Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

By rounding up the buffer size to power of 2, several expensive
modulus operations can be avoided.  This patch also solves a bug where
the gap need when ring gets full was not being accounted for.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

GC is non-existent in netns, so after you hit GC threshold, no new
dst entries will be created until someone triggers cleanup in init_net.

Make xfrm4_dst_ops and xfrm6_dst_ops per-netns.
This is not done in a generic way, because it woule waste
(AF_MAX - 2) * sizeof(struct dst_ops) bytes per-netns.

Reorder GC threshold initialization so it'd be done before registering
XFRM policies.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

No matter whether connection track is enabled, an end host should send 
an ICMPv4 "Fragment Reassembly Timeout" message when defrag timeout. 
The reasons are following two points:  

1. RFC 792 says:
   >>>> >> > >   If a host reassembling a fragmented datagram cannot complete the
   >>>> >> > >   reassembly due to missing fragments within its time limit it
   >>>> >> > >   discards the datagram, and it may send a time exceeded message.
   >>>> >> > > 
   >>>> >> > >   If fragment zero is not available then no time exceeded need be
   >>>> >> > >   sent at all.
   >>>> >> > > 
   >>>> >> > > Read more: http://www.faqs.org/rfcs/rfc792.html#ixzz0aOXRD7Wp

2. Patrick McHardy also agrees with this opinion.   :-)   
   About the discussion of this opinion, refer to http://patchwork.ozlabs.org/patch/41649

The patch fixed the problem like this:
When enabling connection track, fragments are received at PRE_ROUTING HOOK.
If they are failed to reassemble, ip_expire() will be called. 
Before sending an ICMP "Fragment Reassembly Timeout" message, 
the patch searches router table to get the destination entry only for host type.

The patch has been tested on both host type and route type.
Signed-off-by: NShan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The s-port should be compared.
Signed-off-by: NRoel Kluin <roel.kluin@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If the per device ARP_ACCEPT option is enable, currently we only allow
creating new ARP cache entries for response type gratuitous ARP.

Allowing gratuitous ARP to create new ARP entries (not only to update
existing ones) is useful when we want to avoid unnecessary delays for
the first packet of a stream.

This patch allows request type gratuitous ARP to create new ARP cache
entries as well. This is useful when we want to populate the ARP cache
entries for a large number of hosts on the same LAN.
Signed-off-by: NOctavian Purdila <opurdila@ixiacom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

/proc/net/rt_acct is not created if NET_CLS_ROUTE=n.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

__net_init/__net_exit are apparently not going away, so use them
to full extent.

In some cases __net_init was removed, because it was called from
__net_exit code.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently we don't increment SYN-ACK timeouts & retransmissions
although we do increment the same stats for SYN. We seem to have lost
the SYN-ACK accounting with the introduction of tcp_syn_recv_timer
(commit 2248761e in the netdev-vger-cvs tree).

This patch fixes this issue. In the process we also rename the v4/v6
syn/ack retransmit functions for clarity. We also add a new
request_socket operations (syn_ack_timeout) so we can keep code in
inet_connection_sock.c protocol agnostic.
Signed-off-by: NOctavian Purdila <opurdila@ixiacom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As noticed by H Hartley Sweeten, since change_nexthops() uses 'nh'
as it's iterator variable, it can conflict with other existing
local vars.

Use "nexthop_nh" to avoid the conflict and make it easier to figure
out where this magic variable comes from.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds the kernel portions needed to implement
RFC 5082 Generalized TTL Security Mechanism (GTSM).
It is a lightweight security measure against forged
packets causing DoS attacks (for BGP). 

This is already implemented the same way in BSD kernels.
For the necessary Quagga patch 
  http://www.gossamer-threads.com/lists/quagga/dev/17389

Description from Cisco
  http://www.cisco.com/en/US/docs/ios/12_3t/12_3t7/feature/guide/gt_btsh.html

It does add one byte to each socket structure, but I did
a little rearrangement to reuse a hole (on 64 bit), but it
does grow the structure on 32 bit

This should be documented on ip(4) man page and the Glibc in.h
file also needs update.  IPV6_MINHOPLIMIT should also be added
(although BSD doesn't support that).  

Only TCP is supported, but could also be added to UDP, DCCP, SCTP
if desired.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This is to be used together with switch technologies, like RFC3069,
that where the individual ports are not allowed to communicate with
each other, but they are allowed to talk to the upstream router.  As
described in RFC 3069, it is possible to allow these hosts to
communicate through the upstream router by proxy_arp'ing.

This patch basically allow proxy arp replies back to the same
interface (from which the ARP request/solicitation was received).

Tunable per device via proc "proxy_arp_pvlan":
  /proc/sys/net/ipv4/conf/*/proxy_arp_pvlan

This switch technology is known by different vendor names:
 - In RFC 3069 it is called VLAN Aggregation.
 - Cisco and Allied Telesyn call it Private VLAN.
 - Hewlett-Packard call it Source-Port filtering or port-isolation.
 - Ericsson call it MAC-Forced Forwarding (RFC Draft).
Signed-off-by: NJesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When we have L3 tunnels with different inner/outer families
(i.e. IPV4/IPV6) which use a multicast address as the outer tunnel
destination address, multicast packets will be loopbacked back to the
sending socket even if IP*_MULTICAST_LOOP is set to disabled.

The mc_loop flag is present in the family specific part of the socket
(e.g. the IPv4 or IPv4 specific part).  setsockopt sets the inner
family mc_loop flag. When the packet is pushed through the L3 tunnel
it will eventually be processed by the outer family which if different
will check the flag in a different part of the socket then it was set.
Signed-off-by: NOctavian Purdila <opurdila@ixiacom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

when using policy routing and the skb mark:
there are cases where a back path validation requires us
to use a different routing table for src ip validation than
the one used for mapping ingress dst ip.
One such a case is transparent proxying where we pretend to be
the destination system and therefore the local table
is used for incoming packets but possibly a main table would
be used on outbound.
Make the default behavior to allow the above and if users
need to turn on the symmetry via sysctl src_valid_mark
Signed-off-by: NJamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add rtnetlink init_rcvwnd to set the TCP initial receive window size
advertised by passive and active TCP connections.
The current Linux TCP implementation limits the advertised TCP initial
receive window to the one prescribed by slow start. For short lived
TCP connections used for transaction type of traffic (i.e. http
requests), bounding the advertised TCP initial receive window results
in increased latency to complete the transaction.
Support for setting initial congestion window is already supported
using rtnetlink init_cwnd, but the feature is useless without the
ability to set a larger TCP initial receive window.
The rtnetlink init_rcvwnd allows increasing the TCP initial receive
window, allowing TCP connection to advertise larger TCP receive window
than the ones bounded by slow start.
Signed-off-by: NLaurent Chavey <chavey@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Slightly optimize tcp_sendmsg since NETIF_F_SG is used many
times iteratively in the loop. The only other modification is
to change:
			} else if (i == MAX_SKB_FRAGS ||
				   (!i &&
				   !(sk->sk_route_caps & NETIF_F_SG))) {
	to:
			} else if (i == MAX_SKB_FRAGS || !sg) {

The reason why this change is correct: this code (other than
the MAX_SKB_FRAGS case) executes only due to the else part
of: "if (skb_tailroom(skb) > 0) {" - i.e. there was no space
in the skb to put the data inline. Hence SG is false is a
sufficient condition, and there is no way a fragment can be
added to the skb.

Changelog:
	- Added the above explanation for the change
Signed-off-by: NKrishna Kumar <krkumar2@in.ibm.com>
Acked-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove unrequired operations in tcp_push()

Changelog:
	Removed a temporary skb variable from tcp_push()
Signed-off-by: NKrishna Kumar <krkumar2@in.ibm.com>
Acked-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

tcp_push checks tcp_send_head and calls __tcp_push_pending_frames,
which again checks tcp_send_head, and this unnecessary check is
done for every other caller of __tcp_push_pending_frames.

Remove tcp_send_head check in __tcp_push_pending_frames and add
the check to tcp_push_pending_frames. Other functions call
__tcp_push_pending_frames only when tcp_send_head would evaluate
to true.
Signed-off-by: NKrishna Kumar <krkumar2@in.ibm.com>
Acked-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It creates a regression, triggering badness for SYN_RECV
sockets, for example:

[19148.022102] Badness at net/ipv4/inet_connection_sock.c:293
[19148.022570] NIP: c02a0914 LR: c02a0904 CTR: 00000000
[19148.023035] REGS: eeecbd30 TRAP: 0700   Not tainted  (2.6.32)
[19148.023496] MSR: 00029032 <EE,ME,CE,IR,DR>  CR: 24002442  XER: 00000000
[19148.024012] TASK = eee9a820[1756] 'privoxy' THREAD: eeeca000

This is likely caused by the change in the 'estab' parameter
passed to tcp_parse_options() when invoked by the functions
in net/ipv4/tcp_minisocks.c

But even if that is fixed, the ->conn_request() changes made in
this patch series is fundamentally wrong.  They try to use the
listening socket's 'dst' to probe the route settings.  The
listening socket doesn't even have a route, and you can't
get the right route (the child request one) until much later
after we setup all of the state, and it must be done by hand.

This stuff really isn't ready, so the best thing to do is a
full revert.  This reverts the following commits:

f55017a9
022c3f7d
1aba721e
cda42ebd
345cda2f
dc343475
05eaade2
6a2a2d6bSigned-off-by: NDavid S. Miller <davem@davemloft.net>

When fragments from bridge netfilter are passed to IPv4 or IPv6 conntrack
and a reassembly queue with the same fragment key already exists from
reassembling a similar packet received on a different device (f.i. with
multicasted fragments), the reassembled packet might continue on a different
codepath than where the head fragment originated. This can cause crashes
in bridge netfilter when a fragment received on a non-bridge device (and
thus with skb->nf_bridge == NULL) continues through the bridge netfilter
code.

Add a new reassembly identifier for packets originating from bridge
netfilter and use it to put those packets in insolated queues.

Fixes http://bugzilla.kernel.org/show_bug.cgi?id=14805Reported-and-Tested-by: NChong Qiao <qiaochong@loongson.cn>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

Now we can have a large udp hash table, udp_lib_get_port() loop
should be converted to a do {} while (cond) form,
or we dont enter it at all if hash table size is exactly 65536.
Reported-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It can happen, that tcp_retransmit_skb fails due to some error.
In such cases we might end up into a state where tp->retrans_out
is zero but that's only because we removed the TCPCB_SACKED_RETRANS
bit from a segment but couldn't retransmit it because of the error
that happened. Therefore some assumptions that retrans_out checks
are based do not necessarily hold, as there still can be an old
retransmission but that is only visible in TCPCB_EVER_RETRANS bit.
As retransmission happen in sequential order (except for some very
rare corner cases), it's enough to check the head skb for that bit.

Main reason for all this complexity is the fact that connection dying
time now depends on the validity of the retrans_stamp, in particular,
that successive retransmissions of a segment must not advance
retrans_stamp under any conditions. It seems after quick thinking that
this has relatively low impact as eventually TCP will go into CA_Loss
and either use the existing check for !retrans_stamp case or send a
retransmission successfully, setting a new base time for the dying
timer (can happen only once). At worst, the dying time will be
approximately the double of the intented time. In addition,
tcp_packet_delayed() will return wrong result (has some cc aspects
but due to rarity of these errors, it's hardly an issue).

One of retrans_stamp clearing happens indirectly through first going
into CA_Open state and then a later ACK lets the clearing to happen.
Thus tcp_try_keep_open has to be modified too.

Thanks to Damian Lukowski <damian@tvk.rwth-aachen.de> for hinting
that this possibility exists (though the particular case discussed
didn't after all have it happening but was just a debug patch
artifact).
Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch moves retransmits_timed_out() from include/net/tcp.h
to tcp_timer.c, where it is used.
Reported-by: NFrederic Leroy <fredo@starox.org>
Signed-off-by: NDamian Lukowski <damian@tvk.rwth-aachen.de>
Acked-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adds kerneldoc for inet_twsk_unhash() & inet_twsk_bind_unhash().

With help from Randy Dunlap.
Suggested-by: NEvgeniy Polyakov <zbr@ioremap.net>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When we find a timewait connection in __inet_hash_connect() and reuse
it for a new connection request, we have a race window, releasing bind
list lock and reacquiring it in __inet_twsk_kill() to remove timewait
socket from list.

Another thread might find the timewait socket we already chose, leading to
list corruption and crashes.

Fix is to remove timewait socket from bind list before releasing the bind lock.

Note: This problem happens if sysctl_tcp_tw_reuse is set.
Reported-by: Nkapil dakhane <kdakhane@gmail.com>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

First patch changes __inet_hash_nolisten() and __inet6_hash()
to get a timewait parameter to be able to unhash it from ehash
at same time the new socket is inserted in hash.

This makes sure timewait socket wont be found by a concurrent
writer in __inet_check_established()
Reported-by: Nkapil dakhane <kdakhane@gmail.com>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

GCC even warns about it, as reported by Andrew Morton:

net/ipv4/tcp.c: In function 'do_tcp_getsockopt':
net/ipv4/tcp.c:2544: warning: comparison is always false due to limited range of data type
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

That is "success", "unknown", "through", "performance", "[re|un]mapping"
, "access", "default", "reasonable", "[con]currently", "temperature"
, "channel", "[un]used", "application", "example","hierarchy", "therefore"
, "[over|under]flow", "contiguous", "threshold", "enough" and others.
Signed-off-by: NAndré Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

After TCP RCU conversion, tw->tw_refcnt should not be set to 1 in
inet_twsk_alloc(). It allows a RCU reader to get this timewait socket,
while we not yet stabilized it.

Only choice we have is to set tw_refcnt to 0 in inet_twsk_alloc(),
then atomic_add() it later, once everything is done.

Location of this atomic_add() is tricky, because we dont want another
writer to find this timewait in ehash, while tw_refcnt is still zero !

Thanks to Kapil Dakhane tests and reports.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Its currently possible that several threads issuing a connect() find
the same timewait socket and try to reuse it, leading to list
corruptions.

Condition for bug is that these threads bound their socket on same
address/port of to-be-find timewait socket, and connected to same
target. (SO_REUSEADDR needed)

To fix this problem, we could unhash timewait socket while holding
ehash lock, to make sure lookups/changes will be serialized. Only
first thread finds the timewait socket, other ones find the
established socket and return an EADDRNOTAVAIL error.

This second version takes into account Evgeniy's review and makes sure
inet_twsk_put() is called outside of locked sections.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Both netlink and /proc/net/tcp interfaces can report transient
negative values for rx queue.

ss ->
State   Recv-Q Send-Q  Local Address:Port  Peer Address:Port
ESTAB   -6     6       127.0.0.1:45956     127.0.0.1:3333 

netstat ->
tcp   4294967290      6 127.0.0.1:37784  127.0.0.1:3333 ESTABLISHED

This is because we dont lock socket while computing 
tp->rcv_nxt - tp->copied_seq,
and another CPU can update copied_seq before rcv_next in RX path.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This function walks the whole hashtable so there is no point in
passing it a network namespace.  Instead I purge all timewait
sockets from dead network namespaces that I find.  If the namespace
is one of the once I am trying to purge I am guaranteed no new timewait
sockets can be formed so this will get them all.  If the namespace
is one I am not acting for it might form a few more but I will
call inet_twsk_purge again and  shortly to get rid of them.  In
any even if the network namespace is dead timewait sockets are
useless.

Move the calls of inet_twsk_purge into batch_exit routines so
that if I am killing a bunch of namespaces at once I will just
call inet_twsk_purge once and save a lot of redundant unnecessary
work.

My simple 4k network namespace exit test the cleanup time dropped from
roughly 8.2s to 1.6s.  While the time spent running inet_twsk_purge fell
to about 2ms.  1ms for ipv4 and 1ms for ipv6.
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While we are looking up entries to free there is no reason to take
the lock in inet_twsk_purge.  We have to drop locks and restart
occassionally anyway so adding a few more in case we get on the
wrong list because of a timewait move is no big deal.  At the
same time not taking the lock for long periods of time is much
more polite to the rest of the users of the hash table.

In my test configuration of killing 4k network namespaces
this change causes 4k back to back runs of inet_twsk_purge on an
empty hash table to go from roughly 20.7s to 3.3s, and the total
time to destroy 4k network namespaces goes from roughly 44s to
3.3s.
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Refactor the code so fib_rules_register always takes a template instead
of the actual fib_rules_ops structure that will be used.  This is
required for network namespace support so 2 out of the 3 callers already
do this, it allows the error handling to be made common, and it allows
fib_rules_unregister to free the template for hte caller.

Modify fib_rules_unregister to use call_rcu instead of syncrhonize_rcu
to allw multiple namespaces to be cleaned up in the same rcu grace
period.
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit 8ec1e0ebe26087bfc5c0394ada5feb5758014fc8
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:16:35 2009 +0100

    ipv4: add sysctl to accept packets with local source addresses

    Change fib_validate_source() to accept packets with a local source address when
    the "accept_local" sysctl is set for the incoming inet device. Combined with the
    previous patches, this allows to communicate between multiple local interfaces
    over the wire.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit d124356ce314fff22a047ea334379d5105b2d834
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:16:35 2009 +0100

    net: fib_rules: allow to delete local rule

    Allow to delete the local rule and recreate it with a higher priority. This
    can be used to force packets with a local destination out on the wire instead
    of routing them to loopback. Additionally this patch allows to recreate rules
    with a priority of 0.

    Combined with the previous patch to allow oif classification, a socket can
    be bound to the desired interface and packets routed to the wire like this:

    # move local rule to lower priority
    ip rule add pref 1000 lookup local
    ip rule del pref 0

    # route packets of sockets bound to eth0 to the wire independant
    # of the destination address
    ip rule add pref 100 oif eth0 lookup 100
    ip route add default dev eth0 table 100
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Otherwise:

ERROR: "sysctl_tcp_cookie_size" [net/ipv6/ipv6.ko] undefined!
make[1]: *** [__modpost] Error 1
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

net/ipv4/tcp_output.c: In function ‘tcp_make_synack’:
net/ipv4/tcp_output.c:2488: warning: cast from pointer to integer of different size
Signed-off-by: NDavid S. Miller <davem@davemloft.net>