linux.git/net/ipv4, branch v3.3-rc2

tcp: fix tcp_trim_head() to adjust segment count with skb MSS

2012-01-30T17:42:58+00:00

This commit fixes tcp_trim_head() to recalculate the number of
segments in the skb with the skb's existing MSS, so trimming the head
causes the skb segment count to be monotonically non-increasing - it
should stay the same or go down, but not increase.

Previously tcp_trim_head() used the current MSS of the connection. But
if there was a decrease in MSS between original transmission and ACK
(e.g. due to PMTUD), this could cause tcp_trim_head() to
counter-intuitively increase the segment count when trimming bytes off
the head of an skb. This violated assumptions in tcp_tso_acked() that
tcp_trim_head() only decreases the packet count, so that packets_acked
in tcp_tso_acked() could underflow, leading tcp_clean_rtx_queue() to
pass u32 pkts_acked values as large as 0xffffffff to
ca_ops->pkts_acked().

As an aside, if tcp_trim_head() had really wanted the skb to reflect
the current MSS, it should have called tcp_set_skb_tso_segs()
unconditionally, since a decrease in MSS would mean that a
single-packet skb should now be sliced into multiple segments.

Signed-off-by: Neal Cardwell 
Acked-by: Nandita Dukkipati 
Acked-by: Ilpo Järvinen 
Signed-off-by: David S. Miller

net/tcp: Fix tcp memory limits initialization when !CONFIG_SYSCTL

2012-01-30T17:41:06+00:00

sysctl_tcp_mem() initialization was moved to sysctl_tcp_ipv4.c
in commit 3dc43e3e4d0b52197d3205214fe8f162f9e0c334, since it
became a per-ns value.

That code, however, will never run when CONFIG_SYSCTL is
disabled, leading to bogus values on those fields - causing hung
TCP sockets.

This patch fixes it by keeping an initialization code in
tcp_init(). It will be overwritten by the first net namespace
init if CONFIG_SYSCTL is compiled in, and do the right thing if
it is compiled out.

It is also named properly as tcp_init_mem(), to properly signal
its non-sysctl side effect on TCP limits.

Reported-by: Ingo Molnar 
Signed-off-by: Glauber Costa 
Cc: David S. Miller 
Link: http://lkml.kernel.org/r/4F22D05A.8030604@parallels.com
[ renamed the function, tidied up the changelog a bit ]
Signed-off-by: Ingo Molnar 
Signed-off-by: David S. Miller

ipv6: Fix ip_gre lockless xmits.

2012-01-26T21:34:08+00:00

Tunnel devices set NETIF_F_LLTX to bypass HARD_TX_LOCK.  Sit and
ipip set this unconditionally in ops->setup, but gre enables it
conditionally after parameter passing in ops->newlink. This is
not called during tunnel setup as below, however, so GRE tunnels are
still taking the lock.

modprobe ip_gre
ip tunnel add test0 mode gre remote 10.5.1.1 dev lo
ip link set test0 up
ip addr add 10.6.0.1 dev test0
 # cat /sys/class/net/test0/features
 # $DIR/test_tunnel_xmit 10 10.5.2.1
ip route add 10.5.2.0/24 dev test0
ip tunnel del test0

The newlink callback is only called in rtnl_netlink, and only if
the device is new, as it calls register_netdevice internally. Gre
tunnels are created at 'ip tunnel add' with ioctl SIOCADDTUNNEL,
which calls ipgre_tunnel_locate, which calls register_netdev.
rtnl_newlink is called at 'ip link set', but skips ops->newlink
and the device is up with locking still enabled. The equivalent
ipip tunnel works fine, btw (just substitute 'method gre' for
'method ipip').

On kernels before /sys/class/net/*/features was removed [1],
the first commented out line returns 0x6000 with method gre,
which indicates that NETIF_F_LLTX (0x1000) is not set. With ipip,
it reports 0x7000. This test cannot be used on recent kernels where
the sysfs file is removed (and ETHTOOL_GFEATURES does not currently
work for tunnel devices, because they lack dev->ethtool_ops).

The second commented out line calls a simple transmission test [2]
that sends on 24 cores at maximum rate. Results of a single run:

ipip:			19,372,306
gre before patch:	 4,839,753
gre after patch:	19,133,873

This patch replicates the condition check in ipgre_newlink to
ipgre_tunnel_locate. It works for me, both with oseq on and off.
This is the first time I looked at rtnetlink and iproute2 code,
though, so someone more knowledgeable should probably check the
patch. Thanks.

The tail of both functions is now identical, by the way. To avoid
code duplication, I'll be happy to rework this and merge the two.

[1] http://patchwork.ozlabs.org/patch/104610/
[2] http://kernel.googlecode.com/files/xmit_udp_parallel.c

Signed-off-by: Willem de Bruijn 
Acked-by: Eric Dumazet 
Signed-off-by: David S. Miller

tcp: bind() optimize port allocation

2012-01-26T02:50:43+00:00

Port autoselection finds a port and then drop the lock,
then right after that, gets the hash bucket again and lock it.

Fix it to go direct.

Signed-off-by: Flavio Leitner 
Signed-off-by: Marcelo Ricardo Leitner 
Acked-by: Eric Dumazet 
Signed-off-by: David S. Miller

tcp: bind() fix autoselection to share ports

2012-01-26T02:50:43+00:00

The current code checks for conflicts when the application
requests a specific port.  If there is no conflict, then
the request is granted.

On the other hand, the port autoselection done by the kernel
fails when all ports are bound even when there is a port
with no conflict available.

The fix changes port autoselection to check if there is a
conflict and use it if not.

Signed-off-by: Flavio Leitner 
Signed-off-by: Marcelo Ricardo Leitner 
Acked-by: Eric Dumazet 
Signed-off-by: David S. Miller

tcp: md5: using remote adress for md5 lookup in rst packet

2012-01-22T20:08:45+00:00

md5 key is added in socket through remote address.
remote address should be used in finding md5 key when
sending out reset packet.

Signed-off-by: shawnlu 
Signed-off-by: David S. Miller

tcp: detect loss above high_seq in recovery

2012-01-22T20:08:44+00:00

Correctly implement a loss detection heuristic: New sequences (above
high_seq) sent during the fast recovery are deemed lost when higher
sequences are SACKed.

Current code does not catch these losses, because tcp_mark_head_lost()
does not check packets beyond high_seq. The fix is straight-forward by
checking packets until the highest sacked packet. In addition, all the
FLAG_DATA_LOST logic are in-effective and redundant and can be removed.

Update the loss heuristic comments. The algorithm above is documented
as heuristic B, but it is redundant too because heuristic A already
covers B.

Note that this change only marks some forward-retransmitted packets LOST.
It does NOT forbid TCP performing further CWR on new losses. A potential
follow-up patch under preparation is to perform another CWR on "new"
losses such as
1) sequence above high_seq is lost (by resetting high_seq to snd_nxt)
2) retransmission is lost.

Signed-off-by: Yuchung Cheng 
Signed-off-by: David S. Miller

tcp: fix undo after RTO for CUBIC

2012-01-20T19:17:26+00:00

This patch fixes CUBIC so that cwnd reductions made during RTOs can be
undone (just as they already can be undone when using the default/Reno
behavior).

When undoing cwnd reductions, BIC-derived congestion control modules
were restoring the cwnd from last_max_cwnd. There were two problems
with using last_max_cwnd to restore a cwnd during undo:

(a) last_max_cwnd was set to 0 on state transitions into TCP_CA_Loss
(by calling the module's reset() functions), so cwnd reductions from
RTOs could not be undone.

(b) when fast_covergence is enabled (which it is by default)
last_max_cwnd does not actually hold the value of snd_cwnd before the
loss; instead, it holds a scaled-down version of snd_cwnd.

This patch makes the following changes:

(1) upon undo, revert snd_cwnd to ca->loss_cwnd, which is already, as
the existing comment notes, the "congestion window at last loss"

(2) stop forgetting ca->loss_cwnd on TCP_CA_Loss events

(3) use ca->last_max_cwnd to check if we're in slow start

Signed-off-by: Neal Cardwell 
Acked-by: Stephen Hemminger 
Acked-by: Sangtae Ha 
Signed-off-by: David S. Miller

tcp: fix undo after RTO for BIC

2012-01-20T19:17:26+00:00

This patch fixes BIC so that cwnd reductions made during RTOs can be
undone (just as they already can be undone when using the default/Reno
behavior).

When undoing cwnd reductions, BIC-derived congestion control modules
were restoring the cwnd from last_max_cwnd. There were two problems
with using last_max_cwnd to restore a cwnd during undo:

(a) last_max_cwnd was set to 0 on state transitions into TCP_CA_Loss
(by calling the module's reset() functions), so cwnd reductions from
RTOs could not be undone.

(b) when fast_covergence is enabled (which it is by default)
last_max_cwnd does not actually hold the value of snd_cwnd before the
loss; instead, it holds a scaled-down version of snd_cwnd.

This patch makes the following changes:

(1) upon undo, revert snd_cwnd to ca->loss_cwnd, which is already, as
the existing comment notes, the "congestion window at last loss"

(2) stop forgetting ca->loss_cwnd on TCP_CA_Loss events

(3) use ca->last_max_cwnd to check if we're in slow start

Signed-off-by: Neal Cardwell 
Signed-off-by: David S. Miller

Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

2012-01-18T06:26:41+00:00

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (47 commits)
  tg3: Fix single-vector MSI-X code
  openvswitch: Fix multipart datapath dumps.
  ipv6: fix per device IP snmp counters
  inetpeer: initialize ->redirect_genid in inet_getpeer()
  net: fix NULL-deref in WARN() in skb_gso_segment()
  net: WARN if skb_checksum_help() is called on skb requiring segmentation
  caif: Remove bad WARN_ON in caif_dev
  caif: Fix typo in Vendor/Product-ID for CAIF modems
  bnx2x: Disable AN KR work-around for BCM57810
  bnx2x: Remove AutoGrEEEn for BCM84833
  bnx2x: Remove 100Mb force speed for BCM84833
  bnx2x: Fix PFC setting on BCM57840
  bnx2x: Fix Super-Isolate mode for BCM84833
  net: fix some sparse errors
  net: kill duplicate included header
  net: sh-eth: Fix build error by the value which is not defined
  net: Use device model to get driver name in skb_gso_segment()
  bridge: BH already disabled in br_fdb_cleanup()
  net: move sock_update_memcg outside of CONFIG_INET
  mwl8k: Fixing Sparse ENDIAN CHECK warning
  ...