linux/net
Eric Dumazet c8628155ec tcp: reduce out_of_order memory use
With increasing receive window sizes, but speed of light not improved
that much, out of order queue can contain a huge number of skbs, waiting
to be moved to receive_queue when missing packets can fill the holes.

Some devices happen to use fat skbs (truesize of 4096 + sizeof(struct
sk_buff)) to store regular (MTU <= 1500) frames. This makes highly
probable sk_rmem_alloc hits sk_rcvbuf limit, which can be 4Mbytes in
many cases.

When limit is hit, tcp stack calls tcp_collapse_ofo_queue(), a true
latency killer and cpu cache blower.

Doing the coalescing attempt each time we add a frame in ofo queue
permits to keep memory use tight and in many cases avoid the
tcp_collapse() thing later.

Tested on various wireless setups (b43, ath9k, ...) known to use big skb
truesize, this patch removed the "packets collapsed in receive queue due
to low socket buffer" I had before.

This also reduced average memory used by tcp sockets.

With help from Neal Cardwell.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: H.K. Jerry Chu <hkchu@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-03-19 16:53:08 -04:00
..
9p virtio: rename virtqueue_add_buf_gfp to virtqueue_add_buf 2012-01-12 15:44:42 +10:30
802
8021q vlan: static functions 2011-12-14 02:39:30 -05:00
appletalk
atm ppp: Replace uses of <linux/if_ppp.h> with <linux/ppp-ioctl.h> 2012-03-04 20:41:38 -05:00
ax25 ax25: avoid overflows in ax25_setsockopt() 2011-12-28 14:08:08 -05:00
batman-adv Merge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge 2012-03-11 15:36:34 -07:00
bluetooth Bluetooth: Set security level on incoming pairing request 2012-03-08 02:26:04 -03:00
bridge Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2012-03-09 14:34:20 -08:00
caif caif: make zero a legal caif connetion id. 2012-03-11 15:38:16 -07:00
can
ceph Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client 2012-02-02 15:47:33 -08:00
core net: get rid of some pointless casts to sockaddr 2012-03-11 19:11:22 -07:00
dcb
dccp dccp: fix bug in sequence number validation during connection setup 2012-03-03 09:02:52 -07:00
decnet decnet: remove unused variable from dn_output() 2012-02-05 17:44:55 -05:00
dns_resolver
dsa dsa: Move switch drivers to new directory drivers/net/dsa 2011-11-29 00:21:36 -05:00
econet
ethernet eth: reset addr_assign_type if eth_mac_addr() called 2012-02-13 00:46:41 -05:00
ieee802154 net/ieee802154/6lowpan.c: reuse eth_mac_addr() 2012-02-22 14:46:37 -05:00
ipv4 tcp: reduce out_of_order memory use 2012-03-19 16:53:08 -04:00
ipv6 Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2012-03-18 23:29:41 -04:00
ipx
irda ppp: Replace uses of <linux/if_ppp.h> with <linux/ppp-ioctl.h> 2012-03-04 20:41:38 -05:00
iucv af_iucv: add shutdown for HS transport 2012-03-07 22:52:24 -08:00
key net: use IS_ENABLED(CONFIG_IPV6) 2011-12-11 18:25:16 -05:00
l2tp ppp: Replace uses of <linux/if_ppp.h> with <linux/ppp-ioctl.h> 2012-03-04 20:41:38 -05:00
lapb
llc llc: Fix race condition in llc_ui_recvmsg 2012-01-24 15:33:19 -05:00
mac80211 mac80211: make uapsd_* keys per-vif 2012-03-15 13:43:12 -04:00
netfilter Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2012-03-18 23:29:41 -04:00
netlabel net: reintroduce missing rcu_assign_pointer() calls 2012-01-12 12:26:56 -08:00
netlink netlink: allow to pass data pointer to netlink_dump_start() callback 2012-02-26 14:10:44 -05:00
netrom netrom: avoid overflows in nr_setsockopt() 2011-12-28 14:08:08 -05:00
nfc NFC: NCI code identation fixes 2012-03-06 15:16:25 -05:00
openvswitch Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2012-03-09 14:34:20 -08:00
packet net: Add framework to allow sending packets with customized CRC. 2012-02-24 01:37:35 -08:00
phonet net: reintroduce missing rcu_assign_pointer() calls 2012-01-12 12:26:56 -08:00
rds rds: Make rds_sock_lock BH rather than IRQ safe. 2012-01-24 17:03:44 -05:00
rfkill Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next into for-davem 2012-01-05 10:13:24 -05:00
rose
rxrpc RxRPC: Fix kcalloc parameters swapped 2012-02-14 14:41:55 -05:00
sched Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2012-03-18 23:29:41 -04:00
sctp sctp: Export sctp_do_peeloff 2012-03-08 13:52:08 -08:00
sunrpc SUNRPC: Fix machine creds in generic_create_cred and generic_match 2012-01-23 14:03:46 -08:00
tipc tipc: Optimize setting of immutable payload message header fields 2012-02-29 11:45:35 -05:00
unix netlink: add netlink_dump_control structure for netlink_dump_start() 2012-02-26 14:10:06 -05:00
wanrouter
wimax
wireless cfg80211: clarify timestamp in cfg80211_inform_bss 2012-03-13 14:54:20 -04:00
x25 net:x25: use IS_ENABLED 2011-12-16 15:49:52 -05:00
xfrm netlink: add netlink_dump_control structure for netlink_dump_start() 2012-02-26 14:10:06 -05:00
compat.c net: get rid of some pointless casts to sockaddr 2012-03-11 19:11:22 -07:00
Kconfig net: Add Open vSwitch kernel components. 2011-12-03 09:35:17 -08:00
Makefile net: Add Open vSwitch kernel components. 2011-12-03 09:35:17 -08:00
nonet.c
socket.c net: get rid of some pointless casts to sockaddr 2012-03-11 19:11:22 -07:00
sysctl_net.c