linux/net/core
Tom Herbert 1d24eb4815 xps: Transmit Packet Steering
This patch implements transmit packet steering (XPS) for multiqueue
devices.  XPS selects a transmit queue during packet transmission based
on configuration.  This is done by mapping the CPU transmitting the
packet to a queue.  This is the transmit side analogue to RPS-- where
RPS is selecting a CPU based on receive queue, XPS selects a queue
based on the CPU (previously there was an XPS patch from Eric
Dumazet, but that might more appropriately be called transmit completion
steering).

Each transmit queue can be associated with a number of CPUs which will
use the queue to send packets.  This is configured as a CPU mask on a
per queue basis in:

/sys/class/net/eth<n>/queues/tx-<n>/xps_cpus

The mappings are stored per device in an inverted data structure that
maps CPUs to queues.  In the netdevice structure this is an array of
num_possible_cpu structures where each structure holds and array of
queue_indexes for queues which that CPU can use.

The benefits of XPS are improved locality in the per queue data
structures.  Also, transmit completions are more likely to be done
nearer to the sending thread, so this should promote locality back
to the socket on free (e.g. UDP).  The benefits of XPS are dependent on
cache hierarchy, application load, and other factors.  XPS would
nominally be configured so that a queue would only be shared by CPUs
which are sharing a cache, the degenerative configuration woud be that
each CPU has it's own queue.

Below are some benchmark results which show the potential benfit of
this patch.  The netperf test has 500 instances of netperf TCP_RR test
with 1 byte req. and resp.

bnx2x on 16 core AMD
   XPS (16 queues, 1 TX queue per CPU)  1234K at 100% CPU
   No XPS (16 queues)                   996K at 100% CPU

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-24 11:44:20 -08:00
..
datagram.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6 2010-10-23 11:47:02 -07:00
dev_addr_lists.c net: include linux/proc_fs.h in dev_addr_lists.c 2010-04-07 16:46:36 -07:00
dev.c xps: Transmit Packet Steering 2010-11-24 11:44:20 -08:00
drop_monitor.c drop_monitor: use genl_register_family_with_ops() 2010-07-26 20:59:42 -07:00
dst.c net/dst: dst_dev_event() called after other notifiers 2010-11-09 12:17:16 -08:00
ethtool.c ethtool: Add support for vlan accleration. 2010-10-21 01:26:54 -07:00
fib_rules.c fib_rules: __rcu annotates ctarget 2010-10-27 11:37:32 -07:00
filter.c Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6 2010-11-19 13:13:47 -08:00
flow.c net: return operator cleanup 2010-09-23 14:33:39 -07:00
gen_estimator.c pkt_sched: remov unnecessary bh_disable 2010-09-10 12:47:59 -07:00
gen_stats.c net/core: EXPORT_SYMBOL cleanups 2010-07-12 12:57:55 -07:00
iovec.c net: Limit socket I/O iovec total length to INT_MAX. 2010-10-28 11:47:52 -07:00
kmap_skb.h
link_watch.c net/core: EXPORT_SYMBOL cleanups 2010-07-12 12:57:55 -07:00
Makefile net: support time stamping in phy devices. 2010-07-18 19:15:26 -07:00
neighbour.c net/neighbour: cancel_delayed_work() + flush_scheduled_work() -> cancel_delayed_work_sync() 2010-10-21 04:25:48 -07:00
net_namespace.c net_ns: add __rcu annotations 2010-10-25 14:18:27 -07:00
net-sysfs.c xps: Transmit Packet Steering 2010-11-24 11:44:20 -08:00
net-sysfs.h xps: Transmit Packet Steering 2010-11-24 11:44:20 -08:00
net-traces.c netdev: Add tracepoints to netdev layer 2010-09-07 17:51:33 +02:00
netevent.c net/core: EXPORT_SYMBOL cleanups 2010-07-12 12:57:55 -07:00
netpoll.c netpoll: Revert napi_poll fix for bonding driver 2010-10-20 01:44:30 -07:00
pktgen.c pktgen: allow faster module unload 2010-11-21 10:26:44 -08:00
request_sock.c
rtnetlink.c rtnetlink: Link address family API 2010-11-17 11:28:24 -08:00
scm.c scm: lower SCM_MAX_FD 2010-11-24 11:16:43 -08:00
skbuff.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6 2010-10-23 11:47:02 -07:00
sock.c net: avoid limits overflow 2010-11-10 12:12:00 -08:00
stream.c net: Fix the condition passed to sk_wait_event() 2010-10-03 20:41:32 -07:00
sysctl_net_core.c rps: add __rcu annotations 2010-10-25 14:18:27 -07:00
timestamping.c filter: optimize sk_run_filter 2010-11-19 09:49:59 -08:00
user_dma.c net/core/user_dma.c: Use frag list abstraction interfaces. 2009-06-09 00:19:10 -07:00
utils.c net: return operator cleanup 2010-09-23 14:33:39 -07:00