After Florian patches, there is no need for XT_TABLE_INFO_SZ anymore :
Only one copy of table is kept, instead of one copy per cpu.

We also can avoid a dereference if we put table data right after
xt_table_info. It reduces register pressure and helps compiler.

Then, we attempt a kmalloc() if total size is under order-3 allocation,
to reduce TLB pressure, as in many cases, rules fit in 32 KB.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Jozsef Kadlecsik says:

====================
ipset patches for nf-next

Please consider to apply the next bunch of patches for ipset. First
comes the small changes, then the bugfixes and at the end the RCU
related patches.

* Use MSEC_PER_SEC consistently instead of the number.
* Use SET_WITH_*() helpers to test set extensions from Sergey Popovich.
* Check extensions attributes before getting extensions from Sergey Popovich.
* Permit CIDR equal to the host address CIDR in IPv6 from Sergey Popovich.
* Make sure we always return line number on batch in the case of error
  from Sergey Popovich.
* Check CIDR value only when attribute is given from Sergey Popovich.
* Fix cidr handling for hash:*net* types, reported by Jonathan Johnson.
* Fix parallel resizing and listing of the same set so that the original
  set is kept for the whole dumping.
* Make sure listing doesn't grab a set which is just being destroyed.
* Remove rbtree from ip_set_hash_netiface.c in order to introduce RCU.
* Replace rwlock_t with spinlock_t in "struct ip_set", change the locking
  in the core and simplifications in the timeout routines.
* Introduce RCU locking in bitmap:* types with a slight modification in the
  logic on how an element is added.
* Introduce RCU locking in hash:* types. This is the most complex part of
  the changes.
* Introduce RCU locking in list type where standard rculist is used.
* Fix coding styles reported by checkpatch.pl.
====================
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

According to the reporter, they are not needed.
Reported-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Standard rculist is used.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Three types of data need to be protected in the case of the hash types:

a. The hash buckets: standard rcu pointer operations are used.
b. The element blobs in the hash buckets are stored in an array and
   a bitmap is used for book-keeping to tell which elements in the array
   are used or free.
c. Networks per cidr values and the cidr values themselves are stored
   in fix sized arrays and need no protection. The values are modified
   in such an order that in the worst case an element testing is repeated
   once with the same cidr value.

The ipset hash approach uses arrays instead of lists and therefore is
incompatible with rhashtable.

Performance is tested by Jesper Dangaard Brouer:

Simple drop in FORWARD
~~~~~~~~~~~~~~~~~~~~~~

Dropping via simple iptables net-mask match::

 iptables -t raw -N simple || iptables -t raw -F simple
 iptables -t raw -I simple  -s 198.18.0.0/15 -j DROP
 iptables -t raw -D PREROUTING -j simple
 iptables -t raw -I PREROUTING -j simple

Drop performance in "raw": 11.3Mpps

Generator: sending 12.2Mpps (tx:12264083 pps)

Drop via original ipset in RAW table
~~~~~~~~~~~~~~~~~~~~~~~~~~~

Create a set with lots of elements::

 sudo ./ipset destroy test
 echo "create test hash:ip hashsize 65536" > test.set
 for x in `seq 0 255`; do
    for y in `seq 0 255`; do
        echo "add test 198.18.$x.$y" >> test.set
    done
 done
 sudo ./ipset restore < test.set

Dropping via ipset::

 iptables -t raw -F
 iptables -t raw -N net198 || iptables -t raw -F net198
 iptables -t raw -I net198 -m set --match-set test src -j DROP
 iptables -t raw -I PREROUTING -j net198

Drop performance in "raw" with ipset: 8Mpps

Perf report numbers ipset drop in "raw"::

 +   24.65%  ksoftirqd/1  [ip_set]           [k] ip_set_test
 -   21.42%  ksoftirqd/1  [kernel.kallsyms]  [k] _raw_read_lock_bh
    - _raw_read_lock_bh
       + 99.88% ip_set_test
 -   19.42%  ksoftirqd/1  [kernel.kallsyms]  [k] _raw_read_unlock_bh
    - _raw_read_unlock_bh
       + 99.72% ip_set_test
 +    4.31%  ksoftirqd/1  [ip_set_hash_ip]   [k] hash_ip4_kadt
 +    2.27%  ksoftirqd/1  [ixgbe]            [k] ixgbe_fetch_rx_buffer
 +    2.18%  ksoftirqd/1  [ip_tables]        [k] ipt_do_table
 +    1.81%  ksoftirqd/1  [ip_set_hash_ip]   [k] hash_ip4_test
 +    1.61%  ksoftirqd/1  [kernel.kallsyms]  [k] __netif_receive_skb_core
 +    1.44%  ksoftirqd/1  [kernel.kallsyms]  [k] build_skb
 +    1.42%  ksoftirqd/1  [kernel.kallsyms]  [k] ip_rcv
 +    1.36%  ksoftirqd/1  [kernel.kallsyms]  [k] __local_bh_enable_ip
 +    1.16%  ksoftirqd/1  [kernel.kallsyms]  [k] dev_gro_receive
 +    1.09%  ksoftirqd/1  [kernel.kallsyms]  [k] __rcu_read_unlock
 +    0.96%  ksoftirqd/1  [ixgbe]            [k] ixgbe_clean_rx_irq
 +    0.95%  ksoftirqd/1  [kernel.kallsyms]  [k] __netdev_alloc_frag
 +    0.88%  ksoftirqd/1  [kernel.kallsyms]  [k] kmem_cache_alloc
 +    0.87%  ksoftirqd/1  [xt_set]           [k] set_match_v3
 +    0.85%  ksoftirqd/1  [kernel.kallsyms]  [k] inet_gro_receive
 +    0.83%  ksoftirqd/1  [kernel.kallsyms]  [k] nf_iterate
 +    0.76%  ksoftirqd/1  [kernel.kallsyms]  [k] put_compound_page
 +    0.75%  ksoftirqd/1  [kernel.kallsyms]  [k] __rcu_read_lock

Drop via ipset in RAW table with RCU-locking
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

With RCU locking, the RW-lock is gone.

Drop performance in "raw" with ipset with RCU-locking: 11.3Mpps
Performance-tested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

There's nothing much required because the bitmap types use atomic
bit operations. However the logic of adding elements slightly changed:
first the MAC address updated (which is not atomic), then the element
activated (added). The extensions may call kfree_rcu() therefore we
call rcu_barrier() at module removal.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Replace rwlock_t with spinlock_t in "struct ip_set" and change the locking
accordingly. Convert the comment extension into an rcu-avare object. Also,
simplify the timeout routines.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Remove rbtree in order to introduce RCU instead of rwlock in ipset
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

There was a small window when all sets are destroyed and a concurrent
listing of all sets could grab a set which is just being destroyed.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

When elements added to a hash:* type of set and resizing triggered,
parallel listing could start to list the original set (before resizing)
and "continue" with listing the new set. Fix it by references and
using the original hash table for listing. Therefore the destroying of
the original hash table may happen from the resizing or listing functions.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Commit "Simplify cidr handling for hash:*net* types" broke the cidr
handling for the hash:*net* types when the sets were used by the SET
target: entries with invalid cidr values were added to the sets.
Reported by Jonathan Johnson.

Testsuite entry is added to verify the fix.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

There is no reason to check CIDR value regardless attribute
specifying CIDR is given.

Initialize cidr array in element structure on element structure
declaration to let more freedom to the compiler to optimize
initialization right before element structure is used.

Remove local variables cidr and cidr2 for netnet and netportnet
hashes as we do not use packed cidr value for such set types and
can store value directly in e.cidr[].
Signed-off-by: Sergey Popovich <popovich_sergei@mail.ua>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Even if we return with generic IPSET_ERR_PROTOCOL it is good idea
to return line number if we called in batch mode.

Moreover we are not always exiting with IPSET_ERR_PROTOCOL. For
example hash:ip,port,net may return IPSET_ERR_HASH_RANGE_UNSUPPORTED
or IPSET_ERR_INVALID_CIDR.
Signed-off-by: Sergey Popovich <popovich_sergei@mail.ua>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Permit userspace to supply CIDR length equal to the host address CIDR
length in netlink message. Prohibit any other CIDR length for IPv6
variant of the set.

Also return -IPSET_ERR_HASH_RANGE_UNSUPPORTED instead of generic
-IPSET_ERR_PROTOCOL in IPv6 variant of hash:ip,port,net when
IPSET_ATTR_IP_TO attribute is given.
Signed-off-by: Sergey Popovich <popovich_sergei@mail.ua>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Make all extensions attributes checks within ip_set_get_extensions()
and reduce number of duplicated code.
Signed-off-by: Sergey Popovich <popovich_sergei@mail.ua>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Signed-off-by: Sergey Popovich <popovich_sergei@mail.ua>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

We store the rule blob per (possible) cpu.  Unfortunately this means we can
waste lot of memory on big smp machines. ipt_entry structure ('rule head')
is 112 byte, so e.g. with maxcpu=64 one single rule eats
close to 8k RAM.

Since previous patch made counters percpu it appears there is nothing
left in the rule blob that needs to be percpu.

On my test system (144 possible cpus, 400k dummy rules) this
change saves close to 9 Gigabyte of RAM.
Reported-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

The binary arp/ip/ip6tables ruleset is stored per cpu.

The only reason left as to why we need percpu duplication are the rule
counters embedded into ipt_entry et al -- since each cpu has its own copy
of the rules, all counters can be lockless.

The downside is that the more cpus are supported, the more memory is
required.  Rules are not just duplicated per online cpu but for each
possible cpu, i.e. if maxcpu is 144, then rule is duplicated 144 times,
not for the e.g. 64 cores present.

To save some memory and also improve utilization of shared caches it
would be preferable to only store the rule blob once.

So we first need to separate counters and the rule blob.

Instead of using entry->counters, allocate this percpu and store the
percpu address in entry->counters.pcnt on CONFIG_SMP.

This change makes no sense as-is; it is merely an intermediate step to
remove the percpu duplication of the rule set in a followup patch.
Suggested-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reported-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

If bridge netfilter is used with both
bridge-nf-call-iptables and bridge-nf-filter-vlan-tagged enabled
then ip fragments in VLAN frames are sent without the vlan header.

This has never worked reliably.  Turns out this relied on pre-3.5
behaviour where skb frag_list was used to store ip fragments;
ip_fragment() then re-used these skbs.

But since commit 3cc49492
("ipv4: use skb coalescing in defragmentation") this is no longer
the case.  ip_do_fragment now needs to allocate new skbs, but these
don't contain the vlan tag information anymore.

Fix it by storing vlan information of the ressembled skb in the
br netfilter percpu frag area, and restore them for each of the
fragments.

Fixes: 3cc49492 ("ipv4: use skb coalescing in defragmentation")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

since commit d6b915e2
("ip_fragment: don't forward defragmented DF packet") the largest
fragment size is available in the IPCB.

Therefore we no longer need to care about 'encapsulation'
overhead of stripped PPPOE/VLAN headers since ip_do_fragment
doesn't use device mtu in such cases.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

IPv6 fragmented packets are not forwarded on an ethernet bridge
with netfilter ip6_tables loaded. e.g. steps to reproduce

1) create a simple bridge like this

        modprobe br_netfilter
        brctl addbr br0
        brctl addif br0 eth0
        brctl addif br0 eth2
        ifconfig eth0 up
        ifconfig eth2 up
        ifconfig br0 up

2) place a host with an IPv6 address on each side of the bridge

        set IPv6 address on host A:
        ip -6 addr add fd01:2345:6789:1::1/64 dev eth0

        set IPv6 address on host B:
        ip -6 addr add fd01:2345:6789:1::2/64 dev eth0

3) run a simple ping command on host A with packets > MTU

        ping6 -s 4000 fd01:2345:6789:1::2

4) wait some time and run e.g. "ip6tables -t nat -nvL" on the bridge

IPv6 fragmented packets traverse the bridge cleanly until somebody runs.
"ip6tables -t nat -nvL". As soon as it is run (and netfilter modules are
loaded) IPv6 fragmented packets do not traverse the bridge any more (you
see no more responses in ping's output).

After applying this patch IPv6 fragmented packets traverse the bridge
cleanly in above scenario.
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
[pablo@netfilter.org: small changes to br_nf_dev_queue_xmit]
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Prepare check_hbh_len() to be called from newly introduced
br_validate_ipv6() in next commit.
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

br_parse_ip_options() does not parse any IP options, it validates IP
packets as a whole and the function name is misleading.

Rename br_parse_ip_options() to br_validate_ipv4() and remove unneeded
commments.
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Currently frag_max_size is member of br_input_skb_cb and copied back and
forth using IPCB(skb) and BR_INPUT_SKB_CB(skb) each time it is changed or
used.

Attach frag_max_size to nf_bridge_info and set value in pre_routing and
forward functions. Use its value in forward and xmit functions.
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

IPv4 iptables allows to REDIRECT/DNAT/SNAT any traffic over a bridge.

e.g. REDIRECT
$ sysctl -w net.bridge.bridge-nf-call-iptables=1
$ iptables -t nat -A PREROUTING -p tcp -m tcp --dport 8080 \
  -j REDIRECT --to-ports 81

This does not work with ip6tables on a bridge in NAT66 scenario
because the REDIRECT/DNAT/SNAT is not correctly detected.

The bridge pre-routing (finish) netfilter hook has to check for a possible
redirect and then fix the destination mac address. This allows to use the
ip6tables rules for local REDIRECT/DNAT/SNAT REDIRECT similar to the IPv4
iptables version.

e.g. REDIRECT
$ sysctl -w net.bridge.bridge-nf-call-ip6tables=1
$ ip6tables -t nat -A PREROUTING -p tcp -m tcp --dport 8080 \
  -j REDIRECT --to-ports 81

This patch makes it possible to use IPv6 NAT66 on a bridge. It was tested
on a bridge with two interfaces using SNAT/DNAT NAT66 rules.
Reported-by: Artie Hamilton <artiemhamilton@yahoo.com>
Signed-off-by: Sven Eckelmann <sven@open-mesh.com>
[bernhard.thaler@wvnet.at: rebased, add indirect call to ip6_route_input()]
[bernhard.thaler@wvnet.at: rebased, split into separate patches]
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Put br_nf_pre_routing_finish_ipv6() after daddr_was_changed() and
br_nf_pre_routing_finish_bridge() to prepare calling these functions
from there.
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

use binary AND on complement of BRNF_NF_BRIDGE_PREROUTING to unset
bit in nf_bridge->mask.
Signed-off-by: Bernhard Thaler <bernhard.thaler@wvnet.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

After db29a950 ("netfilter: conntrack: disable generic tracking for
known protocols"), if the specific helper is built but not loaded
(a standard for most distributions) systems with a restrictive firewall
but weak configuration regarding netfilter modules to load, will
silently stop working.

This patch then puts a warning message so the sysadmin knows where to
start looking into. It's a pr_warn_once regardless of protocol itself
but it should be enough to give a hint on where to look.

Cc: Florian Westphal <fw@strlen.de>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Eric Dumazet says:

====================
tcp: defer shinfo->gso_size|type settings

We put shinfo->gso_segs in TCP_SKB_CB(skb) a while back for performance
reasons.

This was in commit cd7d8498 ("tcp: change tcp_skb_pcount() location")

This patch series complete the job for gso_size and gso_type, so that
we do not bring 2 extra cache lines in tcp write xmit fast path,
and making tcp_init_tso_segs() simpler and faster.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

We had various issues in the past when TCP stack was modifying
gso_size/gso_segs while clones were in flight.

Commit c52e2421 ("tcp: must unclone packets before mangling them")
fixed these bugs and added a WARN_ON_ONCE(skb_cloned(skb)); in
tcp_set_skb_tso_segs()

These bugs are now fixed, and because TCP stack now only sets
shinfo->gso_size|segs on the clone itself, the check can be removed.

As a result of this change, compiler inlines tcp_set_skb_tso_segs() in
tcp_init_tso_segs()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In commit cd7d8498 ("tcp: change tcp_skb_pcount() location") we stored
gso_segs in a temporary cache hot location.

This patch does the same for gso_size.

This allows to save 2 cache line misses in tcp xmit path for
the last packet that is considered but not sent because of
various conditions (cwnd, tso defer, receiver window, TSQ...)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_set_skb_tso_segs() & tcp_init_tso_segs() no longer
use the sock pointer.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Our goal is to touch skb_shinfo(skb) only when absolutely needed,
to avoid two cache line misses in TCP output path for last skb
that is considered but not sent because of various conditions
(cwnd, tso defer, receiver window, TSQ...)

A packet is GSO only when skb_shinfo(skb)->gso_size is not zero.

We can set skb_shinfo(skb)->gso_type to sk->sk_gso_type even for
non GSO packets.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_gso_segment() and tcp_gro_receive() are not strictly
part of TCP stack. They should not assume tcp_skb_mss(skb)
is in fact skb_shinfo(skb)->gso_size.

This will allow us to change tcp_skb_mss() in following patches.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix a BUG_ON() where CONFIG_NET_SWITCHDEV is set but the driver for a
bridged port does not support switchdev_port_attr_set op.  Don't BUG_ON()
if -EOPNOTSUPP is returned.

Also change BUG_ON() to netdev_err since this is a normal error path and
does not warrant the use of BUG_ON(), which is reserved for unrecoverable
errs.
Signed-off-by: Scott Feldman <sfeldma@gmail.com>
Reported-by: Brenden Blanco <bblanco@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Ivan Vecera says:

====================
bna: clean-up

The patches clean the bna driver.

v2: changes & comments requested by Joe
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

...and remove some of them. It is not necessary to log when .probe() and
.remove() are called or when TxQ is started or stopped. Also log level
of some of them was changed to more appropriate one (link up/down,
firmware loading failure.
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Timeout functions are defined with 'void *' ptr argument. They should
be defined directly with 'struct bfa_ioc *' type to avoid type conversions.
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>