linux/net
Krishna Kumar bbd8a0d3a3 net: Avoid enqueuing skb for default qdiscs
dev_queue_xmit enqueue's a skb and calls qdisc_run which
dequeue's the skb and xmits it. In most cases, the skb that
is enqueue'd is the same one that is dequeue'd (unless the
queue gets stopped or multiple cpu's write to the same queue
and ends in a race with qdisc_run). For default qdiscs, we
can remove the redundant enqueue/dequeue and simply xmit the
skb since the default qdisc is work-conserving.

The patch uses a new flag - TCQ_F_CAN_BYPASS to identify the
default fast queue. The controversial part of the patch is
incrementing qlen when a skb is requeued - this is to avoid
checks like the second line below:

+  } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
>>         !q->gso_skb &&
+          !test_and_set_bit(__QDISC_STATE_RUNNING, &q->state)) {

Results of a 2 hour testing for multiple netperf sessions (1,
2, 4, 8, 12 sessions on a 4 cpu system-X). The BW numbers are
aggregate Mb/s across iterations tested with this version on
System-X boxes with Chelsio 10gbps cards:

----------------------------------
Size |  ORG BW          NEW BW   |
----------------------------------
128K |  156964          159381   |
256K |  158650          162042   |
----------------------------------

Changes from ver1:

1. Move sch_direct_xmit declaration from sch_generic.h to
   pkt_sched.h
2. Update qdisc basic statistics for direct xmit path.
3. Set qlen to zero in qdisc_reset.
4. Changed some function names to more meaningful ones.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-08-06 20:10:18 -07:00
..
9p net/9p: Fix crash due to bad mount parameters. 2009-07-02 13:17:01 -07:00
802 net: remove COMPAT_NET_DEV_OPS 2009-05-25 01:53:53 -07:00
8021q net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
appletalk net: correct off-by-one write allocations reports 2009-06-18 00:29:12 -07:00
atm net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
ax25 net: Move rx skb_orphan call to where needed 2009-06-23 16:36:25 -07:00
bluetooth net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
bridge net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
can net/can: add module alias to can protocol drivers 2009-07-15 11:20:38 -07:00
core net: Avoid enqueuing skb for default qdiscs 2009-08-06 20:10:18 -07:00
dcb DCB: fix kfree(skb) 2009-01-04 17:29:21 -08:00
dccp net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
decnet net: remove NET_RX_BAD and NET_RX_CN* defines 2009-07-05 19:15:35 -07:00
dsa dsa: fix 88e6xxx statistics counter snapshotting 2009-07-05 18:03:35 -07:00
econet econet: use NET_RX_SUCCESS instead of magic number 0 for econet_rcv successful return 2009-07-06 18:07:59 -07:00
ethernet net: remove COMPAT_NET_DEV_OPS 2009-05-25 01:53:53 -07:00
ieee802154 ieee802154: move headers out of extra directory 2009-07-23 17:08:51 +04:00
ipv4 net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
ipv6 net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
ipx net: correct off-by-one write allocations reports 2009-06-18 00:29:12 -07:00
irda net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
iucv net: adding memory barrier to the poll and receive callbacks 2009-07-09 17:06:57 -07:00
key net: correct off-by-one write allocations reports 2009-06-18 00:29:12 -07:00
lapb net: remove NET_RX_BAD and NET_RX_CN* defines 2009-07-05 19:15:35 -07:00
llc net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
mac80211 mac80211: FIF_PSPOLL filter flag 2009-08-04 16:44:35 -04:00
netfilter net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
netlabel Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6 2009-07-30 19:22:43 -07:00
netlink net/compat/wext: send different messages to compat tasks 2009-07-15 08:53:39 -07:00
netrom net: use NETDEV_TX_OK instead of 0 in ndo_start_xmit() functions 2009-07-05 19:16:04 -07:00
packet af_packet: style cleanups 2009-07-23 18:01:10 -07:00
phonet phonet: phonet_device_get() fix 2009-07-27 08:03:18 -07:00
rds net/rds: Use AF_INET for sin_family field 2009-08-05 20:30:13 -07:00
rfkill rfkill: add the GPS radio type 2009-08-04 16:44:23 -04:00
rose Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6 2009-07-30 19:22:43 -07:00
rxrpc net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
sched net: Avoid enqueuing skb for default qdiscs 2009-08-06 20:10:18 -07:00
sctp net: mark read-only arrays as const 2009-08-05 10:42:58 -07:00
sunrpc sunrpc: Use rcu_barrier() on unload. 2009-06-26 13:51:34 -07:00
tipc genetlink: make netns aware 2009-07-12 14:03:27 -07:00
unix net: adding memory barrier to the poll and receive callbacks 2009-07-09 17:06:57 -07:00
wanrouter wanrouter: fix sparse warnings: context imbalance 2009-02-26 23:13:36 -08:00
wimax wimax: fix warning caused by not checking retval of rfkill_set_hw_state() 2009-06-11 11:12:48 -07:00
wireless cfg80211: clear SSID on disconnect/no connection 2009-08-04 16:44:26 -04:00
x25 net: correct off-by-one write allocations reports 2009-06-18 00:29:12 -07:00
xfrm xfrm: use xfrm_addr_cmp() instead of compare addresses directly 2009-06-29 19:41:46 -07:00
compat.c net/compat/wext: send different messages to compat tasks 2009-07-15 08:53:39 -07:00
Kconfig net/compat/wext: send different messages to compat tasks 2009-07-15 08:53:39 -07:00
Makefile net: remove redundant sched/ in net/Makefile 2009-07-12 20:11:14 -07:00
nonet.c
socket.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6 2009-04-06 18:05:43 -07:00
sysctl_net.c net: sysctl_net - use net_eq to compare nets 2009-03-16 16:23:30 +01:00
TUNABLE