linux_dsm_epyc7002/net
Yuchung Cheng ddf1af6fa0 tcp: new delivery accounting
This patch changes the accounting of how many packets are
newly acked or sacked when the sender receives an ACK.

The current approach basically computes

   newly_acked_sacked = (prior_packets - prior_sacked) -
                        (tp->packets_out - tp->sacked_out)

   where prior_packets and prior_sacked out are snapshot
   at the beginning of the ACK processing.

The new approach tracks the delivery information via a new
TCP state variable "delivered" which monotically increases
as new packets are delivered in order or out-of-order.

The reason for this change is that the current approach is
brittle that produces negative or inaccurate estimate.

   1) For non-SACK connections, an ACK that advances the SND.UNA
   could reset the DUPACK counters (tp->sacked_out) in
   tcp_process_loss() or tcp_fastretrans_alert(). This inflates
   the inflight suddenly and causes under-estimate or even
   negative estimate. Here is a real example:

                   before   after (processing ACK)
   packets_out     75       73
   sacked_out      23        0
   ca state        Loss     Open

   The old approach computes (75-23) - (73 - 0) = -21 delivered
   while the new approach computes 1 delivered since it
   considers the 2nd-24th packets are delivered OOO.

   2) MSS change would re-count packets_out and sacked_out so
   the estimate is in-accurate and can even become negative.
   E.g., the inflight is doubled when MSS is halved.

   3) Spurious retransmission signaled by DSACK is not accounted

The new approach is simpler and more robust. For SACK connections,
tp->delivered increments as packets are being acked or sacked in
SACK and ACK processing.

For non-sack connections, it's done in tcp_remove_reno_sacks() and
tcp_add_reno_sack(). When an ACK advances the SND.UNA, tp->delivered
is incremented by the number of packets ACKed (less the current
number of DUPACKs received plus one packet hole).  Upon receiving
a DUPACK, tp->delivered is incremented assuming one out-of-order
packet is delivered.

Upon receiving a DSACK, tp->delivered is incremtened assuming one
retransmission is delivered in tcp_sacktag_write_queue().

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-07 14:09:51 -05:00
..
6lowpan 6lowpan: fix debugfs interface entry name 2015-12-20 08:21:00 +01:00
9p Rework and error handling fixes, primarily in the fscatch and fd transports. 2016-01-24 12:39:09 -08:00
802
8021q net: Rename NETIF_F_ALL_CSUM to NETIF_F_CSUM_MASK 2015-12-15 16:50:08 -05:00
appletalk
atm net: Generalise wq_has_sleeper helper 2015-11-30 14:47:33 -05:00
ax25 net: add validation for the socket syscall protocol argument 2015-12-14 16:09:30 -05:00
batman-adv batman-adv: Drop immediate orig_node free function 2016-01-16 22:50:00 +08:00
bluetooth Bluetooth: Fix incorrect removing of IRKs 2016-01-29 11:47:24 +01:00
bridge switchdev: Require RTNL mutex to be held when sending FDB notifications 2016-01-28 16:21:31 -08:00
caif net: rename SOCK_ASYNC_NOSPACE and SOCK_ASYNC_WAITDATA 2015-12-01 15:45:05 -05:00
can can: avoid using timeval for uapi 2015-10-13 17:42:34 +02:00
ceph libceph: remove outdated comment 2016-01-21 19:36:09 +01:00
core ethtool: Declare netdev_rss_key as __read_mostly. 2016-02-06 03:13:49 -05:00
dcb net/dcb: make dcbnl.c explicitly non-modular 2015-10-09 07:52:27 -07:00
dccp Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2015-12-03 21:09:12 -05:00
decnet net: add validation for the socket syscall protocol argument 2015-12-14 16:09:30 -05:00
dns_resolver net: dns_resolver: convert time_t to time64_t 2015-11-18 16:27:46 -05:00
dsa dsa: Register netdev before phy 2016-01-07 14:31:26 -05:00
ethernet net: Add eth_platform_get_mac_address() helper. 2016-01-06 16:31:56 -05:00
hsr net/hsr: fix a warning message 2015-11-23 14:56:15 -05:00
ieee802154 inet: kill unused skb_free op 2016-01-05 22:25:57 -05:00
ipv4 tcp: new delivery accounting 2016-02-07 14:09:51 -05:00
ipv6 Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2016-02-01 15:56:08 -08:00
ipx
irda irda: fix a potential use-after-free in ircomm_param_request 2016-01-29 22:56:46 -08:00
iucv af_iucv: Validate socket address length in iucv_sock_bind() 2016-01-19 14:21:08 -05:00
key af_key: fix two typos 2015-10-23 03:05:19 -07:00
l2tp l2tp: rely on ppp layer for skb scrubbing 2016-01-04 16:45:24 -05:00
l3mdev net: Add netif_is_l3_slave 2015-10-07 04:27:43 -07:00
lapb
llc tcp: fix recv with flags MSG_WAITALL | MSG_PEEK 2015-07-27 01:06:53 -07:00
mac80211 Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2016-02-01 15:56:08 -08:00
mac802154 mac802154: constify ieee802154_llsec_ops structure 2016-01-04 20:40:41 +01:00
mpls Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2015-12-17 22:08:28 -05:00
netfilter netfilter: nf_conntrack: use safer way to lock all buckets 2016-01-20 14:15:31 +01:00
netlabel
netlink netlink: not trim skb for mmaped socket when dump 2016-01-29 20:25:17 -08:00
netrom
nfc NFC 4.5 pull request 2016-01-04 21:48:15 -05:00
openvswitch ovs: limit ovs recursions in ovs_execute_actions to not corrupt stack 2016-01-18 12:09:45 -05:00
packet packet: Allow packets with only a header (but no payload) 2015-11-29 22:17:17 -05:00
phonet phonet: properly unshare skbs in phonet_rcv() 2016-01-12 12:05:38 -05:00
rds Initial roundup of 4.5 merge window patches 2016-01-23 18:45:06 -08:00
rfkill rfkill: fix rfkill_fop_read wait_event usage 2016-01-26 11:32:05 +01:00
rose
rxrpc Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next 2016-01-12 18:57:02 -08:00
sched net_sched: drr: check for NULL pointer in drr_dequeue 2016-01-29 17:26:44 -08:00
sctp sctp: remove the dead field of sctp_transport 2016-01-28 15:59:32 -08:00
sunrpc Initial roundup of 4.5 merge window patches 2016-01-23 18:45:06 -08:00
switchdev switchdev: Require RTNL mutex to be held when sending FDB notifications 2016-01-28 16:21:31 -08:00
tipc tipc: use alloc_ordered_workqueue() instead of WQ_UNBOUND w/ max_active = 1 2016-02-06 03:41:58 -05:00
unix net: drop write-only stack variable 2016-02-07 14:06:26 -05:00
vmw_vsock Revert "Merge branch 'vsock-virtio'" 2015-12-08 21:55:49 -05:00
wimax net:wimax: Fix doucble word "the the" in networking.xml 2015-08-09 22:43:52 -07:00
wireless regulatory: fix world regulatory domain data 2016-01-14 11:10:13 +01:00
x25
xfrm net: preserve IP control block during GSO segmentation 2016-01-15 14:35:24 -05:00
compat.c
Kconfig net, sched: add clsact qdisc 2016-01-10 22:13:15 -05:00
Makefile net: Introduce L3 Master device abstraction 2015-09-29 20:40:32 -07:00
socket.c kmemcg: account certain kmem allocations to memcg 2016-01-14 16:00:49 -08:00
sysctl_net.c net: sysctl: fix a kmemleak warning 2015-10-23 06:22:08 -07:00