diff options
| author | Yaogong Wang <wygivan@google.com> | 2016-09-07 23:49:28 +0200 |
|---|---|---|
| committer | David S. Miller <davem@davemloft.net> | 2016-09-09 02:25:58 +0200 |
| commit | 9f5afeae51526b3ad7b7cb21ee8b145ce6ea7a7a (patch) | |
| tree | f434343314c30020025c7a84507107c4aad60fd4 /include | |
| parent | Merge branch 'ovs-802.1ad' (diff) | |
| download | linux-9f5afeae51526b3ad7b7cb21ee8b145ce6ea7a7a.tar.xz linux-9f5afeae51526b3ad7b7cb21ee8b145ce6ea7a7a.zip | |
tcp: use an RB tree for ooo receive queue
Over the years, TCP BDP has increased by several orders of magnitude,
and some people are considering to reach the 2 Gbytes limit.
Even with current window scale limit of 14, ~1 Gbytes maps to ~740,000
MSS.
In presence of packet losses (or reorders), TCP stores incoming packets
into an out of order queue, and number of skbs sitting there waiting for
the missing packets to be received can be in the 10^5 range.
Most packets are appended to the tail of this queue, and when
packets can finally be transferred to receive queue, we scan the queue
from its head.
However, in presence of heavy losses, we might have to find an arbitrary
point in this queue, involving a linear scan for every incoming packet,
throwing away cpu caches.
This patch converts it to a RB tree, to get bounded latencies.
Yaogong wrote a preliminary patch about 2 years ago.
Eric did the rebase, added ofo_last_skb cache, polishing and tests.
Tested with network dropping between 1 and 10 % packets, with good
success (about 30 % increase of throughput in stress tests)
Next step would be to also use an RB tree for the write queue at sender
side ;)
Signed-off-by: Yaogong Wang <wygivan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-By: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'include')
| -rw-r--r-- | include/linux/skbuff.h | 2 | ||||
| -rw-r--r-- | include/linux/tcp.h | 7 | ||||
| -rw-r--r-- | include/net/tcp.h | 2 |
3 files changed, 6 insertions, 5 deletions
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index cfb7219be665..4c5662f05bda 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -2402,6 +2402,8 @@ static inline void __skb_queue_purge(struct sk_buff_head *list) kfree_skb(skb); } +void skb_rbtree_purge(struct rb_root *root); + void *netdev_alloc_frag(unsigned int fragsz); struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 7be9b1242354..c723a465125d 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -281,10 +281,9 @@ struct tcp_sock { struct sk_buff* lost_skb_hint; struct sk_buff *retransmit_skb_hint; - /* OOO segments go in this list. Note that socket lock must be held, - * as we do not use sk_buff_head lock. - */ - struct sk_buff_head out_of_order_queue; + /* OOO segments go in this rbtree. Socket lock must be held. */ + struct rb_root out_of_order_queue; + struct sk_buff *ooo_last_skb; /* cache rb_last(out_of_order_queue) */ /* SACKs data, these 2 need to be together (see tcp_options_write) */ struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ diff --git a/include/net/tcp.h b/include/net/tcp.h index d6ae36512429..fdfbedd61c67 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -640,7 +640,7 @@ static inline void tcp_fast_path_check(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); - if (skb_queue_empty(&tp->out_of_order_queue) && + if (RB_EMPTY_ROOT(&tp->out_of_order_queue) && tp->rcv_wnd && atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && !tp->urg_data) |