diff options
| author | Eric Dumazet <edumazet@google.com> | 2016-06-21 23:16:49 -0700 |
|---|---|---|
| committer | David S. Miller <davem@davemloft.net> | 2016-06-25 12:19:35 -0400 |
| commit | 520ac30f45519b0a82dd92117c181d1d6144677b (patch) | |
| tree | a3189ae1ab3b5be6213716e42ddaec082578e774 /net/sched/sch_fq_codel.c | |
| parent | 36195d869e4b1dbf81bafa3a31bb095a3c013bdd (diff) | |
| download | linux-520ac30f45519b0a82dd92117c181d1d6144677b.tar.bz2 | |
net_sched: drop packets after root qdisc lock is released
Qdisc performance suffers when packets are dropped at enqueue()
time because drops (kfree_skb()) are done while qdisc lock is held,
delaying a dequeue() draining the queue.
Nominal throughput can be reduced by 50 % when this happens,
at a time we would like the dequeue() to proceed as fast as possible.
Even FQ is vulnerable to this problem, while one of FQ goals was
to provide some flow isolation.
This patch adds a 'struct sk_buff **to_free' parameter to all
qdisc->enqueue(), and in qdisc_drop() helper.
I measured a performance increase of up to 12 %, but this patch
is a prereq so that future batches in enqueue() can fly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/sched/sch_fq_codel.c')
| -rw-r--r-- | net/sched/sch_fq_codel.c | 15 |
1 files changed, 9 insertions, 6 deletions
diff --git a/net/sched/sch_fq_codel.c b/net/sched/sch_fq_codel.c index 2dc0a849515a..f715195459c9 100644 --- a/net/sched/sch_fq_codel.c +++ b/net/sched/sch_fq_codel.c @@ -139,7 +139,8 @@ static inline void flow_queue_add(struct fq_codel_flow *flow, skb->next = NULL; } -static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets) +static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets, + struct sk_buff **to_free) { struct fq_codel_sched_data *q = qdisc_priv(sch); struct sk_buff *skb; @@ -172,7 +173,7 @@ static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets) skb = dequeue_head(flow); len += qdisc_pkt_len(skb); mem += skb->truesize; - kfree_skb(skb); + __qdisc_drop(skb, to_free); } while (++i < max_packets && len < threshold); flow->dropped += i; @@ -184,7 +185,8 @@ static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets) return idx; } -static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch) +static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch, + struct sk_buff **to_free) { struct fq_codel_sched_data *q = qdisc_priv(sch); unsigned int idx, prev_backlog, prev_qlen; @@ -197,7 +199,7 @@ static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch) if (idx == 0) { if (ret & __NET_XMIT_BYPASS) qdisc_qstats_drop(sch); - kfree_skb(skb); + __qdisc_drop(skb, to_free); return ret; } idx--; @@ -229,7 +231,7 @@ static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch) * So instead of dropping a single packet, drop half of its backlog * with a 64 packets limit to not add a too big cpu spike here. */ - ret = fq_codel_drop(sch, q->drop_batch_size); + ret = fq_codel_drop(sch, q->drop_batch_size, to_free); prev_qlen -= sch->q.qlen; prev_backlog -= sch->qstats.backlog; @@ -276,7 +278,8 @@ static void drop_func(struct sk_buff *skb, void *ctx) { struct Qdisc *sch = ctx; - qdisc_drop(skb, sch); + kfree_skb(skb); + qdisc_qstats_drop(sch); } static struct sk_buff *fq_codel_dequeue(struct Qdisc *sch) |