Yuchung Cheng
3e59cb0ddf
tcp: remove bad timeout logic in fast recovery ...
tcp_timeout_skb() was intended to trigger fast recovery on timeout,
unfortunately in reality it often causes spurious retransmission
storms during fast recovery. The particular sign is a fast retransmit
over the highest sacked sequence (SND.FACK).
Currently the RTO timer re-arming (as in RFC6298) offers a nice cushion
to avoid spurious timeout: when SND.UNA advances the sender re-arms
RTO and extends the timeout by icsk_rto. The sender does not offset
the time elapsed since the packet at SND.UNA was sent.
But if the next (DUP)ACK arrives later than ~RTTVAR and triggers
tcp_fastretrans_alert(), then tcp_timeout_skb() will mark any packet
sent before the icsk_rto interval lost, including one that's above the
highest sacked sequence. Most likely a large part of scorebard will be
marked.
If most packets are not lost then the subsequent DUPACKs with new SACK
blocks will cause the sender to continue to retransmit packets beyond
SND.FACK spuriously. Even if only one packet is lost the sender may
falsely retransmit almost the entire window.
The situation becomes common in the world of bufferbloat: the RTT
continues to grow as the queue builds up but RTTVAR remains small and
close to the minimum 200ms. If a data packet is lost and the DUPACK
triggered by the next data packet is slightly delayed, then a spurious
retransmission storm forms.
As the original comment on tcp_timeout_skb() suggests: the usefulness
of this feature is questionable. It also wastes cycles walking the
sack scoreboard and is actually harmful because of false recovery.
It's time to remove this.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Nandita Dukkipati <nanditad@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-05-19 23:51:17 -07:00
9p: turn fid->dlist into hlist
2013-02-27 22:51:08 -05:00
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
2013-05-01 17:51:54 -07:00
caif: Remove my bouncing email address.
2013-04-23 13:25:51 -04:00
irda: small read past the end of array in debug code
2013-04-19 17:32:31 -04:00
af_iucv: fix recvmsg by replacing skb_pull() function
2013-04-08 17:16:57 -04:00
netfilter: move skb_gso_segment into nfnetlink_queue module
2013-04-29 20:09:05 +02:00
netfilter: nf_log: prepare net namespace support for loggers
2013-04-05 20:12:54 +02:00
NFC: RFKILL support
2013-04-12 16:54:45 +02:00
net: remove my future former mail address
2012-06-17 16:29:38 -07:00
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
2013-05-01 14:08:52 -07:00
act_police: move struct tcf_police to act_police.c
2013-02-12 18:59:45 -05:00
ipv6: statically link register_inet6addr_notifier()
2013-04-14 15:24:17 -04:00
af_unix: fix a fatal race with bit fields
2013-05-01 15:13:49 -04:00
net: Dont use ifindices in hash fns
2012-08-09 16:18:06 -07:00
hlist: drop the node parameter from iterators
2013-02-27 19:10:24 -08:00
cfg80211: introduce critical protocol indication from user-space
2013-04-22 15:48:00 +02:00
net: core: add function for incremental IPv6 pseudo header checksum updates
2012-08-30 03:00:16 +02:00
cipso: handle CIPSO options correctly when NetLabel is disabled
2012-06-01 14:18:29 -04:00
cls_cgroup: remove task_struct parameter from sock_update_classid()
2013-04-09 13:19:35 -04:00
codel: refine one condition to avoid a nul rec_inv_sqrt
2012-08-10 16:52:54 -07:00
net: cleanup unsigned to unsigned int
2012-04-15 12:44:40 -04:00
net/dcb: Add an optional max rate attribute
2012-04-05 05:08:04 -04:00
net: cleanup unsigned to unsigned int
2012-04-15 12:44:40 -04:00
decnet: Parse netlink attributes on our own
2013-03-22 10:31:16 -04:00
decnet: use correct RCU API to deref sk_dst_cache field
2013-01-28 00:15:27 -05:00
dsa: Include linux/if_ether.h to fix build error
2011-12-01 11:41:06 -05:00
ipv6: Optimize ipv6_change_dsfield().
2013-01-09 23:59:53 -08:00
Fix dst_neigh_lookup/dst_neigh_lookup_skb return value handling bug
2013-03-15 09:06:58 -04:00
net: Fix warnings in dst_ops.h
2012-07-19 10:43:03 -07:00
ipv4: Elide fib_validate_source() completely when possible.
2012-06-29 01:36:36 -07:00
firewire net, ipv4 arp: Extend hardware address and remove driver-level packet inspection.
2013-03-26 12:32:13 -04:00
ipv4: Add FLOWI_FLAG_KNOWN_NH
2012-10-08 17:42:36 -04:00
flow_keys: include thoff into flow_keys for later usage
2013-03-20 12:14:36 -04:00
genl: Allow concurrent genl callbacks.
2013-04-25 01:43:15 -04:00
GRE: Refactor GRE tunneling code.
2013-03-26 12:27:18 -04:00
gro: Fix kcalloc argument order
2013-01-27 22:46:33 -05:00
ipv4: fix error handling in icmp_protocol.
2013-02-22 15:10:18 -05:00
mac80211: support (partial) VHT radiotap information
2012-11-27 11:56:18 +01:00
ieee802154/nl-mac.c: make some MLME operations optional
2013-04-08 12:00:16 -04:00
ipv6: add support of peer address
2013-05-19 15:09:26 -07:00
ipv6: Add helper inet6_csk_update_pmtu().
2012-07-16 03:44:56 -07:00
ipv6: use a stronger hash for tcp
2013-02-21 18:15:58 -05:00
net-tcp: Fast Open client - sendmsg(MSG_FASTOPEN)
2012-07-19 11:02:03 -07:00
tcp: Tail loss probe (TLP)
2013-03-12 08:30:34 -04:00
tunnel: drop packet if ECN present with not-ECT
2012-09-27 18:12:37 -04:00
net: frag, fix race conditions in LRU list maintenance
2013-05-06 11:06:51 -04:00
hlist: drop the node parameter from iterators
2013-02-27 19:10:24 -08:00
ipv6: use a stronger hash for tcp
2013-02-21 18:15:58 -05:00
hlist: drop the node parameter from iterators
2013-02-27 19:10:24 -08:00
ipv4: Maintain redirect and PMTU info in struct rtable again.
2012-07-10 22:40:14 -07:00
ipv4: Add a socket release callback for datagram sockets
2013-01-21 14:17:05 -05:00
ipv6: move csum_ipv6_magic() and udp6_csum_init() into static library
2013-01-08 17:56:10 -08:00
ipv6: fix race condition regarding dst->expires and dst->from.
2013-02-20 15:11:45 -05:00
ipv6: Remove unused neigh argument for icmp6_dst_alloc() and its callers.
2013-01-18 14:41:13 -05:00
GRE: Refactor GRE tunneling code.
2013-03-26 12:27:18 -04:00
ipv4: fix definition of FIB_TABLE_HASHSZ
2013-03-13 10:47:09 -04:00
GRE: Refactor GRE tunneling code.
2013-03-26 12:27:18 -04:00
ipvs: fix sparse warnings for some parameters
2013-04-23 11:43:05 +09:00
ipv6: implement RFC3168 5.3 (ecn protection) for ipv6 fragmentation handling
2013-03-24 17:16:30 -04:00
lapb: Neaten debugging
2012-05-17 18:45:20 -04:00
hostap: Don't use create_proc_read_entry()
2013-04-29 15:41:56 -04:00
llc: Remove stray reference to sysctl_llc_station_ack_timeout.
2012-09-17 13:13:24 -04:00
net: cleanup unsigned to unsigned int
2012-04-15 12:44:40 -04:00
net: delete all instances of special processing for token ring
2012-05-15 20:14:35 -04:00
mac80211: improve the rate control API
2013-04-22 16:16:41 +02:00
mac802154: add wpan device-class support
2012-06-26 21:06:11 -07:00
net/802: Implement Multiple Registration Protocol (MRP)
2013-02-10 20:37:22 -05:00
ndisc: Move ndisc_opt_addr_space() to include/net/ndisc.h.
2013-01-21 13:33:14 -05:00
net neighbour, decnet: Ensure to align device private data on preferred alignment.
2013-02-11 00:21:44 -05:00
netfilter: make /proc/net/netfilter pernet
2013-04-05 19:35:02 +02:00
ipv6 netevent: Remove old_neigh from netevent_redirect.
2013-01-14 15:04:59 -05:00
userns: Convert the audit loginuid to be a kuid
2012-09-17 18:08:54 -07:00
netlink: Rename pid to portid to avoid confusion
2012-09-10 15:30:41 -04:00
netprio_cgroup: remove task_struct parameter from sock_update_netprio()
2013-04-09 13:19:37 -04:00
hlist: drop the node parameter from iterators
2013-02-27 19:10:24 -08:00
pkt_sched: namespace aware act_mirred
2013-01-14 15:09:36 -05:00
sch_api: introduce qdisc_watchdog_schedule_ns()
2013-02-12 18:59:45 -05:00
net: Remove code duplication between offload structures
2012-11-15 17:39:51 -05:00
ipv6: bool/const conversions phase2
2012-05-19 01:08:16 -04:00
net_sched: red: Make minor corrections to comments
2012-04-16 23:53:11 -04:00
regulatory: use RCU to protect last_request
2013-01-03 13:01:30 +01:00
net: remove a stale comment for dl_next
2013-04-22 15:55:48 -04:00
ipv4: avoid a test in ip_rt_put()
2012-11-03 14:59:04 -04:00
rtnetlink: Remove passing of attributes into rtnl_doit functions
2013-03-22 10:31:16 -04:00
hlist: drop the node parameter from iterators
2013-02-27 19:10:24 -08:00
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
2013-04-22 20:32:51 -04:00
net: defer net_secret[] initialization
2013-04-29 15:14:02 -04:00
net: avoid reloads in SNMP_UPD_PO_STATS
2012-08-06 13:40:47 -07:00
ipv6: do not clear pinet6 field
2013-05-11 16:26:38 -07:00
tcp: remove bad timeout logic in fast recovery
2013-05-19 23:51:17 -07:00
cgroup: pass struct mem_cgroup instead of struct cgroup to socket memcg
2012-04-10 10:04:07 -07:00
[PATCH] tcp: Cache inetpeer in timewait socket, and only when necessary.
2012-06-09 14:56:12 -07:00
ipv6: rename datagram_send_ctl and datagram_recv_ctl
2013-01-31 13:53:08 -05:00
net/ipv6/udp: UDP encapsulation: introduce encap_rcv hook into IPv6
2012-04-28 22:21:51 -04:00
net: ipv4: Standardize prefixes for message logging
2012-03-12 17:05:21 -07:00
net: cleanup unsigned to unsigned int
2012-04-15 12:44:40 -04:00
mac802154: monitor device support
2012-05-16 15:17:08 -04:00
net: cleanup unsigned to unsigned int
2012-04-15 12:44:40 -04:00
xfrm: allow to avoid copying DSCP during encapsulation
2013-03-06 07:02:45 +01:00
3e59cb0ddf