch_generic.h |    2 +-
 net/sched/sch_generic.c   |   30 ++++++++++++++++++++----------
 2 files changed, 21 insertions(+), 11 deletions(-)

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 2c57278..95c11ed 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -128,7 +128,7 @@ static inline void qdisc_run_end(struct Qdisc *qdisc)
 
 static inline bool qdisc_may_bulk(const struct Qdisc *qdisc)
 {
-       return qdisc->flags & TCQ_F_ONETXQUEUE;
+       return (qdisc->flags & TCQ_F_ONETXQUEUE) & !(qdisc->flags & 
TCQ_F_NOLOCK);
 }
 
 static inline int qdisc_avail_bulklimit(const struct netdev_queue *txq)
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 480cf63..ec5e78e 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -507,25 +507,35 @@ static int pfifo_fast_enqueue(struct sk_buff *skb, struct 
Qdisc *qdisc)
 static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
 {
        struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
-       struct sk_buff *skb = NULL;
-       int band;
+       struct sk_buff *skb[8+1] = {NULL};
+       int band, i, elems = 0;
 
-       for (band = 0; band < PFIFO_FAST_BANDS && !skb; band++) {
+       if (this_cpu_ptr(qdisc->cpu_qstats)->qlen < 8)
+               return NULL;
+
+       for (band = 0; band < PFIFO_FAST_BANDS && !skb[0]; band++) {
                struct alf_queue *q = band2list(priv, band);
 
                if (alf_queue_empty(q))
                        continue;
 
-               alf_mc_dequeue(q, &skb, 1);
+               elems = alf_mc_dequeue(q, skb, 8);
+
+               /* link array of skbs for driver to process */
+               for (i = 0; i < elems; i++)
+                       skb[i]->next = skb[i+1];
        }
 
-       if (likely(skb)) {
-               qdisc_qstats_cpu_backlog_dec(qdisc, skb);
-               qdisc_bstats_cpu_update(qdisc, skb);
-               qdisc_qstats_cpu_qlen_dec(qdisc);
+       if (likely(skb[0])) {
+               for (i = 0; i < elems; i++) {
+                       qdisc_qstats_cpu_backlog_dec(qdisc, skb[i]);
+                       qdisc_bstats_cpu_update(qdisc, skb[i]);
+               }
+
+               this_cpu_ptr(qdisc->cpu_qstats)->qlen -= elems;
        }
 
-       return skb;
+       return skb[0];
 }
 
 static void pfifo_fast_reset(struct Qdisc *qdisc)
@@ -579,7 +589,7 @@ static int pfifo_fast_init(struct Qdisc *qdisc, struct 
nlattr *opt)
        }
 
        /* Can by-pass the queue discipline */
-       qdisc->flags |= TCQ_F_CAN_BYPASS;
+       //qdisc->flags |= TCQ_F_CAN_BYPASS;
        qdisc->flags |= TCQ_F_NOLOCK;
        qdisc->flags |= TCQ_F_CPUSTATS;
 

--
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