linux-stable.git/Documentation/networking/ip-sysctl.txt, branch linux-3.0.y

tcp: implement RFC 5961 3.2

2013-01-11T17:03:48+00:00

[ Upstream commit 282f23c6ee343126156dd41218b22ece96d747e3 ]

Implement the RFC 5691 mitigation against Blind
Reset attack using RST bit.

Idea is to validate incoming RST sequence,
to match RCV.NXT value, instead of previouly accepted
window : (RCV.NXT <= SEG.SEQ < RCV.NXT+RCV.WND)

If sequence is in window but not an exact match, send
a "challenge ACK", so that the other part can resend an
RST with the appropriate sequence.

Add a new sysctl, tcp_challenge_ack_limit, to limit
number of challenge ACK sent per second.

Add a new SNMP counter to count number of challenge acks sent.
(netstat -s | grep TCPChallengeACK)

Signed-off-by: Eric Dumazet 
Cc: Kiran Kumar Kella 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

tcp: change tcp_adv_win_scale and tcp_rmem[2]

2012-05-21T16:40:00+00:00

[ Upstream commit b49960a05e32121d29316cfdf653894b88ac9190 ]

tcp_adv_win_scale default value is 2, meaning we expect a good citizen
skb to have skb->len / skb->truesize ratio of 75% (3/4)

In 2.6 kernels we (mis)accounted for typical MSS=1460 frame :
1536 + 64 + 256 = 1856 'estimated truesize', and 1856 * 3/4 = 1392.
So these skbs were considered as not bloated.

With recent truesize fixes, a typical MSS=1460 frame truesize is now the
more precise :
2048 + 256 = 2304. But 2304 * 3/4 = 1728.
So these skb are not good citizen anymore, because 1460 < 1728

(GRO can escape this problem because it build skbs with a too low
truesize.)

This also means tcp advertises a too optimistic window for a given
allocated rcvspace : When receiving frames, sk_rmem_alloc can hit
sk_rcvbuf limit and we call tcp_prune_queue()/tcp_collapse() too often,
especially when application is slow to drain its receive queue or in
case of losses (netperf is fast, scp is slow). This is a major latency
source.

We should adjust the len/truesize ratio to 50% instead of 75%

This patch :

1) changes tcp_adv_win_scale default to 1 instead of 2

2) increase tcp_rmem[2] limit from 4MB to 6MB to take into account
better truesize tracking and to allow autotuning tcp receive window to
reach same value than before. Note that same amount of kernel memory is
consumed compared to 2.6 kernels.

Signed-off-by: Eric Dumazet 
Cc: Neal Cardwell 
Cc: Tom Herbert 
Cc: Yuchung Cheng 
Acked-by: Neal Cardwell 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

net: Fix default in docs for tcp_orphan_retries.

2011-07-08T16:31:31+00:00

Default should be listed at 8 instead of 7.

Reported-by: Denys Fedoryshchenko 
Signed-off-by: David S. Miller

tcp: document tcp_max_ssthresh (Limited Slow-Start)

2011-02-20T19:10:14+00:00

Base on Ilpo's patch about documenting tcp_max_ssthresh.
(see http://marc.info/?l=linux-netdev&m=117950581307310&w=2)

According to errata of RFC3742, fix the number of segments increased
during RTT time.

Just to state the occasion to use this parameter, But
about how to set parameter value, maybe some others can do it.

Signed-off-by: Ilpo Järvinen 
Signed-off-by: Shan Wei 
Signed-off-by: David S. Miller

tcp_ecn is an integer not a boolean

2011-02-02T23:39:58+00:00

There was some confusion at LCA as to why the sysctl tcp_ecn took one
of three values when it was documented as a Boolean.  This patch fixes
the documentation.

Signed-off-by: Peter Chubb 
Signed-off-by: David S. Miller

net: change ip_default_ttl documentation

2010-12-13T20:50:49+00:00

Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller

Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

2010-12-08T21:47:38+00:00

Conflicts:
	drivers/net/wireless/ath/ath9k/ar9003_eeprom.c
	net/llc/af_llc.c

tcp: restrict net.ipv4.tcp_adv_win_scale (#20312)

2010-11-28T18:39:45+00:00

tcp_win_from_space() does the following:

      if (sysctl_tcp_adv_win_scale <= 0)
              return space >> (-sysctl_tcp_adv_win_scale);
      else
              return space - (space >> sysctl_tcp_adv_win_scale);

"space" is int.

As per C99 6.5.7 (3) shifting int for 32 or more bits is
undefined behaviour.

Indeed, if sysctl_tcp_adv_win_scale is exactly 32,
space >> 32 equals space and function returns 0.

Which means we busyloop in tcp_fixup_rcvbuf().

Restrict net.ipv4.tcp_adv_win_scale to [-31, 31].

Fix https://bugzilla.kernel.org/show_bug.cgi?id=20312

Steps to reproduce:

      echo 32 >/proc/sys/net/ipv4/tcp_adv_win_scale
      wget www.kernel.org
      [softlockup]

Signed-off-by: Alexey Dobriyan 
Signed-off-by: David S. Miller

clarify documentation for net.ipv4.igmp_max_memberships

2010-11-17T19:22:06+00:00

This patch helps clarify documentation for
net.ipv4.igmp_max_memberships by providing a formula for
calculating the maximum number of multicast groups that can be
subscribed to, plus defining the theoretical limit.

Signed-off-by: Jiri Pirko 
Signed-off-by: Jeremy Eder 
Signed-off-by: David S. Miller

docs: Add neigh/gc_thresh3 and route/max_size documentation.

2010-11-12T22:03:20+00:00

Signed-off-by: Ben Greear 
Signed-off-by: David S. Miller