When a software scan starts, it first sets sw_scanning, but
leaves the scan_channel "unset" (it currently actually gets
initialised to a default). Now, when something else tries
to (re)configure the hardware in the window between these two
events (after sw_scanning = true, but before scan_channel is
set), the current code switches to the (unset!) scan_channel.
This causes trouble, especially when switching bands and
sending frames on the wrong channel.

To work around this, leave scan_channel initialised to NULL
and use it to determine whether or not a switch to a different
channel should occur (and also use the same condition to check
whether to adjust power for scan or not).

Additionally, avoid reconfiguring the hardware completely when
recalculating idle resulted in no changes, this was the problem
that originally led us to discover the race condition in the
first place, which was helpfully bisected by Pavel. This part
of the patch should not be necessary with the other fixes, but
not calling the ieee80211_hw_config function when we know it to
be unnecessary is certainly a correct thing to do.

Unfortunately, this patch cannot and does not fix the race
condition completely, but due to the way the scan code is
structured it makes the particular problem Pavel discovered
(race while changing channel at the same time as transmitting
frames) go away. To fix it completely, more work especially
with locking configuration is needed.
Bisected-by: NPavel Roskin <proski@gnu.org>
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

NL80211_CMD_ASSOCIATE request must be able to indicate whether
management frame protection (IEEE 802.11w) is being used. mac80211 was
able to use MFP in client mode only with WEXT, but the new
NL80211_ATTR_USE_MFP attribute will allow this to be done with
nl80211, too.

Since we are currently using nl80211 for MFP only with drivers that
use user space SME, only MFP disabled and required values are
used. However, the NL80211_ATTR_USE_MFP attribute is an enum that can
be extended with MFP optional in the future, if that is needed with
some drivers (e.g., if the RSN IE is generated by the driver).
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Add missed checking of dev_addr_init return value in alloc_netdev_mq.
Signed-off-by: NJiri Pirko <jpirko@redhat.com>

 net/core/dev.c |   15 ++++++++++++---
 1 files changed, 12 insertions(+), 3 deletions(-)
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit ead2ceb0 ("Network Drop Monitor:
Adding kfree_skb_clean for non-drops and modifying end-of-line points
for skbs") established new conventions for identifying dropped packets.

Align skb_kill_datagram() with these conventions so that packets that
get dropped just before the copy to userspace are properly tracked.
Signed-off-by: NJohn Dykstra <john.dykstra1@gmail.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NAshish Karkare <akarkare@marvell.com>
Signed-off-by: NLennert Buytenhek <buytenh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

By separating the freeing code from the refcounting decrementing.
Probably reducing icache pressure when we still have reference counts to
go.
Signed-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit ac45f602 ("net: infrastructure
for hardware time stamping") added two skb initialization actions to
__alloc_skb(), which need to be added to skb_recycle_check() as well.
Signed-off-by: NLennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: NPatrick Ohly <patrick.ohly@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When no limit is given, the bfifo uses a default of tx_queue_len * mtu.
Packets handled by qdiscs include the link layer header, so this should
be taken into account, similar to what other qdiscs do.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a new wimax_dev is created, it's state has to be __WIMAX_ST_NULL
until wimax_dev_add() is succesfully called. This allows calls into
the stack that happen before said time to be rejected.

Until now, the state was being set (by mistake) to UNINITIALIZED,
which was allowing calls such as wimax_report_rfkill_hw() to go
through even when a call to wimax_dev_add() had failed; that was
causing an oops when touching uninitialized data.

This situation is normal when the device starts reporting state before
the whole initialization has been completed. It just has to be dealt
with.
Signed-off-by: NInaky Perez-Gonzalez <inaky@linux.intel.com>

When sending a message to user space using wimax_msg(), if nla_put()
fails, correctly interpret the return code from wimax_msg_alloc() as
an err ptr and return the error code instead of crashing (as it is
assuming than non-NULL means the pointer is ok).
Signed-off-by: NInaky Perez-Gonzalez <inaky@linux.intel.com>

We are currently processing block ack reordering as a separate task
before all other RX handlers. In theory, this is wrong since this step
should be done only after duplicate removal (see Figure 6-1 in IEEE
802.11n). However, moving this needs some work and the current
situation is not too bad. Add a comment here so that this small detail
does not get forgotten and who knows, maybe someone has some extra
time to take a look at cleaning this up.
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This patch allows skbs to be released from the RX reorder buffer in
case they have been there for an unexpectedly long time without us
having received the missing frames before them. Previously, these
frames were only released when the reorder window moved and that could
take very long time unless new frames were received constantly (e.g.,
TCP connections could be killed more or less indefinitely).

This situation should not happen very frequently, but it looks like
there are some scenarious that trigger it for some reason. As such,
this should be considered mostly a workaround to speed up recovery
from unexpected siutation that could result in connections hanging for
long periods of time.

The changes here will only check for timeout situation when adding new
RX frames to the reorder buffer. It does not handle all possible
cases, but seems to help for most cases that could result from common
network usage (e.g., TCP retrying at least couple of times). For more
completely coverage, a timer could be used to periodically check
whether there are any frames remaining in the reorder buffer if no new
frames are received.
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

No need to duplicate the same code in two places (and that would be
three after the followup patch).
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

net/mac80211/mlme.c:2079:28: warning: symbol 'ssid_len' shadows an earlier one
net/mac80211/mlme.c:2022:12: originally declared here

ssid_len is already being declared and checked above so there is
no need for it again.
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

It's not very helpful to see, in iwconfig, the current frequency
the card is tuned to if that frequency is currently somewhere
across the board because we're scanning. Since we keep track of
the frequency the user wants, display that instead.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Fixes sparse complaint:

  CHECK   net/wireless/nl80211.c
net/wireless/nl80211.c:3694:6:
	warning: symbol 'nl80211_send_mlme_timeout'
	was not declared. Should it be static?
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

When we aren't doing anything in mac80211, we can turn off
much of the hardware, depending on the driver/hw. Not doing
anything, aka being idle, means:

 * no monitor interfaces
 * no AP/mesh/wds interfaces
 * any station interfaces are in DISABLED state
 * any IBSS interfaces aren't trying to be in a network
 * we aren't trying to scan

By creating a new function that verifies these conditions and calling
it at strategic points where the states of those conditions change,
we can easily make mac80211 tell the driver when we are idle to save
power.

Additionally, this fixes a small quirk where a recalculated powersave
state is passed to the driver even if the hardware is about to stopped
completely.

This patch intentionally doesn't touch radio_enabled because that is
currently implemented to be a soft rfkill which is inappropriate here
when we need to be able to wake up with low latency.

One thing I'm not entirely sure about is this:

  phy0: device no longer idle - in use
  wlan0: direct probe to AP 00:11:24:91:07:4d try 1
  wlan0 direct probe responded
  wlan0: authenticate with AP 00:11:24:91:07:4d
  wlan0: authenticated
> phy0: device now idle
> phy0: device no longer idle - in use
  wlan0: associate with AP 00:11:24:91:07:4d
  wlan0: RX AssocResp from 00:11:24:91:07:4d (capab=0x401 status=0 aid=1)
  wlan0: associated

Is it appropriate to go into idle state for a short time when we have
just authenticated, but not associated yet? This happens only with the
userspace SME, because we cannot really know how long it will wait
before asking us to associate. Would going idle after a short timeout
be more appropriate? We may need to revisit this, depending on what
happens.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

To deter future rate scaling algorithm writers from requesting NO_ACK
packets to be retried, throw a WARN_ON_ONCE if the algorithm hands us
a try count over 1 for NO_ACK packet.
Signed-off-by: NGábor Stefanik <netrolller.3d@gmail.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Make PID check for IEEE80211_TX_CTL_NO_ACK instead of
is_multicast_ether_addr when determining whether to use the lowest
rate, and set the retry count to 0 (total try count = 1) if
IEEE80211_TX_CTL_NO_ACK is set.
Signed-off-by: NGábor Stefanik <netrolller.3d@gmail.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Make the retry count zero (total try count = 1) for frames with
IEEE80211_TX_CTL_NO_ACK set.

Also remove the check for is_multicast_ether_addr in use_low_rate,
which is redundant because all multicasts have IEEE80211_TX_CTL_NO_ACK
set.
Signed-off-by: NGábor Stefanik <netrolller.3d@gmail.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

If the IE buffer was allocated, the pub.information_elements pointer
was also changed to the allocated space. So we must not assume anymore
that the pointer points at the "found" tail.

So if it was allocated previously, take the codebranch that grows the
buffer size (if necessary) and put the data into the allocated buffer.
Signed-off-by: NMichael Buesch <mb@bu3sch.de>
Acked-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This removes an unnecessary ksize() call. krealloc() will do this
test internally and won't perform any allocation if the space is
already sufficient to hold the data.
So remove the redundant check.
Signed-off-by: NMichael Buesch <mb@bu3sch.de>
Reviewed-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

net/rfkill/rfkill.c: In function 'update_rfkill_state':
net/rfkill/rfkill.c:99: error: implicit declaration of function 'rfkill_led_trigger'

Caused by

: commit 492301fb
: Author: Larry Finger <Larry.Finger@lwfinger.net>
: Date:   Thu Apr 9 22:14:19 2009 -0500
:
:     rfkill: Fix broken rfkill LED in 2.6.30-rc1

Cc: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: Ivo van Doorn <IvDoorn@gmail.com>
Cc: Michael Buesch <mb@bu3sch.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Commit 0ad8acaf "cfg80211: fix NULL pointer deference in
reg_device_remove()" added a check that last_request is non-NULL,
rendering the 2nd check superfluous.  While there, rearrange the code a
bit so it's a little more straight forward.
Signed-off-by: NChris Wright <chrisw@sous-sol.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Due to the use of a _REQ_DIRECT_PROBE bit, which is
unnecessary (and I wonder why it was done that way),
an interesting situation can arise:
 1) we try to probe an access point
 2) the AP doesn't response in time
 3) we tell userspace that we gave up
 4) the AP suddenly responds
 5) we auth/assoc with the AP

I've seen 4) happen in testing with hostapd SIGSTOPped,
and when SIGCONTinued it processes the probe requests
that came in and send responses. But 5) is not supposed
to happen after we tell everybody we've given up on the
AP.

To fix this, remove the _REQ_DIRECT_PROBE request bit,
and process probe responses when we're in the relevant
MLME state, namely IEEE80211_STA_MLME_DIRECT_PROBE.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

If the direct probe times out, we need to send the authentication
timeout event to notify SME in the same way as we notify on timeout
with authentication frames since the direct probe is run as part of
the authentication attempt.
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

In order to later add tracing or verifications to the driver
calls mac80211 makes, this patch adds static inline wrappers
for all operations.

All calls are now written as

	drv_<op>(local, ...);

instead of

	local->ops-><op>(&local->hw, ...);

Where necessary, the wrappers also do existence checking and
return default values as appropriate.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The config_interface method is a little strange, it contains the
BSSID and beacon updates, while bss_info_changed contains most
other BSS information for each interface. This patch removes
config_interface and rolls all the information it previously
passed to drivers into bss_info_changed.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

We currently have two beacon interval configuration knobs:
hw.conf.beacon_int and vif.bss_info.beacon_int. This is
rather confusing, even though the former is used when we
beacon ourselves and the latter when we are associated to
an AP.

This just deprecates the hw.conf.beacon_int setting in favour
of always using vif.bss_info.beacon_int. Since it touches all
the beaconing IBSS code anyway, we can also add support for
the cfg80211 IBSS beacon interval configuration easily.

NOTE: The hw.conf.beacon_int setting is retained for now due
      to drivers still using it -- I couldn't untangle all
      drivers, some are updated in this patch.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

There are some places marked
	/* XXX maybe racy? */
and they really are racy because there's no locking.

This patch reworks much of the scan code, and introduces proper
locking for the scan request as well as the internal scanning
(which is necessary for IBSS/managed modes). Helper functions
are added to call the scanning code whenever necessary. The
scan deferring is changed to simply queue the scanning work
instead of trying to start the scan in place, the scanning work
will then take care of the rest.

Also, currently when internal scans are requested for an interface
that is trying to associate, we reject such scans. This was not
intended, the mlme code has provisions to scan twice when it can't
find the BSS to associate with right away; this has never worked
properly. Fix this by not rejecting internal scan requests for an
interface that is associating.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

When the IBSS code wants to scan, but that fails, we can
get stuck in a situation where you can never scan again.
Fix this by properly notifying ourselves when the scan
request has failed.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Kalle points out that max_sleep_interval is somewhat confusing
because the value is measured in beacon intervals, and not in
TU. Rename it to max_sleep_period to be consistent with things
like DTIM period that are also measured in beacon intervals.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

When somebody changes the PS parameters while scanning
is in progress, we enable PS -- during the scan. This
is clearly not desirable, and we can just abort enabling
PS when scanning since when the scan finishes it will
be taken care of.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Reviewed-by: NKalle Valo <kalle.valo@iki.fi>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

There are a few problems in the IBSS code:
 a) it tries to activate interfaces that are down after scanning
 b) it crashes after scanning on an IBSS iface that isn't active
 c) since the ssid_len is used as a flag, need to make it visible
    only after all other settings are set, this helps protect
    against b)

For b), we get a system crash:

wlan0: Creating new IBSS network, BSSID ce:f9:88:76:1e:4d
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<...>] ieee80211_sta_find_ibss+0x294/0x37d [mac80211]
Call Trace:
 [<...>] ieee80211_ibss_notify_scan_completed+0x0/0x88 [mac80211]
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Setting the name of a sysfs device has to be done in a context that can
actually sleep. It allocates its memory with GFP_KERNEL. Previously it
was a static (size limited) string and that got changed to accommodate
longer device names. So move the dev_set_name() just before calling
device_add() which is executed in a work queue.

This fixes the following error:

[  110.012125] BUG: sleeping function called from invalid context at mm/slub.c:1595
[  110.012135] in_atomic(): 1, irqs_disabled(): 0, pid: 0, name: swapper
[  110.012141] 2 locks held by swapper/0:
[  110.012145]  #0:  (hci_task_lock){++.-.+}, at: [<ffffffffa01f822f>] hci_rx_task+0x2f/0x2d0 [bluetooth]
[  110.012173]  #1:  (&hdev->lock){+.-.+.}, at: [<ffffffffa01fb9e2>] hci_event_packet+0x72/0x25c0 [bluetooth]
[  110.012198] Pid: 0, comm: swapper Tainted: G        W 2.6.30-rc4-g953cdaa #1
[  110.012203] Call Trace:
[  110.012207]  <IRQ>  [<ffffffff8023eabd>] __might_sleep+0x14d/0x170
[  110.012228]  [<ffffffff802cfbe1>] __kmalloc+0x111/0x170
[  110.012239]  [<ffffffff803c2094>] kvasprintf+0x64/0xb0
[  110.012248]  [<ffffffff803b7a5b>] kobject_set_name_vargs+0x3b/0xa0
[  110.012257]  [<ffffffff80465326>] dev_set_name+0x76/0xa0
[  110.012273]  [<ffffffffa01fb9e2>] ? hci_event_packet+0x72/0x25c0 [bluetooth]
[  110.012289]  [<ffffffffa01ffc1d>] hci_conn_add_sysfs+0x3d/0x70 [bluetooth]
[  110.012303]  [<ffffffffa01fba2c>] hci_event_packet+0xbc/0x25c0 [bluetooth]
[  110.012312]  [<ffffffff80516eb0>] ? sock_def_readable+0x80/0xa0
[  110.012328]  [<ffffffffa01fee0c>] ? hci_send_to_sock+0xfc/0x1c0 [bluetooth]
[  110.012343]  [<ffffffff80516eb0>] ? sock_def_readable+0x80/0xa0
[  110.012347]  [<ffffffff805e88c5>] ? _read_unlock+0x75/0x80
[  110.012354]  [<ffffffffa01fee0c>] ? hci_send_to_sock+0xfc/0x1c0 [bluetooth]
[  110.012360]  [<ffffffffa01f8403>] hci_rx_task+0x203/0x2d0 [bluetooth]
[  110.012365]  [<ffffffff80250ab5>] tasklet_action+0xb5/0x160
[  110.012369]  [<ffffffff8025116c>] __do_softirq+0x9c/0x150
[  110.012372]  [<ffffffff805e850f>] ? _spin_unlock+0x3f/0x80
[  110.012376]  [<ffffffff8020cbbc>] call_softirq+0x1c/0x30
[  110.012380]  [<ffffffff8020f01d>] do_softirq+0x8d/0xe0
[  110.012383]  [<ffffffff80250df5>] irq_exit+0xc5/0xe0
[  110.012386]  [<ffffffff8020e71d>] do_IRQ+0x9d/0x120
[  110.012389]  [<ffffffff8020c3d3>] ret_from_intr+0x0/0xf
[  110.012391]  <EOI>  [<ffffffff80431832>] ? acpi_idle_enter_bm+0x264/0x2a6
[  110.012399]  [<ffffffff80431828>] ? acpi_idle_enter_bm+0x25a/0x2a6
[  110.012403]  [<ffffffff804f50d5>] ? cpuidle_idle_call+0xc5/0x130
[  110.012407]  [<ffffffff8020a4b4>] ? cpu_idle+0xc4/0x130
[  110.012411]  [<ffffffff805d2268>] ? rest_init+0x88/0xb0
[  110.012416]  [<ffffffff807e2fbd>] ? start_kernel+0x3b5/0x412
[  110.012420]  [<ffffffff807e2281>] ? x86_64_start_reservations+0x91/0xb5
[  110.012424]  [<ffffffff807e2394>] ? x86_64_start_kernel+0xef/0x11b

Based on a report by Davide Pesavento <davidepesa@gmail.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>
Tested-by: NHugo Mildenberger <hugo.mildenberger@namir.de>
Tested-by: NBing Zhao <bzhao@marvell.com>

Signed-off-by: Shan Wei<shanwei@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

v5 -> v6 (current):
-removed so far unused static functions
-corrected dev_addr_del_multiple to call del instead of add

v4 -> v5:
-added device address type (suggested by davem)
-removed refcounting (better to have simplier code then safe potentially few
 bytes)

v3 -> v4:
-changed kzalloc to kmalloc in __hw_addr_add_ii()
-ASSERT_RTNL() avoided in dev_addr_flush() and dev_addr_init()

v2 -> v3:
-removed unnecessary rcu read locking
-moved dev_addr_flush() calling to ensure no null dereference of dev_addr

v1 -> v2:
-added forgotten ASSERT_RTNL to dev_addr_init and dev_addr_flush
-removed unnecessary rcu_read locking in dev_addr_init
-use compare_ether_addr_64bits instead of compare_ether_addr
-use L1_CACHE_BYTES as size for allocating struct netdev_hw_addr
-use call_rcu instead of rcu_synchronize
-moved is_etherdev_addr into __KERNEL__ ifdef

This patch introduces a new list in struct net_device and brings a set of
functions to handle the work with device address list. The list is a replacement
for the original dev_addr field and because in some situations there is need to
carry several device addresses with the net device. To be backward compatible,
dev_addr is made to point to the first member of the list so original drivers
sees no difference.
Signed-off-by: NJiri Pirko <jpirko@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes the wrong message type that are triggered by
user updates, the following commands:

(term1)# conntrack -I -p tcp -s 1.1.1.1 -d 2.2.2.2 -t 10 --sport 10 --dport 20 --state LISTEN
(term1)# conntrack -U -p tcp -s 1.1.1.1 -d 2.2.2.2 -t 10 --sport 10 --dport 20 --state SYN_SENT
(term1)# conntrack -U -p tcp -s 1.1.1.1 -d 2.2.2.2 -t 10 --sport 10 --dport 20 --state SYN_RECV

only trigger event message of type NEW, when only the first is NEW
while others should be UPDATE.

(term2)# conntrack -E
    [NEW] tcp      6 10 LISTEN src=1.1.1.1 dst=2.2.2.2 sport=10 dport=20 [UNREPLIED] src=2.2.2.2 dst=1.1.1.1 sport=20 dport=10 mark=0
    [NEW] tcp      6 10 SYN_SENT src=1.1.1.1 dst=2.2.2.2 sport=10 dport=20 [UNREPLIED] src=2.2.2.2 dst=1.1.1.1 sport=20 dport=10 mark=0
    [NEW] tcp      6 10 SYN_RECV src=1.1.1.1 dst=2.2.2.2 sport=10 dport=20 [UNREPLIED] src=2.2.2.2 dst=1.1.1.1 sport=20 dport=10 mark=0

This patch also removes IPCT_REFRESH from the bitmask since it is
not of any use.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

This patch fixes a problem when you use 32 nodes in the cluster
match:

% iptables -I PREROUTING -t mangle -i eth0 -m cluster \
  --cluster-total-nodes  32  --cluster-local-node  32 \
  --cluster-hash-seed 0xdeadbeef -j MARK --set-mark 0xffff
iptables: Invalid argument. Run `dmesg' for more information.
% dmesg | tail -1
xt_cluster: this node mask cannot be higher than the total number of nodes

The problem is related to this checking:

if (info->node_mask >= (1 << info->total_nodes)) {
	printk(KERN_ERR "xt_cluster: this node mask cannot be "
			"higher than the total number of nodes\n");
	return false;
}

(1 << 32) is 1. Thus, the checking fails.

BTW, I said this before but I insist: I have only tested the cluster
match with 2 nodes getting ~45% extra performance in an active-active setup.
The maximum limit of 32 nodes is still completely arbitrary. I'd really
appreciate if people that have more nodes in their setups let me know.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: NPatrick McHardy <kaber@trash.net>