[IPV6]: Introduce the INET6_TW_MATCH macro.

We have INET_MATCH, INET_TW_MATCH and INET6_MATCH to test sockets and
twbuckets for matching, but ipv6 twbuckets are tested manually.

Here's the INET6_TW_MATCH to help with it.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: xt_iprange: fix sparse warnings

  CHECK   net/netfilter/xt_iprange.c
net/netfilter/xt_iprange.c:104:19: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:37: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:19: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:37: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:19: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:37: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:19: warning: restricted degrades to integer
net/netfilter/xt_iprange.c:104:37: warning: restricted degrades to integer

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_nat: fix sparse warning

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: fix sparse warning

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: {ip,ip6}_queue: fix build error

Reported by Ingo Molnar:

 net/built-in.o: In function `ip_queue_init':
 ip_queue.c:(.init.text+0x322c): undefined reference to `net_ipv4_ctl_path'

Fix the build error and also handle CONFIG_PROC_FS=n properly.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: annotate l3protos with const

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_icmp: constify and annotate

Constify a few data tables use const qualifiers on variables where
possible in the nf_conntrack_icmp* sources.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_proto_gre: annotate with const

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_proto_udp{,lite}: annotate with const

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_proto_tcp: constify and annotate TCP modules

Constify a few data tables use const qualifiers on variables where
possible in the nf_*_proto_tcp sources.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_sane: annotate SANE helper with const

Annotate nf_conntrack_sane variables with const qualifier and remove
a few casts.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_pptp: annotate PPtP helper with const

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_tftp: annotate TFTP helper with const

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_{conntrack,nat}_sip: annotate SIP helper with const

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_h323: constify and annotate H.323 helper

Constify data tables (predominantly in nf_conntrack_h323_types.c, but
also a few in nf_conntrack_h323_asn1.c) and use const qualifiers on
variables where possible in the h323 sources.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: x_tables: create per-netns /proc/net/*_tables_*

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: x_tables: netns propagation for /proc/net/*_tables_names

Propagate netns together with AF down to ->start/->next/->stop
iterators. Choose table based on netns and AF for showing.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: x_tables: semi-rewrite of /proc/net/foo_tables_*

There are many small but still wrong things with /proc/net/*_tables_*
so I decided to do overhaul simultaneously making it more suitable for
per-netns /proc/net/*_tables_* implementation.

Fix
a) xt_get_idx() duplicating now standard seq_list_start/seq_list_next
   iterators
b) tables/matches/targets list was chosen again and again on every ->next
c) multiple useless "af >= NPROTO" checks -- we simple don't supply invalid
   AFs there and registration function should BUG_ON instead.

   Regardless, the one in ->next() is the most useless -- ->next doesn't
   run at all if ->start fails.
d) Don't use mutex_lock_interruptible() -- it can fail and ->stop is
   executed even if ->start failed, so unlock without lock is possible.

As side effect, streamline code by splitting xt_tgt_ops into xt_target_ops,
xt_matches_ops, xt_tables_ops.

xt_tables_ops hooks will be changed by per-netns code. Code of
xt_matches_ops, xt_target_ops is identical except the list chosen for
iterating, but I think consolidating code for two files not worth it
given "<< 16" hacks needed for it.

[Patrick: removed unused enum in x_tables.c]

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: xt_hashlimit match, revision 1

Introduces the xt_hashlimit match revision 1. It adds support for
kernel-level inversion and grouping source and/or destination IP
addresses, allowing to limit on a per-subnet basis. While this would
technically obsolete xt_limit, xt_hashlimit is a more expensive due
to the hashbucketing.

Kernel-level inversion: Previously you had to do user-level inversion:

iptables -N foo
iptables -A foo -m hashlimit --hashlimit(-upto) 5/s -j RETURN
iptables -A foo -j DROP
iptables -A INPUT -j foo

now it is simpler:

iptables -A INPUT -m hashlimit --hashlimit-over 5/s -j DROP

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: kill unused static inline (do_iter)

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ipt_CLUSTERIP: kill clusterip_config_entry_get

It's unused static inline.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_netlink: transmit mark during all events

The following feature was submitted some months ago. It forces the dump
of mark during the connection destruction event. The induced load is
quiet small and the patch is usefull to provide an easy way to filter
event on user side without having to keep an hash in userspace.

Signed-off-by: Eric Leblond <eric@inl.fr>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_h323: clean up code a bit

-total: 81 errors, 3 warnings, 876 lines checked
+total: 44 errors, 3 warnings, 876 lines checked

There is still work to be done, but that's for another patch.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_nat: switch rwlock to spinlock

Since we're using RCU, all users of nf_nat_lock take a write_lock.
Switch it to a spinlock.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_nat: use RCU for bysource hash

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: naming unification

Rename all "conntrack" variables to "ct" for more consistency and
avoiding some overly long lines.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: don't inline early_drop()

early_drop() is only called *very* rarely, unfortunately gcc inlines it
into the hotpath because there is only a single caller. Explicitly mark
it noinline.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: reorder struct nf_conntrack_l4proto

Reorder struct nf_conntrack_l4proto so all members used during packet
processing are in the same cacheline.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: optimize hash_conntrack()

Avoid calling jhash three times and hash the entire tuple in one go.

  __hash_conntrack | -485 # 760 -> 275, # inlines: 3 -> 1, size inlines: 717 -> 252
 1 function changed, 485 bytes removed

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: avoid duplicate protocol comparison in nf_ct_tuple_equal()

nf_ct_tuple_src_equal() and nf_ct_tuple_dst_equal() both compare the protocol
numbers. Unfortunately gcc doesn't optimize out the second comparison, so
remove it and prefix both functions with __ to indicate that they should not
be used directly.

Saves another 16 byte of text in __nf_conntrack_find() on x86_64:

  nf_conntrack_tuple_taken |  -20 # 320 -> 300, size inlines: 181 -> 161
  __nf_conntrack_find      |  -16 # 267 -> 251, size inlines: 127 -> 115
  __nf_conntrack_confirm   |  -40 # 875 -> 835, size inlines: 570 -> 537
 3 functions changed, 76 bytes removed

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: optimize __nf_conntrack_find()

Ignoring specific entries in __nf_conntrack_find() is only needed by NAT
for nf_conntrack_tuple_taken(). Remove it from __nf_conntrack_find()
and make nf_conntrack_tuple_taken() search the hash itself.

Saves 54 bytes of text in the hotpath on x86_64:

  __nf_conntrack_find      |  -54 # 321 -> 267, # inlines: 3 -> 2, size inlines: 181 -> 127
  nf_conntrack_tuple_taken | +305 # 15 -> 320, lexblocks: 0 -> 3, # inlines: 0 -> 3, size inlines: 0 -> 181
  nf_conntrack_find_get    |   -2 # 90 -> 88
 3 functions changed, 305 bytes added, 56 bytes removed, diff: +249

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: switch rwlock to spinlock

With the RCU conversion only write_lock usages of nf_conntrack_lock are
left (except one read_lock that should actually use write_lock in the
H.323 helper). Switch to a spinlock.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: use RCU for conntrack hash

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_expect: use RCU for expectation hash

Use RCU for expectation hash. This doesn't buy much for conntrack
runtime performance, but allows to reduce the use of nf_conntrack_lock
for /proc and nf_netlink_conntrack.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_core: avoid taking nf_conntrack_lock in nf_conntrack_alter_reply

The conntrack is unconfirmed, so we have an exclusive reference, which
means that the write_lock is definitely unneeded. A read_lock used to
be needed for the helper lookup, but since we're using RCU for helpers
now rcu_read_lock is enough.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: use RCU for conntrack helpers

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: fix accounting with fixed timeouts

Don't skip accounting for conntracks with the FIXED_TIMEOUT bit.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_netlink: fix unbalanced locking

Properly drop nf_conntrack_lock on tuple parsing error.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_ipv6: fix sparse warnings

  CHECK   net/ipv6/netfilter/nf_conntrack_reasm.c
  net/ipv6/netfilter/nf_conntrack_reasm.c:77:18: warning: symbol 'nf_ct_ipv6_sysctl_table' was not declared. Should it be static?
  net/ipv6/netfilter/nf_conntrack_reasm.c:586:16: warning: symbol 'nf_ct_frag6_gather' was not declared. Should it be static?
  net/ipv6/netfilter/nf_conntrack_reasm.c:662:6: warning: symbol 'nf_ct_frag6_output' was not declared. Should it be static?
  net/ipv6/netfilter/nf_conntrack_reasm.c:683:5: warning: symbol 'nf_ct_frag6_kfree_frags' was not declared. Should it be static?
  net/ipv6/netfilter/nf_conntrack_reasm.c:698:5: warning: symbol 'nf_ct_frag6_init' was not declared. Should it be static?
  net/ipv6/netfilter/nf_conntrack_reasm.c:717:6: warning: symbol 'nf_ct_frag6_cleanup' was not declared. Should it be static?

Based on patch by Stephen Hemminger with suggestions by Yasuyuki KOZAKAI.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: {ip,arp,ip6}_tables: fix sparse warnings in compat code

  CHECK   net/ipv4/netfilter/ip_tables.c
net/ipv4/netfilter/ip_tables.c:1453:8: warning: incorrect type in argument 3 (different signedness)
net/ipv4/netfilter/ip_tables.c:1453:8:    expected int *size
net/ipv4/netfilter/ip_tables.c:1453:8:    got unsigned int [usertype] *size
net/ipv4/netfilter/ip_tables.c:1458:44: warning: incorrect type in argument 3 (different signedness)
net/ipv4/netfilter/ip_tables.c:1458:44:    expected int *size
net/ipv4/netfilter/ip_tables.c:1458:44:    got unsigned int [usertype] *size
net/ipv4/netfilter/ip_tables.c:1603:2: warning: incorrect type in argument 2 (different signedness)
net/ipv4/netfilter/ip_tables.c:1603:2:    expected unsigned int *i
net/ipv4/netfilter/ip_tables.c:1603:2:    got int *<noident>
net/ipv4/netfilter/ip_tables.c:1627:8: warning: incorrect type in argument 3 (different signedness)
net/ipv4/netfilter/ip_tables.c:1627:8:    expected int *size
net/ipv4/netfilter/ip_tables.c:1627:8:    got unsigned int *size
net/ipv4/netfilter/ip_tables.c:1634:40: warning: incorrect type in argument 3 (different signedness)
net/ipv4/netfilter/ip_tables.c:1634:40:    expected int *size
net/ipv4/netfilter/ip_tables.c:1634:40:    got unsigned int *size
net/ipv4/netfilter/ip_tables.c:1653:8: warning: incorrect type in argument 5 (different signedness)
net/ipv4/netfilter/ip_tables.c:1653:8:    expected unsigned int *i
net/ipv4/netfilter/ip_tables.c:1653:8:    got int *<noident>
net/ipv4/netfilter/ip_tables.c:1666:2: warning: incorrect type in argument 2 (different signedness)
net/ipv4/netfilter/ip_tables.c:1666:2:    expected unsigned int *i
net/ipv4/netfilter/ip_tables.c:1666:2:    got int *<noident>
  CHECK   net/ipv4/netfilter/arp_tables.c
net/ipv4/netfilter/arp_tables.c:1285:40: warning: incorrect type in argument 3 (different signedness)
net/ipv4/netfilter/arp_tables.c:1285:40:    expected int *size
net/ipv4/netfilter/arp_tables.c:1285:40:    got unsigned int *size
net/ipv4/netfilter/arp_tables.c:1543:44: warning: incorrect type in argument 3 (different signedness)
net/ipv4/netfilter/arp_tables.c:1543:44:    expected int *size
net/ipv4/netfilter/arp_tables.c:1543:44:    got unsigned int [usertype] *size
  CHECK   net/ipv6/netfilter/ip6_tables.c
net/ipv6/netfilter/ip6_tables.c:1481:8: warning: incorrect type in argument 3 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1481:8:    expected int *size
net/ipv6/netfilter/ip6_tables.c:1481:8:    got unsigned int [usertype] *size
net/ipv6/netfilter/ip6_tables.c:1486:44: warning: incorrect type in argument 3 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1486:44:    expected int *size
net/ipv6/netfilter/ip6_tables.c:1486:44:    got unsigned int [usertype] *size
net/ipv6/netfilter/ip6_tables.c:1631:2: warning: incorrect type in argument 2 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1631:2:    expected unsigned int *i
net/ipv6/netfilter/ip6_tables.c:1631:2:    got int *<noident>
net/ipv6/netfilter/ip6_tables.c:1655:8: warning: incorrect type in argument 3 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1655:8:    expected int *size
net/ipv6/netfilter/ip6_tables.c:1655:8:    got unsigned int *size
net/ipv6/netfilter/ip6_tables.c:1662:40: warning: incorrect type in argument 3 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1662:40:    expected int *size
net/ipv6/netfilter/ip6_tables.c:1662:40:    got unsigned int *size
net/ipv6/netfilter/ip6_tables.c:1680:8: warning: incorrect type in argument 5 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1680:8:    expected unsigned int *i
net/ipv6/netfilter/ip6_tables.c:1680:8:    got int *<noident>
net/ipv6/netfilter/ip6_tables.c:1693:2: warning: incorrect type in argument 2 (different signedness)
net/ipv6/netfilter/ip6_tables.c:1693:2:    expected unsigned int *i
net/ipv6/netfilter/ip6_tables.c:1693:2:    got int *<noident>

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ipt_recent: fix sparse warnings

net/ipv4/netfilter/ipt_recent.c:215:17: warning: symbol 't' shadows an earlier one
net/ipv4/netfilter/ipt_recent.c:179:22: originally declared here
net/ipv4/netfilter/ipt_recent.c:322:13: warning: context imbalance in 'recent_seq_start' - wrong count at exit
net/ipv4/netfilter/ipt_recent.c:354:13: warning: context imbalance in 'recent_seq_stop' - unexpected unlock

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack_h3223: sparse fixes

Sparse complains when a function is not really static. Putting static
on the function prototype is not enough.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: more sparse fixes

Some lock annotations, and make initializers static.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: conntrack: get rid of sparse warnings

Teach sparse about locking here, and fix signed/unsigned warnings.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nfnetlink_log: sparse warning fixes

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: sparse warnings

The hashtable size is really unsigned so sparse complains when you pass
a signed integer.  Change all uses to make it consistent.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_nat_snmp: sparse warning

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: xt_owner: allow matching UID/GID ranges

Add support for ranges to the new revision. This doesn't affect
compatibility since the new revision was not released yet.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: xt_TCPMSS: consider reverse route's MTU in clamp-to-pmtu

The TCPMSS target in Xtables should consider the MTU of the reverse
route on forwarded packets as part of the path MTU.

Point in case: IN=ppp0, OUT=eth0. MSS set to 1460 in spite of MTU of
ppp0 being 1392.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: netns: put table module on netns stop

When number of entries exceeds number of initial entries, foo-tables code
will pin table module. But during table unregister on netns stop,
that additional pin was forgotten.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: arp_tables: per-netns arp_tables FILTER

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: arp_tables: netns preparation

* Propagate netns from userspace.
* arpt_register_table() registers table in supplied netns.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ip6_tables: per-netns IPv6 FILTER, MANGLE, RAW

Now it's possible to list and manipulate per-netns ip6tables rules.
Filtering decisions are based on init_net's table so far.

P.S.: remove init_net check in inet6_create() to see the effect

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ip6_tables: netns preparation

* Propagate netns from userspace down to xt_find_table_lock()
* Register ip6 tables in netns (modules still use init_net)

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ip_tables: per-netns FILTER, MANGLE, RAW

Now, iptables show and configure different set of rules in different
netnss'. Filtering decisions are still made by consulting only
init_net's set.

Changes are identical except naming so no splitting.

P.S.: one need to remove init_net checks in nf_sockopt.c and inet_create()
      to see the effect.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ip_tables: propagate netns from userspace

.. all the way down to table searching functions.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: x_tables: return new table from {arp,ip,ip6}t_register_table()

Typical table module registers xt_table structure (i.e. packet_filter)
and link it to list during it. We can't use one template for it because
corresponding list_head will become corrupted. We also can't unregister
with template because it wasn't changed at all and thus doesn't know in
which list it is.

So, we duplicate template at the very first step of table registration.
Table modules will save it for use during unregistration time and actual
filtering.

Do it at once to not screw bisection.

P.S.: renaming i.e. packet_filter => __packet_filter is temporary until
      full netnsization of table modules is done.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: x_tables: per-netns xt_tables

In fact all we want is per-netns set of rules, however doing that will
unnecessary complicate routines such as ipt_hook()/ipt_do_table, so
make full xt_table array per-netns.

Every user stubbed with init_net for a while.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: x_tables: change xt_table_register() return value convention

Switch from 0/-E to ptr/PTR_ERR convention.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ebtables: mark matches, targets and watchers __read_mostly

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ebtables: Update modules' descriptions

Update the MODULES_DESCRIPTION() tags for all Ebtables modules.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ebtables: remove casts, use consts

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_log: add netfilter gcc printf format checking

Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: xt_conntrack: add port and direction matching

Extend the xt_conntrack match revision 1 by port matching (all four
{orig,repl}{src,dst}) and by packet direction matching.

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nfnetlink_log: fix typo

It should use htonl for the GID, not htons.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

linux/types.h: Use __u64 for aligned_u64

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: bridge netfilter: remove nf_bridge_info read-only netoutdev member

Before the removal of the deferred output hooks, netoutdev was used in
case of VLANs on top of a bridge to store the VLAN device, so the
deferred hooks would see the correct output device. This isn't
necessary anymore since we're calling the output hooks for the correct
device directly in the IP stack.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_nat: remove double bysource hash initialization

The hash table is already initialized by nf_ct_alloc_hashtable().

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Use const in struct xt_match, xt_target, xt_table

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Supress some sparse warnings

  CHECK   net/netfilter/nf_conntrack_expect.c
net/netfilter/nf_conntrack_expect.c:429:13: warning: context imbalance in 'exp_seq_start' - wrong count at exit
net/netfilter/nf_conntrack_expect.c:441:13: warning: context imbalance in 'exp_seq_stop' - unexpected unlock
  CHECK   net/netfilter/nf_log.c
net/netfilter/nf_log.c:105:13: warning: context imbalance in 'seq_start' - wrong count at exit
net/netfilter/nf_log.c:125:13: warning: context imbalance in 'seq_stop' - unexpected unlock
  CHECK   net/netfilter/nfnetlink_queue.c
net/netfilter/nfnetlink_queue.c:363:7: warning: symbol 'size' shadows an earlier one
net/netfilter/nfnetlink_queue.c:217:9: originally declared here
net/netfilter/nfnetlink_queue.c:847:13: warning: context imbalance in 'seq_start' - wrong count at exit
net/netfilter/nfnetlink_queue.c:859:13: warning: context imbalance in 'seq_stop' - unexpected unlock

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[RAW]: Wrong content of the /proc/net/raw6.

The address of IPv6 raw sockets was shown in the wrong format, from
IPv4 ones.  The problem has been introduced by the commit
42a73808ed4f30b739eb52bcbb33a02fe62ceef5 ("[RAW]: Consolidate proc
interface.")

Thanks to Adrian Bunk who originally noticed the problem.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[RAW]: Cleanup IPv4 raw_seq_show.

There is no need to use 128 bytes on the stack at all. Clean the code
in the IPv6 style.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[RAW]: Family check in the /proc/net/raw[6] is extra.

Different hashtables are used for IPv6 and IPv4 raw sockets, so no
need to check the socket family in the iterator over hashtables. Clean
this out.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPCOMP]: Fix reception of incompressible packets

I made a silly typo by entering IPPROTO_IP (== 0) instead of
IPPROTO_IPIP (== 4).  This broke the reception of incompressible
packets.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: should explicitely initialize atomic_t field in struct dst_ops

All but one struct dst_ops static initializations miss explicit
initialization of entries field.

As this field is atomic_t, we should use ATOMIC_INIT(0), and not
rely on atomic_t implementation.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TCP]: NewReno must count every skb while marking losses

NewReno should add cnt per skb (as with FACK) instead of depending on
SACKED_ACKED bits which won't be set with it at all.  Effectively,
NewReno should always exists after the first iteration anyway (or
immediately if there's already head in lost_out.

This was fixed earlier in net-2.6.25 but got reverted among other
stuff and I didn't notice that this is still necessary (actually
wasn't even considering this case while trying to figure out the
reports because I lived with different kind of code than it in reality
was).

This should solve the WARN_ONs in TCP code that as a result of this
triggered multiple times in every place we check for this invariant.

Special thanks to Dave Young <hidave.darkstar@gmail.com> and Krishna
Kumar2 <krkumar2@in.ibm.com> for trying with my debug patches.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Tested-by: Dave Young <hidave.darkstar@gmail.com>
Tested-by: Krishna Kumar2 <krkumar2@in.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Fix race between put_net() and netlink_kernel_create().

The comment about "race free view of the set of network
namespaces" was a bit hasty. Look (there even can be only
one CPU, as discovered by Alexey Dobriyan and Denis Lunev):

put_net()
  if (atomic_dec_and_test(&net->refcnt))
    /* true */
      __put_net(net);
        queue_work(...);

/*
 * note: the net now has refcnt 0, but still in
 * the global list of net namespaces
 */

== re-schedule ==

register_pernet_subsys(&some_ops);
  register_pernet_operations(&some_ops);
    (*some_ops)->init(net);
      /*
       * we call netlink_kernel_create() here
       * in some places
       */
      netlink_kernel_create();
         sk_alloc();
            get_net(net); /* refcnt = 1 */
         /*
          * now we drop the net refcount not to
          * block the net namespace exit in the
          * future (or this can be done on the
          * error path)
          */
         put_net(sk->sk_net);
             if (atomic_dec_and_test(&...))
                   /*
                    * true. BOOOM! The net is
                    * scheduled for release twice
                    */

When thinking on this problem, I decided, that getting and
putting the net in init callback is wrong. If some init
callback needs to have a refcount-less reference on the struct
net, _it_ has to be careful himself, rather than relying on
the infrastructure to handle this correctly.

In case of netlink_kernel_create(), the problem is that the
sk_alloc() gets the given namespace, but passing the info
that we don't want to get it inside this call is too heavy.

Instead, I propose to crate the socket inside an init_net
namespace and then re-attach it to the desired one right
after the socket is created.

After doing this, we also have to be careful on error paths
not to drop the reference on the namespace, we didn't get
the one on.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: Denis Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[XFRM]: constify 'struct xfrm_type'

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Add missing initialization of nl_info.nl_net in rtm_to_fib6_config()

Add missing initialization of the new nl_info.nl_net field in
rtm_to_fib6_config(). This will be needed the store network namespace
associated to the fib6_config struct.

Signed-off-by: Benjamin Thery <benjamin.thery@bull.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Introducing socket mark socket option.

A userspace program may wish to set the mark for each packets its send
without using the netfilter MARK target. Changing the mark can be used
for mark based routing without netfilter or for packet filtering.

It requires CAP_NET_ADMIN capability.

Signed-off-by: Laszlo Attila Toth <panther@balabit.hu>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[AF_RXRPC]: constify function pointer tables

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Update version to 1.7.3.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Update firmware.

Update firmware to support programmable flow control.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Fine-tune flow control on 5709.

Make use of the programmable high/low water marks in 5709 for
802.3 flow control.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Remove CTX_WR macro.

The CTX_WR macro is unnecessary and obfuscates the code.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Remove REG_WR_IND/REG_RD_IND macros.

The REG_WR_IND/REG_RD_IND macros are unnecessary and obfuscate the
code.  Many callers to these macros read and write shared memory from
the bp->shmem_base, so we add 2 similar functions that automatically
add the shared memory base.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Refine tx coalescing setup.

Make the tx coalescing setup code independent of the MSIX vector.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2]: Fix 5706 serdes link down bug.

1. Correct the MII expansion serdes control register definition.
2. Check an additional RUDI_INVALID bit when determining 5706S link.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BLUETOOTH]: Add conn add/del workqueues to avoid connection fail.

The bluetooth hci_conn sysfs add/del executed in the default
workqueue.  If the del_conn is executed after the new add_conn with
same target, add_conn will failed with warning of "same kobject name".

Here add btaddconn & btdelconn workqueues, flush the btdelconn
workqueue in the add_conn function to avoid the issue.

Signed-off-by: Dave Young <hidave.darkstar@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPCOMP]: Fetch nexthdr before ipch is destroyed

When I moved the nexthdr setting out of IPComp I accidently moved
the reading of ipch->nexthdr after the decompression.  Unfortunately
this means that we'd be reading from a stale ipch pointer which
doesn't work very well.

This patch moves the reading up so that we get the correct nexthdr
value.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4] fib_trie: apply fixes from fib_hash

Update fib_trie with some fib_hash fixes:
- check for duplicate alternative routes for prefix+tos+priority when
replacing route
- properly insert by matching tos together with priority
- fix alias walking to use list_for_each_entry_continue for insertion
and deletion when fa_head is not NULL
- copy state from fa to new_fa on replace (not a problem for now)
- additionally, avoid replacement without error if new route is same,
as Joonwoo Park suggests.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4] fib: fix route replacement, fib_info is shared

fib_info can be shared by many route prefixes but we don't want
duplicate alternative routes for a prefix+tos+priority. Last change
was not correct to check fib_treeref because it accounts usage from
other prefixes. Additionally, avoid replacement without error if new
route is same, as Joonwoo Park suggests.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Fix miss of report unrecognized HMAC Algorithm parameter

This patch fix miss of check for report unrecognized HMAC Algorithm
parameter.  When AUTH is disabled, goto fall through path to report
unrecognized parameter, else, just break

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[INET_DIAG]: Fix inet_diag_lock_handler error path.

Fixes: http://bugzilla.kernel.org/show_bug.cgi?id=9825
The inet_diag_lock_handler function uses ERR_PTR to encode errors but
its callers were testing against NULL.

This only happens when the only inet_diag modular user, DCCP, is not
built into the kernel or available as a module.

Also there was a problem with not dropping the mutex lock when a handler
was not found, also fixed in this patch.

This caused an OOPS and ss would then hang on subsequent calls, as
&inet_diag_table_mutex was being left locked.

Thanks to spike at ml.yaroslavl.ru for report it after trying 'ss -d'
on a kernel that doesn't have DCCP available.

This bug was introduced in cset
d523a328fb0271e1a763e985a21f2488fd816e7e ("Fix inet_diag dead-lock
regression"), after 2.6.24-rc3, so just 2.6.24 seems to be affected.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[AUDIT]: Increase skb->truesize in audit_expand

The recent UDP patch exposed this bug in the audit code.  It
was calling pskb_expand_head without increasing skb->truesize.
The caller of pskb_expand_head needs to do so because that function
is designed to be called in places where truesize is already fixed
and therefore it doesn't update its value.

Because the audit system is using it in a place where the truesize
has not yet been fixed, it needs to update its value manually.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[INET]: Prevent out-of-sync truesize on ip_fragment slow path

When ip_fragment has to hit the slow path the value of skb->truesize
may go out of sync because we would have updated it without changing
the packet length.  This violates the constraints on truesize.

This patch postpones the update of skb->truesize to prevent this.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[AX25]: Kill ax25_bind() user triggable printk.

on the last run overlooked that sfuzz triggable message.
move the message to the corresponding comment.

Signed-off-by: maximilian attems <max@stro.at>
Signed-off-by: David S. Miller <davem@davemloft.net>

[AX25]: Beautify x25_init() version printk.

kill ref to old version and dup Linux.

Signed-off-by: maximilian attems <max@stro.at>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET] 9p: kill dead static inline buf_put_string

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Eric Van Hensbergen <ericvh@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SELinux]: Fix double free in selinux_netlbl_sock_setsid()

As pointed out by Adrian Bunk, commit
45c950e0f839fded922ebc0bfd59b1081cc71b70 ("fix memory leak in netlabel
code") caused a double-free when security_netlbl_sid_to_secattr()
fails.  This patch fixes this by removing the netlbl_secattr_destroy()
call from that function since we are already releasing the secattr
memory in selinux_netlbl_sock_setsid().

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>